Research how to build a 4 quadrant bench power supply, aka “poor man’s source measure unit”
NOTE: This document (currently) only contains unstructured research material mainly for my own use. Might also be useful for others.
Disclaimer: I’m an electronics newbie. Findings, solutions, suggestions, conclusions made in this document should not be viewed as accurate or even correct. They are mainly for my own use, learning and reference.
Refer to section About for some info on why I’ve started this project.
(Updated June 2020) You might also read a bit about my first prototype based on parts of the research gathered here.
My four quadrant endeavour started out with a LT1970 Power Op Amp (which was the closest I find for a IC supporting four quadrants), tried to learn how it works, identify it’s limitations and see if I can create a more “discrete” circuit without all of it’s limitation.
I’ll try to keep it as simple as possible. I want to understand each part of the design. I’ll rather replace multiple components with a single component even if it’s more expensive. 2Hopefully I won’t need too much advanced analog wizardry. Initially, some parts might end up as overkill. Some parts might unstable/inaccurate and not the ideal for the job. Let’s find out… It’s part of the fun… Main goal is to learn about:
- High precision analog circuits
- How to achieve high accuracy
- Stability (also long term)
- High current (>500mA) using discrete output stages
- PCB design
- Combo of software and hardware
Relevant power supply related projects
There are a lot of power supply projects out there. Mostly single quadrant variants, but there is a lot to learn from them.
http://www.djerickson.com/diy_smu/ A really nice project aimed at designing a DIY SMU ! The webpage explains the design decisions made and compares them with the design of the Keithley 236. Impressive stuff and is a must-read !
EEZ H24005 BENCH POWER SUPPLY (Arduino Shield for programmable DIY bench power supply with SCPI support). Tons of useful info here. It’s a must-read, even if it’s not four quadrant.
Lots of design discussions on eevblog forum: DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005) From arduini.cc: https://create.arduino.cc/projecthub/prasimix/diy-programmable-scpi-bench-power-supply-5e59d5.
Or take a look at the newer EEZ-BB3 which is a newer modular design.
Here is the part of the schematics for the control loop for an output power module:
http://www.djerickson.com/ps-load/ Power Supply / Load or 2/4 quadrant voltage current source/meter design proposal
http://wahz.blogspot.nl – lab power supply Tom Biskupic goes through his power supply project from A to Z, learnings and mistakes. Very useful! Code etc. on github. “What I really wanted was one of those Rigol DP832. Of course I could buy one but given I an am using this supply to learn about electronics it seemed a perfect opportunity to build one!… “
https://gerrysweeney.com/fully-programmable-modular-bench-power-supply/ A lot of interesting information/discussions where he in more that 10 parts goes through the design of his power supply. “…back to the regulator stability topic, I have removed all the capitative loading around the driver and output circuit and instead reduced the bandwidth of the control loop by including frequency-dependnat negative feedback on both the voltage and current error amps. The aim was to significantly lower the gain of the loop above the point where the feedback becomes positive and the servo action becomes unstable…. “. In part 10 he changes the error amplifier / output stage completely and explains why: https://youtu.be/xDMpx0s6BwM?t=399
Dave Jones uSupply
Dave jones uSupply, schematics (Rev B , Rev C). Current control uses filter 1uF/ 330R = 480Hz? at the output of the current sense (MAX4080T) which helps reducing noise. (Rev C replaces MAX4080 with “discrete” op amp diff amp and output filter changed to 0.1uF / 1K =15KHz? cutoff) Also uses a transistor to draw the output low in case of current limit. Instead of output from op-amp and diode.
Both revisions also use filter in the input of the voltage regulator (22uF) to “stabilize” the circuit.
Rev C is an update of rev B. It basically saves a lot of components (read: cost). For example: Replacing uCurrent and MAX4080 with INA219 and discrete op-amp based differential amplifier, replaced AD/DA to save cost and pins (more info in EEVblog #259 – PSU Rev C Schematic – Part 12 ).Now powered by only two li-ion batteries. Downside is a slight reduction in precision (10bit).
Variant of Daves uSupply: http://www.instructables.com/id/Digital-Battery-Operated-Powersupply/
(Note also that Ian has designed (and sell) a precision programmable voltage reference Handheld Precision Digital Voltage Source – PDVS2) that can be a very useful tool.
The Modular Bench Power Supply ++, The Essential DIY Build for Every EE Student and Old Timer alike… “We will start with the basics, a backgrounder on the modules purpose and basic functionality, the simpler analogue design and accurate enough for basic projects, this will help you understand Voltage references, OP-AMPS, feedback loops, Current shunts both high side and low side and the benefits of each. Later, once we have the basic PSU working, we will upgrade modules to improve the regulation, accuracy and then some form of digital control in the form of a micro-controller, at this point we will discuss ADC, DAC resolution, over sampling etc“
One of the interesting topics he talks about is how he designed the supply to let the control circuity float with regard to the output in #10 part 1 @ 22:08 . I’m not sure I’m able to wrap my head around that one yet, but I have a feeling that it’s a good solution.
Development of the NA-01 Lab Nutrition
Some limitation of lt1970 is that it cannot set current limits lower that apprx. 4mA. The current limit in general is only within 1%. Output current is max 500mA but requires careful PCB layout to dissepate heat. By adding output stage current can be 5A, but the lowest current limit will then be apprx. 40mA with a 0.1ohm shunt resistance.
It’s absolute current limitation seems to vary with load (must be verified). However, it’s a good candidate for a prototype, or a “VERY poor man’s smu”…
4 quadrant power supply example from: EDN (2004):
EasySMU is a single-channel ±12V/±40mA programmable-voltage/programmable-current source with accurate voltage/current measurement capability. DC2591A shows a demo application that uses 1970a even though the primary purpose of the EasySMU is to demonstrate the LTC4316 I2C address translator. “EasySMU is a single-channel ±12V/±40mA programmable-voltage/programmable-current source with accurate voltage/current measurement capability“
An interesting note about the gain resistor network: “Since the LT1970A power dissipation varies based on loading and output settings, most discrete resistors would exhibit mismatch as the PCB temperature gradients change. The LT5400-3 matched resistor network eliminates this mismatch.”
Read my test here
DC2132A is a high performance, compact, efficient DC bench supply
“The LT3081’s unique current-source reference and voltage-follower output amplifier make it possible to connect two linear regulators in parallel for up to 3A and over 24V of adjustable current and voltage output control. Linear regulators at the output suppress output ripple without requiring large output capacitors, resulting in a truly flat DC output and small size.” Also look at the discussion here: https://www.eevblog.com/forum/projects/linear-technology-dc2132a-cvcc-adj-bench-power-supply-board/
Other relevant power supply discussions
https://www.eevblog.com/forum/projects/anything-wrong-with-this-linear-psu-design/: “…Using diodes as a MIN function the same way I did, and (now) using a transistor to ensure the output of the op amp does not go too high when not controlling the feedback loop… … … I changed out the LT1007 / 1037 ( even though it isn’t the ideal choice in this design) and got discusting parsitic oscillation, how am I supposed to stop this? How do I go about frequency compansation?… … …I used LT1007 in this design, and I was able to get the LT1037 stable in it as well, although it required more capacitance between the output and inverting input. I don’t like that form of compensation because it is essentially slowing the op amp down, reducing bandwidth. I figure the LT1007 is better suited…. … …But just so you know, Rigol DP832 has 1000uF output capacitance, and it counts as a really good use-able power supply, so your supply is already much better and probably fast enough for 99% of the things …”
Relevant commercial power supplies (4 quadrant, 2quadrant, SMUs and others)
The subsections below is a collection of relevant documentation for several commercial available supplies, focusing mainly on devices that operates in 2 and 4 quadrants.
Agilent Model 66312A Dynamic Measurement DC Source
Inside a Two-Quadrant Power Supply – Agilent 66312A Teardown and Experiment “…is a two-quadrant power supply, it not only can source up to 2A of current between 0 and 20V, but also can sink up to 1.2A or 60% of its rated output current as well. Although lacking some key functionality of a source measure unit(SMU), Agilent 66312A can nevertheless be used in similar situations where both current sourcing and sinking capabilities are needed.“
Service manual: https://literature.cdn.keysight.com/litweb/pdf/5962-0874.pdf
Note! I planned to purchase one of these for testing/comparing, but if you look closer at the datasheet, there is a footnote saying “The sink current does not track the programmed current“. I guess this means that it cannot be used as a programmable load (programmable sink). So for a proper two quadrant use, the 66332/663x is better.
Keysight 66332A (Dynamic Measurement DC Source) and 663xB
This series of linear-regulated 80-100 W DC power supplies is designed to maximize the throughput of DUTs through the manufacturing test process. Both programming and measurement are optimized for speed.
These are power supply that also contains “active downprogrammer [that] can sink up to the full rated current of the power supply, which quickly brings the power supply output to zero volts.“. Service manual Keysight 66332A Dynamic Measurement DC Source and Keysight Models 6632B, 6633B, 6634B System DC Power Supply
The user manual states: “…dc source is capable of sourcing as well as sinking current over the output voltage range from zero volts to the rated maximum. The negative quadrant is identical to the positive quadrant. However, the negative current cannot be set independently; it tracks the value programmed for the positive current. Thus, if the positive current is set to 1 A, the negative current is also set to 1 A“. Note that HP/Agilent has a lot of 66xxx power supplies, but it looks like you have to carefully check the manual to find out if it has programmable sink capability. It looks like its the 66332/663x are the ones. Note, however, that the minimum programmable sink current is -5mA (at least for the 20V versions 66332 and 6632)
Principle of operation:
“The CV/CC control circuits provide a CV control loop, a positive CC control loop, and a negative CC control loop. For any value of load resistance, the supply must act either as a constant voltage (CV) or as a constant current (CC) supply. Transfer between these modes is accomplished automatically by the CV/CC control circuit at a value of load resistance equal to the ratio of the programmed voltage value to the programmed current value. The negative CC control circuit is activated when a current source such as another power supply is connected across the output terminals and its voltage is greater than the programmed voltage. A low level CV_Detect*, CC_Detect*, or CCN_Detect* signal is returned to the secondary interface to indicate that the corresponding mode is in effect.
When the CV loop is in control, diode D328 is conducting current. Voltage regulation is accomplished by comparing the programmed voltage signal CV_Prog with the output voltage monitor signal Vmon. The Vmon signal is in the 0 to +5 V range, which corresponds to the zero to full-scale output voltage range of the supply. If the output voltage exceeds the programmed voltage, Vmon goes high and produces a more negative-going CV signal, which reduces the input to the voltage gain stage and lowers the output voltage. Conversely, if the output voltage is less than the programmed voltage, Vmon goes low and produces a more positive-going CV signal, which increases the input to the voltage gain stage and raises the output voltage. Depending upon the position of the sense switch, the output voltage is either monitored at the supply’s output terminals (local), or at the load (remote) using the +S and -S terminals with remote sense leads connected to the load. If the output voltage goes higher than the programmed value, the unit starts sinking current to reduce the output voltage.
When the CC loop is in control, diode D325 is conducting current. Current regulation is accomplished by comparing the programmed current signal CC_Prog with the output current monitor signal Imon_H. The Imon_H signal is produced by measuring the voltage drop across the current monitoring resistor and is in the 0 to +5 V range, which
corresponds to the zero to full-scale output current range of the supply. If the output current exceeds the programmed current, Imon_H goes high and produces a more negative going CC signal, which reduces the input to the voltage gain stage and lowers the output current. Conversely, if the output current is less than the programmed
current, Imon_H goes low and produces a more positive-going CC signal, which increases the input to the voltage gain stage and raises the output current.
When the supply is sinking current, only the CV control circuit or the CCN control circuit can be active. In this case, the supply is acting as a load instead of a power source and will attempt to pull the output voltage down by drawing off current from the externally applied source. The current that will be drawn from the externally supplied source is
determined by the CC_Prog signal. When the current required to reduce the voltage is less than the programmed current value, the CV control circuit is active and regulates the output voltage. When the current required to reduce the voltage exceeds the programmed current value, the CCN control circuit is active. It regulates the output current by comparing the negative Imon_H signal to the inverted CC_Prog signal.“
The service manual also contains full schematics which is very nice. Here is an picture on how the 6632A looks inside, copied from a eevblog 6632A teardown thread.
Keithley 228A Programmable Voltage/Current Source
The Keithley 228A is capable of bipolar source or sink (4 Quadrant Operation) up to a full 100 watts without derating, permitting it to act as a voltage or current supply or as an active load.
The control loop is at the left hand side of the schematics. “U103A supplies a signal which drives the output in the polarity programmed by the operator. Comparators U106 through U109 supply a signal which balances the signal from U103A when the output approaches a programmed limit. The *voltage limits are set with a zero to +l.OlOV signal from the voltage DAC (Vdac) UllO. The acronymn DAC means Digital to Analog Converter. The 4current limits are set with a zero to + l.OlOV signal from the current DAC (Idac) Ulll. The positive limits (U107 and UlO9) are sensed by comparing the (DAC voltage) with the (feedback voltage) and attempting to keep feedback voltage less than or equal to the DAC voltage. The negative limits (U106 and U108) are sensed by comparing the (DAC voltage + feedback voltage) with ground and attempting to keep (DAC voltage +feedback voltagekgreater than or equal to zero or – (feedback voltage) less than or equal to (DAC voltage). Refer to Figures 6-3 and 6-4. “
Seems to be using a summing amplifer from voltage and current measurement amplifers. Not using the diode-or concept I’ve seen in a lot of other devices? Investigate…
Keithley 236 / 237 / 238
The 236, 237, and 238 Source-Measure Units (SMU) are fully programmable instruments, capable of sourcing and measuring voltage or current simultaneously. Link to Operators manual. There are several versions of the service manual out there. The Service manual on Keithleys website does not contain full schematics. Alternative with minor differences: http://exodus.poly.edu/~kurt/manuals/manuals/Keithley/KEI%20238%20Service.pdf (a bit different block diagram for output stage). If you want one that contains schematics: http://www.ko4bb.com/getsimple/index.php?id=download&file=06_Misc_Test_Equipment/Keithley/Keithley_236_237_Source_Measure_Unit_SMU_Service_Manual_and_schematics.pdf (thanks: https://www.eevblog.com/forum/testgear/keitley-236-teardown-and-review/ )
Block diagram (from the service manual) of the control loop:
Description from service manual: “Programming current and voltage sets the output voltage of the two digital-to-analog (DAC) circuits. Program ming current controls the output of the I DAC (U23 and U22), and programming voltage controls the output of the V DAC (U25 and U24). Programming current or voltage for zero output will result with a 0V output from the respective DAC. Programming for a full scale output will result with a -40V output from the respective DAC.
The output voltage from the I DAC is applied to current clamps through resistor networks. Op amp U13 and diode CR11 form the negative current clamp (-I CLAMP). The output from the I DAC is inverted by the xl amplifier U50. Op amp U17 and diode CR10 form the positive cur rent clamp (+I CLAMP).
The output from the V DAC is inverted by the x1 amplifier U12 (xO,1 for the 1.1V range) and similarly applied to current clamps through resistor networks. Op amp U15A and diode CR12 form the positive voltage clamp (+V CLAMP). The inverted output of U12 is again inverted by U19. Op amp U15B and diode CR9 form the negative voltage clamp (-V CLAMP).
During operation, only one of the four precision clamps will be on at one time to control the error amplifier (U14). The controlling function and the programmed polarity (+ or – ) will determine which clamp is on.”…
DAC used are 14 bit current output AD7538 from Analog Device (I’m a bit surprised it’s only 14 bit…). There are LT1007s at the DAC outputs. In the control loop, the voltage clamps use LT1057 and the current clamps use AD744. Why are they different ?
Control loop schematics:
Teardown from eevblog forum: https://www.eevblog.com/forum/testgear/keitley-236-teardown-and-review/msg791944/#msg791944
The Keithley 2303/2304A are 3A/5A “high speed” power supplies with programmable sink capabilities. Quick look at the specs indicates that the resolution is a bit better that the comparable? Keysight 66332A, but the Vpp noise is a bit higher. If you want to know how it looks inside, take a look at xDevs repair page.
Keithley 2400 smu
The Keithley 2400 smu output stage have multiple voltages. This is in order to avoid too much heat dissapation when large current and high voltage difference between supply voltage and output voltage.
Take a look at the Dave Jones eevblog teardown: https://www.youtube.com/watch?v=DJ2nzvX2gBc. It uses AD847 as driver op-amp(?) for the output stage. This is a +/-15V device. That probably? means the the transistor output stage has gain. To keep complexity low, I plan to have no voltage gain in the output stage. This means that the +/-15V probably is a bit on the low side… Learn more…
Other devices used: AMP03(Precision, Unity-Gain Differential Amplifier), nais v214s solid state relay, AD7849 DAC, LT1112(Dual/Quad Low Power Precision, Picoamp Input Op Amps). Keithley Model 2400 review and VFD repair also have teardown details
Theory of operation (from 2400 service manual): “D/A converters control the programmed voltage and current, or voltage compliance and current compliance. Each DAC has two ranges, a 10V output or a 1V output. The DAC outputs are fed to the summing node, FB. Either the V DAC or the I DAC has the ability to control the main loop. If the unit is set for SV (source voltage), it will source voltage until the compliance current is reached (as determined by the I DAC setting), and the current loop will override the voltage loop. If, however, the unit is set for SI (source current), it will source current until the compliance voltage is reached (as determined by the V DAC setting), and the voltage loop will override the current loop. A priority bit in the Vclamp/I clamp circuit controls these functions. The error amplifier adds open-loop gain and slew-rate control to the system to assure accuracy and provide a controllable signal for the output stage, which provides the necessary voltage and current gain to drive the output. Sense resistors in the HI output lead provide output current sensing, and a separate sense resistor is used for each current range. The 1A range uses 0.2V full-scale for a full-range 1A output, while all other ranges use 2V output for full-scale current. Voltage feedback is routed either internally or externally“
“There are four voltage ranges: 0.2V, 2V, 20V, and 200V. The feedback gain changes for only the 20V and 200V ranges, resulting in three unique feedback gain values. A multiplexer directs the voltage feedback, current feedback, reference, or ground signal to the A/D converter. An opto-isolated interface provides control signals for both DACs, analog circuit control, and A/D converter communication to the digital section.“
The 2400 is one of the best known SMUs out there. But it’s worth mentioning that the sink accuracy is not as good as the source accuracy. This fact is hidden in the footnotes in the datasheet: “For sink mode, 1µA to 100mA range, accuracy is: ±(0.15% + offset*4)…1A range, accuracy is : ±(1.5% + offset*8)”. Also for >100mA the current accuracy is not very good. In addition “For operation above 105mA continuous for >1 minute, derate accuracy 10%/35mA above 105mA)”. Could be that the current shunt resistor is not optimal… or maybe there are other reasons…
Keithley 2450 / 246x
The Keithley 2450 is the successor of 2400. The specifications are improved in any all aspects. The user interface is completely modernised with a big LCD display. It was the 2450 that triggered my interest for SMUs. The 2460 adds more power. The 2461 even adds 18bit digitizer as show in this video. Go to the tektronix official pages to find details specifications and differences between the models.
I’ve not been able to find any teardowns of the Keithley 2450, but there are some images at xdevs.com
Signal path also have a review of 2450 where focus is on practical experiments.
Agilent B2912A Source Measure Unit SMU
Agilent B2912A Source Measure Unit SMU Teardown: https://www.youtube.com/watch?v=pKX50E_14MQ AD8208 current sense, 30N06 N-channel mosfet (60V 30A), ixta10p50p P-channel mosfet (-500V 10A), ISO7240A optoisolator, fds8978 mosfet for switching? (optimized for low gate charge, low rDS(on) and fast switching speed), .. LM399 reference, OPA1611 (ultralow noise and distortion op-amp), ISO7240 and ISO7242 isolators, ADS1675 (24 bit delta-sigma precision ADC), AD8676 (precision op-amp with offset 12 μV, drift 0.2 μV/°C, noise 0.10 μV p-p), ADG413 and ADG452 switch, 8N60C output N channel mosfet, ADS1675 24bit ADC: http://www.ti.com/lit/gpn/ads1675,
Here is a screeenshot from Dave Jones video showing current shunt and the switching mosfet. Looks like it uses mosfet switches to connect and disconnect the two? different shunt resistors…
Screenshot from the output stage:
Have found some inerresting units from Yokogawa: GS210, GS820, GS610.
GS210 is “just” a programmable DC source. It’s users manual, appendix 2 includes block diagram and brief theory of operation:
“When the GS200 is operating as a voltage source, SW1 is connected to V, and source voltage Vo is the product of source DAC value Vs and R2/R1. If load current IL increases positively and IL × Rs exceeds positive limiter DAC value Vp, the diode of the limiter circuit will turn on and suppress the load current to Vp/Rs. A similar operation takes place when the load current is negative. When the GS200 is operating as a current source, SW1 is connected to I, and source current Io is the product of source DAC value Vs and (R2/R1)/Rs. If load voltage VL exceeds positive limiter DAC value Vp, the diode of the limiter circuit will turn on and suppress the load voltage to Vp. A similar operation takes place when the load voltage is negative… …The GS200 is equipped with a measurement circuit that is separate from the source and limiter circuits. In voltage source mode, the measurement circuit uses A/D converters (ADC) to measure the current that is received. In current source mode, the measurement circuit uses A/D converters (ADC) to measure the voltage that is received. Differing from the voltage and current source functions described above, when the mV range is selected in voltage source mode, SW2 is connected to mV. The source voltage Vo is voltage-divided by R3 and R4. In this situation, because the output resistance becomes R4 (2 Ω), if the DUT does not have a high impedance, the source voltage is reduced according to the DUT’s impedance. Additionally, because selecting mV causes the limiter and current measurement circuits to be disconnected by Rs and SW2 that are used for current sense, these circuits do not operate.“
The GS820 and GS610 are “real” source measurement units. Block digram can be found in the respective user manuals ( 820 , 610). A discussion and some pcb photos for GS610 on https://www.eevblog.com/forum/testgear/yokogawa-gs610-source-measure-unit-service-manual-needed/
HP 3245 Universal Source (4 quadrant)
HP 3245A universal four quadrant source combines precision dc capabilities with versatile ac performance, including arbitrary waveform generation.
According to https://xdevs.com/fix/hp3245a/ it compares favourably agains modern SMUs both in DC accuracy and AC performance.
Downside is that the standard edition only goes from -10V to 10V and -100mA to 100mA. A special option brings the voltage to 100V. Alternatively an option that gives two independent outputs.
Nice thing is that the schematics are available here: https://xdevs.com/doc/HP_Agilent_Keysight/3245A/3245A-CLIP.pdf
Advantest R6243/6244 ++
I’m a bit confused about the Advantest brand. Seems it’s under ADCMT, Rohde Schwarz. Anyway, there is a model R6243/6244 that seem to be a SMU (at least 4quadrant)
Not too much to find about this unit, but here’s an entry eevblog entry with some info and references.
“6241A is a DC voltage and current source/monitor with high-performance features including a source display resolution of 4½ digits, measuring display resolution of 5½ digits and base accuracy of 0.02 %. It can operate as voltage source or as a real current source in DC, sweeped or pulsed operation. High resolution and extended protection tools make it perfect for testing sensitive electronic components.“
RohdeSchwarz / Hameg HM8143 (two quadrant)
Linear regulated, two-quadrant power supply (current source and sink capabilities)
Service manual for the Hameg HM8142(older version) with schematics https://docs-emea.rs-online.com/webdocs/0296/0900766b80296cf9.pdf. Note that a french version also contains a short one page very high level description of the operating principle (https://drive.google.com/file/d/0Bxd9xq3tnxpiNWxueGQ5blpsZDg/view)
RohdeSchwarz NGM202 NGM201 NGL202 NGL201
A new (2019) series of modern two quadrant power supply, you can read more here: https://goughlui.com/2019/08/20/launch-day-review-rohde-schwarz-ngm202-power-supply-overview/ “This is a two-quadrant programmable DC power supply suitable for bench-top and rack-mount use, being a more featureful version of the NGL-series power supplies launched earlier in late-January. Such power supplies have appeal to users working with DC-to-DC converters, battery operated devices or evaluating rechargeable batteries thanks to its ability to both source and sink current on the same channel with variable internal impedance.“. The author (Dr. Gough Lui) of that article also discusses pros and cons of 2 vs 4 quadrant supplies and has even performed a “marked survey” of various 2 and 4 quadrant supplies out there (2019): https://goughlui.com/wp-content/uploads/2019/08/ngm202-market-survey.png
RohdeSchwarz SMU NGU201, NGU401
In 2021 they released SMU versions which I have not read too much about yet: R&S®NGU source measure units
“With six measurement ranges for current and a resolution
of up to 6½ digits when measuring voltage, current and
power, the R&S®NGU source measure units are perfect
for characterizing devices that work from extremely low
power consumption to high currents in the ampere range.
Using ammeters with feedback-amplifier technology
increases accuracy and widens the sensitivity down
to the nA range.”
A presentation video for the NGUs from R&S on youtube.
Overview of the differences:
R&S now seem to have a pretty good selection of modern 2/4 quadrant supplies sharing save user interface. Seems like the software for the SMUs are still “work in progress” as it (currently?) lacks variable graph timebase etc. according to https://www.eevblog.com/forum/testgear/rohde-schwarz-ngu401/msg3568808/#msg3568808
Useful discussion of its cc circuitry: why do the inputs of this cc circuit op amp need clamping
Agilent E364xA Dual Output DC Power Supplies service manual
Teardown and repair of E3646A: https://www.youtube.com/watch?v=kyY4DXOth2k (The Signal Path)
https://www.eevblog.com/forum/testgear/my-poor-mans-smu-the-agilent-66311b/ ” The Agilent 66311B Mobile Communication DC Source supplies 0 – 15 volts at up to 3 amps and can sink 0-15 volts at up to 2 amps”
Relevant programmable load projects
A four quadrant power supply should be able to sink current. So it’s interesting to take a look at design considerations for an electronic load.
Peter Oakes: https://www.youtube.com/watch?v=vd5IBFFjnOc&t=385s , with improvements: https://www.youtube.com/watch?v=rh32ylmlz-A&t=8s (“In my previous video I left off with the performance response time at between 500uS and a couple of mS depending on the load being switched, this was not good enough for some viewers and they let me know it. So……“
http://www.scullcom.uk/category/projects/dc-load/ Louis Scully goes through the design and build of an electronic load capable of Constant Current, Constant Power and Constant Resistance. Block diagram:
Schematics (PDF) for N3304A 300 Watt Electronic Load Module linked to from a eevblog forum entry: http://www.eevblog.com/forum/projects/diy-bench-power-supply-psl-3604/?action=dlattach;attach=206857.
TODO: Add more relevant stuff ?
Relevant programmable voltage references/sources
Using programmable voltage references can also be handy when experimenting with circuits. A lot can also be learned about how to design stable and accurate circuits, an important part of a Poor mans’s SMU. Here are a few:
Ian Johnstons Handheld Precision Digital Voltage Source – PDVS2 “A handheld Precision Digital Voltage Source, true 0.0000V to 10.0000Vdc range, battery powered…an accuracy/stability down to the uVs it has a multitude of uses as a calibrator, reference & precision voltage source. ” Uses 18bit DAC9881 and LM399.
A lot to be learned in Ian’s informative 3 part video starting here: https://www.youtube.com/watch?v=2Ssg05oFCbk&t=940s. Ses my brief review of the PDVS2 here.
A later version PDVS2mini has a different user interface and replaces the 18bit DAC9881 with 20bit MAX5719A. The new DAC that has higher resolution but poorer linearity. I believe the final linearity is improved by having multipoint calibration. The latter can be purchased from Ians webpage. Note that he does not always have them in stock.
https://www.barbouri.com/2016/07/21/programmable-voltage-reference-v2-12-assembly/ “A precision programmable voltage reference circuit capable of 0.001 to 4.095 volt output in 1 mV steps with an accuracy of 100 uV.“
High Precision Voltage Source (2017) “…showcasing an ultrahigh precision programmable voltage source using ADI/LTC products together. The AD5791 with the LTZ1000, ADA4077, and AD8675/AD8676 can be used to provide a programmable voltage source that achieves 1 ppm resolution with 1 ppm INL and better than 1 ppm FSR long-term drift. “
Scullcom Hobby Electronics – DC Voltage Calibrator “In this project we will design and build a DC Voltage Calibrator, providing a voltage range from 0 to 10 volts in 1 milli volt steps. The user interface will be a TFT display with touchscreen.“
The 20-Bit DAC Is the Easiest Part of a 1-ppm-Accurate Precision Voltage Source “…Advances in semiconductor processing, DAC architecture design, and fast on-chip calibration techniques make possible highly linear, stable, fast-settling digital-to-analog converters that deliver better than 1-ppm relative accuracy, 0.05-ppm/°C temperature drift, 0.1-ppm p-p noise, better than 1-ppm longterm stability, and 1-MHz throughput. These small, single chip devices have guaranteed specifications, do not require calibration, and are easy to use…“
Refduino “The device consists mainly of an REF102, a 10V reference voltage source with a 2.5ppm/°C temperature drift. It mayst be calibrated by a potentiometer R6. This reference is fed into an AD5791 DAC (20 Bit, B version) which has an 1 ppm resolution and a 1 ppm INL. A buffer at the output shall eliminate pulling/pushing effects of the load. We know that this DAC is expensive, but you get what you pay for. And for a reference, only the best components shall be used…“
DRS20 – 20 bit Dual Reference Source “The DRS20 is a dual-channel reference source offering ultra high precision and stability and provides a +/-10V output. The output current can be up to 10mA and the 4 wire connection provides a 1µV/mA load regulation” They also have a lot of other high-spec stuff. Their ATX series uses LTZ1000 reference.
https://www.eevblog.com/forum/projects/project-kx-diy-calibrator-reference-sourcemeter/ A lot of volt-nuts info in this eevblog thread.
DIY ACV/DCV calibrator with AD5791 using AD5791 and LTZ1000A has been presented in the eevblog forum. Includes schematics etc.
BHUMI is an interesting project also using LTZ1000 and AD5791. Three fixed outputs (0.1 V, 0.5 V, 2.5 V) as well as a programmable output (-10 V … +10 V). Driven by an Arduino.
Relevant multimeter related designs
Service manual with schematics: Agilent 34401A Service Guide. Or in pdf format that is searchable here: https://www.ee.ryerson.ca/guides/instrument-manuals/Agilent-HP_34401A_Service_Guide.pdf
Unfortunatelly, 34401 seem to use some custom parts in the analog switching and resistor networks, but a lot can be learned anyway.
Multimeter shield from Digilent: https://store.digilentinc.com/dmm-shield-7-function-digital-multimeter-shield/ “ …built around the Hycon Technology HY3131 DMM chip, and is factory calibrated to provide +-.01% accuracy in voltage measurements, current measurements, and most resistance measurements…“
Highly Integrated 4 1/2 Digit Low Power Handheld Digital Multimeter (DMM) Platform Reference Design “…to demonstrate a highly integrated, low cost, and low power Digital Multimeter platform. The solution delivers 4.5 digits or 60K display count resolution…“
Open source high accuracy DC Multimeter Various info/experiments/projects with voltage reference, calibration etc. “Project for building an open source, DIY, low-cost and high accuracy digital multimeter“
TODO: Add more relevant stuff ?
Current Sources: Options and Circuits “…shows examples of current sources from the microampere range which are integrated in specific devices and also medium to high power discrete applications up to the 1 A range. “
Difference Amplifier Forms Heart of Precision Current Source “…Now, high-precision, low-power, low-cost integrated difference amplifiers, such as the AD8276, can be used to achieve smaller, higher performance current sources…“
General reference designs for high precision design
TI Designs: TIDA-01012 Wireless IoT, Bluetooth® Low Energy, 4½ Digit, 100-kHz True RMS Digital Multimeter Reference Design A bluetooth low energy reference design that also “…implements basic DC voltage and current measurement modes as well as true RMS AC voltage and current measurement modes inherent in most handheld DMMs in the market today.”
Ultrahigh Sensitivity Femtoampere Measurement Platform“…provides a reference design for realworld application by partitioning the system into a low-leakage mezzanine board and a data acquisition board“
10uA-100mA, 0.05% Error, High-Side Current Sensing Solution Reference Design “...we are showcasing an ultrahigh precision programmable voltage source using ADI/LTC products together. The AD5791 with the LTZ1000, ADA4077, and AD8675/AD8676 can be used to provide a programmable voltage source that achieves 1 ppm resolution with 1 ppm INL and better than 1 ppm FSR long-term drift. “
High Precision Voltage Source (Analog Dialogue, 2017) “The AD5791 with the LTZ1000, ADA4077, and AD8675/AD8676 can be used to provide a programmable voltage source that achieves 1 ppm resolution with 1 ppm INL and better than 1 ppm FSR long-term drift“
A Standards Lab Grade 20-Bit DAC with 0.1ppm/°C Drift (Linear Technology, 2001)“A useful development would be a practical, 20-bit (1ppm) DAC that is easily constructed and does not require frequent calibration…“
Unstructured collection of theory, construction ideas/principles, component selection etc…
Interesting reading: download.tek.com/document/LowLevelHandbook_7Ed.pdf
Stability before absolute precision
Modern meters use precision reference, low temperature coefficient and then software calibration. Not necessarily high PRECISION (such as 0.01%) parts. For example if you know the gain of some stages are STABLE, you can design the electronics so that you can measure absolute gain in software. As long as your circuit allow for “setting up” the circuit for (auto)calibration. Ref (https://youtu.be/oXX6GhhoJls?t=821). ….add more on software calibration…
I see several datasheets where low frequency noise is often specified from 0.1Hz to 10Hz. But long-time stability often more difficult to find. How much drift will a component have over several weeks or months? Seems to be difficult to find good proven data. One of the reasons can be found in this article by Bob Pease: What’s All This Long-Term Stability Stuff, Anyhow? ““Why don’t you have full information on [ this op amp for long-term stability]?” I tried to explain that we can’t afford to gather this data on every little product. We can’t afford to take the data and analyze it. We also can’t afford the delays in our time-to-market. Even if we released it soon and updated the datasheet later, people would look at the first datasheet and ask the same question.”
Precision, accuracy, resolution
I tend to mix the terms accuracy and precision. According to Fluke :
Precision: An instrument’s degree of repeatability—how reliably it can reproduce the same measurement over and over.
Accuracy: An instrument’s degree of veracity—how close its measurement comes to the actual or reference value of the signal being measured.
Resolution: The smallest increment an instrument can detect and display—hundredths, thousandths, millionths.
Counts and digits are terms used to describe a digital multimeter’s resolution. Today it is more common to classify DMMs by the total counts than by digits.
- Counts: DMMs that offer higher counts provide better resolution for certain measurements. For example, a 1999-count multimeter cannot measure down to a tenth of a volt if measuring 200 V or more. Fluke offers 3½-digit DMMs with counts of up to 6000 (meaning a max of 5999 on the meter’s display) and 4½-digit meters with counts of either 20000 or 50000.
- Digits: The Fluke product line includes 3½- and 4½-digit digital multimeters. A 3½-digit DMM, for example, can display three full digits and a half digit. The three full digits display a number from 0 to 9. The half digit, considered the most significant digit, displays a 1 or remains blank. A 4½-digit DMM can display four full digits and a half, indicative of higher resolution.
Precision and drift
Minimize Voltage Offsets in Precision Amplifiers“Voltage-offset errors in precision amplifiers are partly caused by input bias currents. This article analyzes the problem and proposes a solution based on resistor networks, both discrete and integrated.“
OPA388 Ultra-Low Offset Voltage: ±0.25 µV, Zero-Drift: ±0.005 µV/°C
Theoretically resolution based on DAC/ADC bits:
16bit = 1/(2^16) = 1/ 65536 = 0.00001525… @1V = 15.25uV @5V= 76uV
18bit = 1/(2^18) = 1/ 262144 = 0.00000381… @1V = 3.81uV @5V = 19uV
Already here we see that in order to obtain total 100uV resolution, we must change the max voltage used for low range measurements… If using 16bits converter.
In practise, this is probably NOT possible to achieve. There are probably other part of the design that will contribute more to offset and noise. If not, we might increase resolution. later.
What is counts vs bits vs percent vs ppm ? The following table is from Understanding and Applying Voltage References (1999):
10ppms of 1V = 10uV ?
1ppm of 1V = 1uV ?, 15ppm of 1V = 15uV ?
Investigation: How Accurate is that 3.5 digit Multimeter?“But in the end, curiosity got the better of me, so I decided I would give a pile of 3.5 digit multimeters a test and see just how accurate they were“
http://www.electronicdesign.com/power/whats-all-error-budget-stuff-anyhow “…many engineers in Europe were quite unfamiliar with the concept of an “error budget.” How can you design a good circuit without being aware of which components will hurt your accuracy?“
https://www.edn.com/design/analog/4368505/Error-budgets-keep-your-analog-signal-path-honest “Some programmers’ claims that you can calibrate out all errors in software may lull you into a false sense of security regarding the errors in your design…“
Errors and Error Budget Analysis in Instrumentation Amplifier Applica AN539: “describes a systematic approach to calculating the overall error in an instrumentation amplifier (in amp) application“
Self “calibration” ?
I would like to design a system so that measurement errors caused by offsets, gain nonlinearity, drift can be (to a large extent) adjusted for in software. I probable need some extra components for this, such as switches, muxes and a good reference. I still, however, want to use precision components with as low drift as possible.
Voltage Reference Application and Design Note (2005) , subsection “Auto-Calibrated 12 Bit Data Acquisition System” suggests “moving” the high precision voltage reference from the ref input on DAC/DAC and instead put it on the “outside” of the muxes etc. in order to account for “all” system inaccuracies other places that the DAC/ADC itself.
In figure 25 the voltage reference is put directly on the A/D. Inaccuracies th opamps, mux etc can NOT be accounted for by automatic calibration.
In figure 26, an internal reference is used in the converter. Instead, the precision reference is placed on one of the mux inputs. It’s now possible to automatic full system calibration that accounts for all component inaccuracies.
DESIGN FOR SELF-CALIBRATION OF INSTRUMENTATION “In a self-calibrated measuring method the input/output relation of a measuring system is directly determined by the self-calibration algorithm with the use of the internal reference quantities and elements, and the measuring errors are self-corrected by the corresponding signal and data processing algorithms…“
https://meettechniek.info/measurement/self-calibration.html (2014) “Electronic circuitry used in measurement equipment are subject to a certain deviation who influences the stability and accuracy. Offset, drift en variations in gain due to temperature changes, aging and power supply changes cause measurement uncertainties of sometimes unacceptable proportions. By making use a self-calibration circuit and protocol the accuracy of a measurement is vastly increased…“
Maxim AP261 Calibration-Multiplexers Ease System Calibration (Maxim AppNote 261)“IC switches and multiplexers are proliferating, thanks to near-continual progress in lowering the supply voltage, incorporating fault-protected inputs, clamping the output voltage, and reducing the switch resistances. The latest of these advances is the inclusion of precision resistors to allow two-point calibration of gain and offset in precision data-acquisition systems…“
From reference design (Ultrahigh Sensitivity Femtoampere Measurement Platform): “To minimize any offset errors due to the ADC, an ADG1419 single-pole/ double-throw (SPDT) analog switch shorts the input of the resistor divider to ground and allows the software to measure the offset error due to the ADC and resistor divider. When offset cancellation is enabled, the software subtracts the measured offset from every reading. Any remaining offset is due only to the ADA4530-1 circuitry“
From AN1711 APPLICATION NOTE SOFTWARE TECHNIQUES FOR COMPENSATING … ADC ERRORS “This document provides some methods of calibrating the ADC. Some ADC errors like Offset and Gain errors can be cancelled using these simple software techniques…“
Improve DAC integral nonlinearity through gain correction (2011)“Linearity errors are the most challenging to handle of the three since, in many applications, the user can null out the offset and gain errors, or compensate for them by building end-point auto calibration into the system design. Linearity errors, however, require more complex correction.“
ADC Calibration (National Instruments, 2009) “…a method used to compensate for an internal DMM gain error, is a feature exclusive to the NI 4070/4071/4072 DMM that allows you to appropriately trade off measurement speed for long-term accuracy…“
Calibrating Amplifiers and ADCs in SoCs “…in real world, there are many errors that get introduced into the system affecting the ADC’s output. The most important errors and the ones that we are going to discuss in this article are the offset and gain errors…“
APPLICATION NOTE 4170 Improve Current Measurement Accuracy by Skewing the Input Offset Voltage on Current-Sense Amplifiers “…presents a method that introduces known input VOS by suitably sizing input resistors for current-sense amplifiers.“
The ABCs of ADCs: Understanding How ADC Errors Affect System Performance “Offset and gain errors can easily be calibrated out using a microcontroller (µC) or a digital signal processor (DSP). With offset error, the measurement is simple when the converter allows bipolar input signals. In bipolar systems, offset error shifts the transfer function but does not reduce the number of available codes…. There are two methodologies to zero out bipolar errors. In one, you shift the x and y axes of the transfer function so that the negative full-scale point aligns with the zero point of a unipolar system….“
Software calibration reduces D/A converter offset and gain errors “In reality, the accuracy of the output voltage is subject to a number of factors, including gain and offset errors from the D/A converter and from other components in the signal chain. The system designer must compensate for these errors in order to get the highest possible output accuracy.“
Oversampling and decimation (Atmel application note)“By using a method called ‘Oversampling and Decimation’ higher resolution might be achieved, without using an external ADC. This Application Note explains the method, and which conditions need to be fulfilled to make this method work properly…“
Design details and selecting parts
Control loop / error amplifier
How Does a Power Supply regulate It’s Output Voltage and Current? “…To accomplish this most all power supplies have separate voltage and current feedback control loops to limit either the output voltage or current, depending on the load. To illustrate this Figure 1 [below] shows a circuit diagram of a basic 5 volt, 1 amp output series regulated power supply operating in CV mode.“
Seems like a lot of power supplies use variants of this conceptual model where the control of voltage is “OR”ed with the current control using diodes or transistors. The question is how this will work when you add an additional negative current limiter. Then the voltage and negative current control loop will “fight” each other… Who is under control ? How to obtain stability ? One of the first parts I want to learn about is how to design a stable and accurate control loop.
From LM12/LM318 Application Note 446 A 150W IC Op Amp Simplifies Design of Power Circuits “External current limit can be provided for an op amp as shown in Figure 19. The positive and negative limiting currents can be set precisely and independently down to zero with potentiometers R3 and R7. Alternately, the limit can be programmed from a voltage supplied to R2 and R6. The input controls the output when not in current limit. This is just the set-up required for an operational power supply or voltage-programmable power source“:
“The power op amp, A4, is connected as an inverting amplifier. Its output current is sensed across R10. This sense voltage is level shifted to ground by A3, a differential amplifier that is made insensitive to the op amp output level by trimming R9. With current below preset levels, the outputs of A1and A2 are clamped by D1 and D2 with Q 1 and Q2 turned off. When the current threshold is reached, the relevant amplifier will come out of clamp, saturate the transistor on its output and take over control of the summing node. The clamp diodes limit the swing on the outputs of the current-control amplifiers while the transistors disconnect frequency compensation until the summing node is engaged. This ensures fast activation of current limit. Recovery back to voltage mode is also fast. The LM318 wideband amplifier is required for A1 through A3.”
According to a post in eevblog forum: “…Q1 and Q2 work as analogue switches, connecting a compensation RC network when either of the two current limit opamps is engaged…”
By first priority will be to create/design the control loop. The following sections discuss some constructions out there.
Measurement current range switching
In order to obtain high dynamic range I probably need to switch between current shunt resistors. Maybe also when measuring volt. Can I get away with two ranges ? Switch using relay? or are there other ways ? Switching extra resistor on/off in parallell ? Or use them in series such as Current measurement with auto-ranging yields 180dB range dynamics:
Additional info about range switching in DMMs…
Switching Low-Level Signals to a DMM (National Instruments, dec 2017) “This document will describe the fundamental principles that affect low-level measurements, which relay types best address these principles…“
Must I use relay for the current switching, or can I use mosfets such as described here: http://www.electronic-products-design.com/geek-area/electronics/mosfets/using-mosfets-as-general-switches ?
LowPowerLab has designed the CurrentRanger which is a nanoAmp current meter featuring auto-ranging, uni/bi-directional modes, bluetooth data logging options and more. “While the amplifiers, topology and shunt configuration of the CurrentRanger are similar to the µCurrent by EEVBlog, CurrentRanger is a product with significantly different features and goals…“. It uses mosfets to switch between the input ranges.
You also might take a look at Dave Jones Designing A Better Multimeter PART 2 where mosfet switching current ranges is discussed.
Or an older Dave Jones video where he explains range switch in his uSupply project (direct link)
In Keithley 236/237 they have an interesting current range switching approach as explained in the service manual: “Without special circuitry, current measurement range changes may cause voltage spikes to occur at the output. Voltage spikes are caused by contact bounce of the selected range relay. These spikes occur too fast for the bootstrap amplifier to react to the change in feedback voltage. As a result, the system cannot adjust itself accordingly to maintain a steady voltage output. The Model 236/237 resolves this problem by using circuitry that, in effect, shunts the current ranges with a 100ohm resistor whose voltage drop ramps in a linear fashion towards 0V. This slow rate of voltage change allows the system to adjust itself keeping the output voltage constant (spike free)…“. A closer description is provided in section 4.2.3.
The eevblog discussion also mentions the X-NUCLEO-LPM01A power measurment shield that uses a simular approach but also uses programmable regulator when in low current mode and separate power amplifier when in high current mode.
RocketLogger: Mobile Data-Logger for Ultra-Low Current and Power Measurements explains details on how the current measurement is implemented with two ranges. https://pub.tik.ee.ethz.ch/students/2016-FS/SA-2016-40.pdf . “For the actual implementation of the dual range we chose to combine a shunt current meter with a linear feedback current meter. The shunt current meter is always active and is used for measurements above 2 mA and for the range switching…“
Question: Should I have two measurements, one for current limiting and one for actual “presented” current measurement ? If less precision is needed for the current limit, maybe it makes sense to separate the two. Dave Jones does that in his uSupply…
Fundamentals of Current Measurement (DigiKey 2019) is the first of a three part article about current measurement: “This three-part series looks at the underestimated nuances of current sensing. Part 1 (here) will discuss the general setup, selection, and implementation of a current sense resistor. Part 2 will discuss associated circuitry such as the critical analog front-end (AFE) and instrumentation amplifier. Part 3 discusses the use of funnel amplifiers to amplify current measurements in applications where the load is being driven by higher voltages.”
Difference amplifier(DA) “In summary, a DA can be utilized for either high-side or low-side sensing. When used for high-side sensing, error can be introduced by the finite common-mode and differential-mode input impedances… However, a DA places a load on the system bus voltage due to its finite common-mode and differential-mode input impedances. This load draws current from the system bus voltage, which introduces uncertainty in the measurement…“
Instrumentational amplifier(IA): “The first advantage over a DA is the ability to easily change the differential gain… Secondly, the inputs are connected to the non-inverting inputs of a buffer amplifier… One disadvantage: input common-mode voltage of IAs is limited by their supply voltage. Therefore, IAs typically is utilised for low-side measurements“
Current sense amplifiers: “Current shunt monitors are devices that place little load on a system and allow for sensing current under high common-mode voltage conditions… typically have fixed gains. The exceptions include current output devices, which require an external precision resistor to set the gain…“
Current Sense Circuit Collection (Linear Technology ApplicationNote 105, 2005) compiles solutions to current sensing problems and organizes the solutions by general application type.
Difference and Current Sense Amplifiers (MT-068 tutorial) “…A simple… difference amplifier can be constructed with four resistors and an op amp… There are several fundamental problems with this simple circuit…“
Current Sensing: Low Side, High Side, and Zero Drift (Texas Instruments video) Info about high side measurement and what too look for in specs (from 2:45)
10uA-100mA, 0.05% Error, High-Side Current Sensing Solution Reference Design “This TI Precision Verified Design provides the theory, component selection, simulation, complete PCB schematic and layout, bill of materials, and measured performance of a split-supply, high-side four decade current sensing solution that can accurately detect load currents from 10uA-100mA. “
High-Side Current Sensing with Wide Dynamic Range: Three Solutions (2010) “This article will focus on providing current-sensing solutions that can help designers accurately monitor wide-ranging dc currents in the presence of high common-mode voltages. Special attention will also be devoted to temperature performance“
Calculating Accuracy in High-Side Current-Sense Amplifiers (2016) “…Determining the optimum value for the current-sense resistor (RSENSE) is crucial. Larger RSENSE values increase the series IR voltage drop and power loss, but minimize the effect of the offset voltage error…“
“I discussed that the common-mode voltage level is critical in determining the implementation … As the maximum common-mode range increases, it typically means a reduction in the achievable accuracy. For example, Texas Instruments’ INA210 current-shunt monitor has a maximum common-mode voltage of 26 V and offers an input offset voltage of 35 µV. Input offset is one of the most important device parameters when determining measurement accuracy. Now compare the specifications of the INA210 to those of the INA283 current-sense amplifier. Its common-mode range extends up to +80 V (as well as down to –16 V), with a maximum input offset of 70 µV.”
Extending Beyond the Max Common-Mode Range of Discrete Current-Sense Amplifiers “The simplest approach to monitor high voltage highside current sensing is a design with a low voltage current sensing amplifier with external input voltage dividers, for example, if a 40V common mode voltage amplifier is selected for a 80V application, the 80V input common mode needs to be divided down to 40V common mode voltage. This voltage division can be accomplished using external resistor… …As voltage dividers has serious consequences with output error and degradation in performance, another alternative approach is to shift the ground reference of the current output amplifier to the high voltage common mode node…“
Precision Current Sensing Solutions Guide Describes considerations and criteria for specification and selection of resistor, difference amp. Includes a lot of resistor details: “This solution guide provides guidance for the selection of low-ohmic components that are constructed from materials that satisfy the design requirements of precision current measurement circuits operating in environments with a high degree of temperature variation. Guidance is also provided for selection of a high-quality difference amplifier.“
Current measurement circuit
Before I started this project I did not know the real difference between opamp, differential amps and instrumentational amps. Now I know a bit more and have currently landed on a topology using three opamps. In a way it’s a instrumentation opamp, but made out of three separate opamps.
I’ll probably go for a combination of AD5422 in from of a LT1997-3. But I’m also interested in see how if ADA4254 can be used. That’s one of the few instrumentation amplifiers I have found which has up to +/-28V supply. In addition, it has programmable gain and a lot of other features !
The three subsections where below can still be relevant to read.
Differential op amp
Use differential amplifier (with bi-directional current sense), examples:
TS1102-50EG5 ( 50x, 0.5% gain error max, 200uV offset max )…are self-poweredONLY UNIDIRECTIONAL and self powered. Not for me.
and feature a wide input common-mode voltage range from 2 to 27 V.
TS1101 has a separate pin indicating polarity as discussed in AN840: Redefining a New State-of-the-Art in Microampere Current-Sense Amplifier. Can it be used ?
INA226 ( 1x, 0.1% gain error max, 10uV offset max ), From: https://www.mouser.se/publicrelations_techarticle_curentsensemonitoring_2015final/: TI INA226 is one of the highest precision current sense monitors on the market today, with an offset voltage of just 10µV and a common mode range of up to 36V. Not for me. No analog output. INA219B ( 0.5% gain error max, 50uV offset max ) No analog output ? INA210-INA215 (different gains) VCM=-3V to 26V. I need better specs on negative side.
INA149 “precision unity-gain difference • Common-Mode Voltage Range: ±275 V amplifier with a very high input common-mode • Minimum CMRR: 90 dB from –40°C to +125°C voltage range”. offset:350uV, max:1100uV. Offset drift: 3µV/°C,max 15 µV/°C. ( VCM –20 to +25 when supply voltage is +/-5V. @15V VCM becomes +/-275V ) From: http://www.ti.com/lit/ds/symlink/ina149.pdf (page 15:): “…the sense resistor imbalances the input resistor matching of the INA149, thus degrading its CMR. Also, the input impedance of the INA149 loads RS, causing gain error in the voltage-to-current conversion. Both of these errors can be easily corrected… addition of a compensation resistor (RC), equal to the value of RS… if RS is less than 5 Ω, degradation in the CMR is negligible and RC can be omitted.”. Video from TI: https://www.youtube.com/watch?v=4by0NI3Pc9g
AMP03 VCM = ±10 V ?
AD8274 Gain of ½ or 2. Offset 100uV- 500uV. Offset drift 6μV/°C (a bit high). 86/92 dB minimum CMRR . Low distortion. “Excellent gain accuracy 0.03% maximum gain error 2 ppm/°C maximum gain drift “. VCM = ±40 V. (Offset is a bit high but as long as the drift is not bad, maybe its usable with some software cal…)
(*) AD8276 Unity gain, Offset 100-500uV. Low offset voltage drift: ±2 μV/°C maximum (B Grade) Low gain drift: 1 ppm/°C maximum(0.5 – 2) (B Grade) gain drift 5ppm/C. Gain error:0.01% max:0.05%, CMRR:86 dB, VCM = ±27 V? Or is 27V too low?
AD8278 Gain of ½ or 2. Offset 50uV- 250uV. Low offset voltage drift: ±1 μV/°C (0.3uV – 1uV )B grade, Low gain drift: 1 ppm/°C maximum (B Grade) 80db CCMR, VCM = ±27 V. AD8279 is a dual version. AD8278 has better spec that 8274/8276, but still only 27V VCM…
(*)AD629 B is better than A. “The AD629 is a difference amplifier with a very high input, common-mode voltage range. It is a precision device that allows the user to accurately measure differential signals in the presence of high common-mode voltages up to ±270 V.” . unity gain. “…allows the user to accurately measure differential signals in the presence of high common-mode voltages up to ±270 V” AD629 is improved over INA117. Measuring −48 V High-Side Current Using the AD629 Difference Amplifier, AD8603 AD780 Reference, and AD7453 12-Bit ADC Single-Supply ComponentsOp Amp. Offset 100uV, max: 500uV, max drift:10uV/C, Gain error:0.01mV max0.03mV. 3 ppm maximum gain nonlinearity (AD629B), 20 μV/°C maximum offset drift (AD629A), 10 μV/°C maximum offset drift (AD629B), 10 ppm/°C maximum gain drift. From AN-1531:
The AD629 looks promising. Expensive, but I only need one.
AD8479 cmm=+/-600V, “…difference amplifier with a very high input common-mode voltage range”. Gain error: 0.005mV (max 0.01). 1mV max offset (double that of AD629?). Analog devices lists this as an alternative to AD629 because AD8479 has “2x wider Common-mode range, higher input impedance, lower voltage drift and gain error, with R-R Output“. It is recommended for new designs. However, it looks like the initial offset is higher…
LT1997 offset and gain drift down to 1ppm if CMM lower that supply voltage (+/-30V?), CMM=160V (over-the-top operation). Max offset 60uV. “…a one-chip solution for accurately amplifying voltages. Gains from –13 to +14 with accuracy of 0.006% (60ppm) can be achieved using no external components…”
If you read the datasheet you will notice that it enters a “over-the-top” mode at CM voltage 1.75V below supply. It will then have significantly less precision. Not neccessarily a problem for me as the supply voltage can be +/-30V.
LT1991/LT1996 seems to be something similar to LT1997, but not so good specs. ???
LT6376 (prelease? oct. 2018) 0.0075% (75ppm) Maximum Gain Error, 1ppm/°C Maximum Gain Error Drift, 2ppm Maximum Gain Nonlinearity. Max offset 200uV. “…is a gain of 10 difference amplifier which combines excellent DC precision, a very high input common mode range and a wide supply voltage range. It includes a precision op amp and a highly-matched thin film resistor network. It features excellent CMRR, extremely low gain error and extremely low gain drift. Comparing the LT6376 to existing difference amplifiers with high common mode voltage range, the gain of 10 and selectable resistor divider ratios of the LT6376 offer superior system performance by allowing the user to achieve low input referred noise with maximum precision and speed.“
Looks like this is a bit more flexible w.r.t. “over-the-top” range. By dividing the input before 10x gain, it enters the “over-the-top” range at a higher voltage than LT1997 while keeping some gain.
AD8210 The AD8210 is a single-supply, difference amplifier ideal for amplifying small differential voltages in the presence of large common-mode voltages. More: http://www.analog.com/media/en/training-seminars/tutorials/MT-068.pdf. “The operating input common-mode voltage range extends from −2 V to +65 V”. Only down to -2…. Not usable for me ? AD8207 for bidirectional current sensing applications. “The AD8207 is a single-supply difference amplifier ideal for amplifying small differential voltages in the presence of large common-mode voltage. The operating input common-mode voltage range extends from −4 V to +65 V”. Only down to -4…. Not usable for me ?
Comparing interesting candidates:
Max offset 1100uV
- 0.0075% (75ppm) Maximum Gain Error
- 1ppm/°C Maximum Gain Error Drift
- 2ppm Maximum Gain Nonlinearity
- ±230V Common Mode Voltage Range (90db minimum CMRR)
- Max offset 200uV
- 0.006% (60ppm) Maximum Gain Error (Gain = 1)
- 1ppm/°C Maximum Gain Error Drift
- 2ppm Maximum Gain Nonlinearity
- ±160V Common Mode Voltage Range
- Max offset 60uV
- 3 ppm maximum gain nonlinearity (AD629B)
- 20 μV/°C maximum offset drift (AD629A)
- 10 μV/°C maximum offset drift (AD629B)
- 10 ppm/°C maximum gain drift
- ±270V Common Mode Voltage Range
- 5 ppm maximum gain nonlinearity
- 10 µV/°C maximum offset voltage drift
- 5 ppm/°C maximum gain drift
- ±600 V common-mode voltage range
- Max offset 1mV?
Parametric search at TI http://www.analog.com/en/parametricsearch/11081#/p4133=Difference%20Amplifier
After testing a prototype with 1997-3 and 1ohm shunt to measure current, I found that it it gave measurement error that increased as the voltage over the shunt increased. This was even more obvious when I tried a higher resistance shunt. This can (probably) be improved by adding a compensation resistor to one input. Is this sufficient ? Could I select a differential opamp with higher input resistance ? What about using buffers on each input ? Is it then the same as an instrumentation amp ?
Instrumentation op amp
No current sensing examples are shown in the datasheets for instrumentationl opamps, only thermocoupler bridge. Is it overkill i my use ? Or not suitable at all ? Did I get something wrong ? CMRR problem ?
One of the problems with the differential amplifiers discusses above is the input impedance. If connected directly to the current sense resistor, especially the low current (>50ohm shunt resistance), the accuracy will suffer. I’ve been thinking of adding a buffer on each of the inputs to get high input impedance. But I guess I then actually have an instrumentation amplifier ? The major drawback is that the common voltage will the be limited by the max input of the opamps. This can be solved by using high voltage opamps such as LT6090. Example from the LT6090 datasheet:
But what about some “read-made” instrumentation opamps. I’ve not included anything that have common mode voltage lower that +/-15V (such as AD8230 and LTC6915). Many have +/-18V. Could be acceptable even though I initially have set the requirement of my SMU to +/-20V.
PGA281: +/-18V (5uV,max 20uV offset, 0.03%typ/0.15%max gain error, multiple gains) Example:10 µA-100 mA, 0.05% Error, High-Side Current Sensing
INA188 (from 2015?) +/- 18V. Gain Error: 0.007%, G = 1, Gain Drift: 5 ppm/°C (max) G = 1, but higher at higher gain! Low Input Offset Voltage: 55 μV (max) – Low Input Offset Drift: 0.2 μV/°C (max). The INA188 is a precision instrumentation amplifier that uses TI proprietary auto-zeroing techniques to achieve low offset voltage, near-zero offset and gain drift, excellent linearity, and exceptionally low-noise density…
ADA4254 (from 2019?) +/-28V !!! Input offset voltage: ±14 µV max, offset voltage drift: ±0.08 µV/°C max, gain drift: ±1 ppm/°C max, This one looks really nice. The ADA4254 is a zero drift, high voltage, low power programmable gain instrumentation amplifier (PGIA) designed for process control and industrial applications. Has a lot of programming possibilities and extras such as 7 GPIO ports with special functions. My only concern is the noise. Could be that I misinterpret the datasheet though.
PGA204/205 +/- 18V, 50-µV offset, offset drift: 0.25µV/°C, Gain programmable PGA204=1, 10, 100, 1000 (PGA205=1,2,4,8). Gain error (x10): ±0.01% for the best version. Gain vs temperature: ±2.5ppm
PGA103 The PGA103 is a programmable-gain amplifier for general purpose applications. Gains of 1, 10, or 100 are digitally selected by two CMOS/TTL-compatible inputs. The PGA103 is ideal for systems that must handle wide dynamic range signals. The PGA103’s high speed circuitry provides fast settling time, even at G=100 (8µs to 0.01%). Bandwidth is 250kHz at G=100, yet quiescent current is only 2.6mA. It operates from ±4.5V to ±18V power supplies.
Current sense amplifiers
High-Side Current Sensing: Difference Amplifier vs. Current-Sense Amplifier (2008) “…We will compare two high-voltage parts, the AD8206 bidirectional difference amplifier and the AD8210 bidirectional current-sense amplifier.Both devices offer the same pinout, and both perform high-side current-shunt monitoring, yet their specifications and architectures are different. So, how does one consider which device is best-suited for the application?”…. “Current-sense amplifiers with this architecture [AD8210] are generally useful only if input common-mode voltage remains above 2 V or 3 V, and if the application doesn’t require that the input common-mode voltage go all the way to ground (or below)….“ !!!!
Performance of Current-Sense Amplifiers with Input Series Resistors“When discussing functional operation, a current-sense amplifier can be considered an instrumentation/differential amplifier with a floating input stage. This means that even when the device is powered from a single-supply with VCC = 3.3V or 5V, it can amplify input differential signals at a common-mode voltage well beyond these power supply rails. The common-mode voltages in a current-sense amplifier can, for example, be up to 28V (MAX4372 and MAX4173) and 76V (MAX4080 and MAX4081).“
MAX9922/23 https://www.maximintegrated.com/en/products/analog/amplifiers/MAX9922.html “…ultra-precision, high-side current-sense amplifiers feature ultra-low offset voltage (VOS) of 25µV (max) and laser-trimmed gain accuracy better than 0.5%. The combination of low VOS and high-gain accuracy allows precise current measurements even at very small sense voltages.”. “The +1.9V to +28V current-sense input common-mode voltage range“…. Not usable for voltages closer to zero ?
LTC6102 10uV max offset (will not use, only unidirectional..).
LTC6103 has 450uV offset… Two unidirectional sense amplifiers can be combined into bipolar:
LTC6104 is bidirectional with higher offset: “…high voltage, high side, bidirectional current sense amplifier…. ±450µV maximum offset”
AD8217 High Resolution, Zero-Drift Current Shunt Monitor. Mentioned in: (High-Side Current Sensing with Wide Dynamic Range: Three Solutions (2010)) VCM:4.5 V to 80 V Not enough range for my use.
AD8418 Bidirectional, Zero-Drift, Current Sense Amplifier, CVM 2 V to +70 V
AD8206 difference amplifier for amplifying small differential voltages in the presence of large common-mode voltages. The operating input common-mode voltage range extends from −2 V to +65 V. Gain=20. Offset 2mV???? Too high…
LTC1787 “…delivers greater than a 12-bit dynamic range with ultralow 40µV input offset voltage compared to a typical 250mV fullscale input voltage. A fixed gain of 8 is set by onboard precision resistors”. Gain = 8. Se also: Sense Milliamps to Kiloamps and Digitize to 12 Bits – Design Note 227. Seems to be powered by the shunt voltage…. or is it different when biasing for bidirectional ???
lmp8601/2/3 -22 to 60V CM. Gain x20, x50 or x100. 0.5% gain error (a bit high?) “…devices are fixed-gain, precision current-sense amplifiers … The input common-mode voltage range is –22 V to +60 V when operating from a single 5-V supply….ideal parts for unidirectional and bidirectional current sensing applications.” Could be a candidate if I accept 22V min voltage…
MAX4081 76V, High-Side, Current-Sense Amplifiers is the bidirectional version of MAX4080 that Dave Jones uses in rev B of his uSupply. ±0.1% Full-Scale Accuracy. 5/20/60 x gain versions available. More info Maxim APPLICATION NOTE 3888. Accuracy is not very good…
PGA281 is a high-precision instrumentation amplifier with a digitally-controllable gain and signal integrity test capabilityWith ±15 V supplies the PGA281 can accept common-mode voltages of ±12.5 V, making it a suitable choice for high-side current sensing only as long as VBUS falls within the common-mode voltage range… Could be a candidate for a first version with reduced voltage span ?
EEVblog #373 – Multimeter Input Protection Tutorial where Dave Jones goes throug input protection on Fluke multimeters, based on the scematics in the Fluke 27 service manual.
From section 3.4 in 10 µA-100 mA, 0.05% Error, High-Side Current Sensing : “…protection for the system against ESD (electrostatic discharge), EFT (electrical fast transients), and surge (simulates a lightning strike). This protection is provided by a Schottky diode and two TVS (transient voltage suppressor) diodes. The BAT54-V-GS08 Schottky diode ensures that no current flows through the split supply when the power terminals are connected in reverse polarity. This diode protects against reverse voltages up to 30V……Since the split supplies to the PGA281 can reach up to ±18 V, the TVS diodes should have a breakdown voltage slightly higher than 18V. The diodes must also be bidirectional and should have a very fast response time in order to provide sufficient protection against fast transients. Based on these requirements the SMBJ20CA was chosen to provide up to 600 W of protection.“
Can this me used to protect 36V opamps? https://assets.nexperia.com/documents/data-sheet/BZB84_SER.pdf
Reverse Current/Battery Protection Circuits (2003): “…The most recent MOSFETs are very low on resistances, and therefore, are ideal for providing reverse current protection with minimal loss…“
APPLICATION NOTE 4035 Overvoltage Protection (OVP) for Sensitive Amplifier Applications (Maxim) “Most amplifier overvoltage-protection methods utilize diodes to shunt overvoltage fault current to ground or to the supply rails. These diodes exhibit capacitance and leakage current that contribute to distortion and limit bandwidth. This article explains the basics of reverse-biased diodes, discusses several protection strategies, and provides a few solutions for reducing parasitic leakage and capacitance…“
Differential Overvoltage Protection Circuits for Current Sense Amplifiers “One way to reduce the value of [input series resistors] is to add external protection diodes with higher current capabilities on the input pins…”
Robust Amplifiers Provide Integrated Overvoltage Protection “This article discusses some common causes and effects of overvoltage conditions, how cumbersome overvoltage protection can be added to an unprotected amplifier, and how the integrated overvoltage protection of newer amplifiers provides designers with a compact, robust, transparent, cost-effective solution…“
Prevent System Damage Via Fast, Accurate Over-Current Detection (2014)“An over-current detection solution can be easily created using a simple, low-cost operational amplifier (op amp), external gain setting resistors, and a low-cost comparator…“
From Agilent DC POWER SUPPLY HANDBOOK Application Note 90B ( http://literature.cdn.keysight.com/litweb/pdf/5952-4020.pdf):
- A. Varistor – A voltage-dependent resistor that protects preregulator triac against ac line spikes. Its resistance decreases abruptly if line voltage exceeds a harmful level. 39
- B. RFI Choke – Minimizes spikes at output of supply by slowing down turn-on of triac.
- C. Rectifier Damping Network – RC network protects other elements in supply against short-duration input line transients.
- D. Series Regulator Diode – Protects the series regulator against reverse voltages which could be delivered by an active load or parallel power supply.
- E. Slow Start Circuit – A long time-constant network that reduces turn-on overshoot and helps limit inrush current. When supply is first turned on, this circuit holds off both the series regulator (to reduce output overshoot) and the preregulator triac (to limit inrush current).
- F. Amplifier Input Clamp Diodes – Limit the maximum input to the amplifier to protect it against excessive voltage excursions.
- G. Output Diode – Protects components in the power supply against reverse voltages that might be generated by an active load or series connected power supply.
- H. Sensing Protection Resistors – Protect the load from receiving full rectifier voltage if remote sensing leads are accidentally open-circuited.
Small signal switching
I might have the need to adjust filters etc depending on measuring modes. This can typically be done by switching in and out capacitors and resistors. There are a lot of types of specialized analog switches for that.
Selecting the Right CMOS Analog Switch “The structure of a conventional analog switch is shown in Figure 1. Connecting an n-channel MOSFET in parallel with a p-channel MOSFET allows signals to pass in either direction with equal ease….“
Output switching (high current)
Do I need an output relay to protect the DUT (device under test) when I turn the power supply on ? A traditional relay ? Solid state relay? Just a mosfet ? The discussion is also relevant for how to switch between current ranges…
When it comes to output protection, a lot of info can be found in the audio society where they need to protect the speakers when amplifier turns on. Here is interesting read about solid state relay solution: MOSFET Solid State Relays (Rod Elliott 2012) “This article shows MOSFET relays for speaker protection, but there are countless uses for them in other applications…”
Looking at older designs they typically use relays for various switching. Is that because that was the only solution at that time ? Or are there other reasons?
I also read that relays typically have a “minimum applicable load” that can typically be 10uA. If I want to make a really precision instrument and measure below that, will relay still work? Why does a relay have a minimum applicable load?
Too keep it simple, I might just go for a normal relay… The Omron G5V-2 or something else. The manufacturer typically have selection guides such as Omron: Electrical Mechanical Relay PCB Relay Selection Guide.
What about reed relays? Seems like they don’t have minimum applicable load ?
There are some notable differences between reed relays and EMRs (electromagnetic relay) which users should be aware of (based on listing in http://www.epdtonthenet.net/article/151720/Comparing-reed-relays-with-electromechanical-relays.aspx:):
- EMRs typically have a lower contact resistance and higher ratings than reed relays. EMRs can also better sustain current surges.
- Reed relays generally exhibit much faster operation (typically between a factor of 5 and 10) than EMRs.
- Reed relays have more consistent switching characteristics at low signal levels and higher insulation values in the open condition.
- Reed relays have longer mechanical life (under light load conditions) than EMRs, typically between a factor of 10 and 100.
- Reed relays require less power to operate the contacts than EMRs.
Solid state relay / mosfet switch
Solid state relay are typically isolated and use a photodiode. I’ve read that they are slow with high on resistance, but it seems that there are solutions out there that looks promising.
Designing a circuit with bidirectional switching looks easy in theory. But what about in pracise? Seems there are various voltage limitations to be aware of. Reading: Using MOSFETs As General Switches “Using this back to back arrangement of P Channel mosfets, when on current will flow in either direction. When off both sides are isolated. You can use any typical P channel mosfet…. ….the P-channel MOSFET has an advantage over the N-channel MOSFET for soem applications due to the simplicity of the on/off control. A N-channel mosfet switching +V requires an additional voltage rail for the gate; the P-channel does not… …Note that this arrangement is only suitable if the voltage being switched is > Vgs switching threshold of the mosfet used…“
So, to get around these limitation, maybe I should use dedicated and/or isolated driver ICs. Seems like there are mosfet drivers with “integrated fast turn off” (53us turn on and 24us turn off) functionality such as VOM1271 – Photovoltaic MOSFET Driver with Integrated Fast Turn-Off, Solid-State Relay
Can that one be used with a set of corresponding mosfets ? >For example FDS8984 which has “been optimized for low gate charge, low RDSon and fast switching speed”. Or FDS8978 (slightly lower on resistance) used in Agilent B2912A.
Do I need fast switching speed? Or maybe a more “controlled” switching to avoid spikes when changing ranges?
“Photovoltaic” based driver, such as the VOM1271, are based on the use of an array of photodiodes connected in series used to generate a gate drive voltage. More info can for example be found in Using Photovoltaic MOSFET Drivers video by Lewis Loflin where TLP1918, PVI5010++, Avago ASSRV62x are also mentioned.
Seems at least VOM1271 combined with FDS8984 will have much faster switch time than the VO14642 solid state relay. Question is: does it matter in my application? Maybe a “slow” transition is better when changing current measurement range; in order to reduce any “glitches” in the voltage stability… Ref. what they seem to do in Keithley 236/237 according to the service manual: “…by using circuitry that, in effect, shunts the current ranges with a 100ohm resistor whose voltage drop ramps in a linear fashion towards 0V. This slow rate of voltage change allows the system to adjust itself keeping the output voltage constant (spike free)…”.
The FDS8984 also have much lower on-resistance compared to the built-in mosfets in VO14642. I might still want to put a relay in parallel to keep the resistance value as close to zero as possible.
For switching smaller currents, there are also integrated solutions such as PhotoMOS AQY221R2S used by Dave Erickson to switch the 100mA range in his DIY SMU.
I need to take a closer look at the PhotoMOS series from Panasonic and see if some of them can be applicable for apprx. 2A with low on resistance and fast switching….
DACs – Digital to analog conversion
I need at least 1 bipolar DAC for main output and one unipolar for limit. Would also be nice to be able to have different limits for positive and negative limits which required to DAC channels.
I was first thinking about 16bits resolution. For voltage that means max 305uA resolution for a swing of -10 to 10V. I want at least 100uA resolution. This means that I should go for at least 18 bit (76uV resolution), maybe 20 (20uV resolution). Maybe I’ll divide operation into two ranges. For example 0-2V with x uV resolution and 0-25V with y uV resolution.
I don’t want the converters to be the main limiting factor from a resolution perspective.
Linear Technology http://www.linear.com/product/LTC2704 bipolar 4 channel 16 bit converter. Linear DAC product selector: http://cds.linear.com/docs/en/product-selector-card/2PB_dacsfc.pdf “…six output spans—two unipolar
and four bipolar…SPI serial interface. INL is accurate to 2LSB[LTC2704-16]. DNL is accurate to 1LSB for all versions“
Analog Device http://www.analog.com/media/en/technical-documentation/data-sheets/AD5764.pdf. AN1411 application note: High Accuracy, Bipolar Voltage Output Digital-to-Analog Conversion Using the AD5764 DAC “… integrated output amplifiers, reference buffers, and proprietary power-up/power-down control circuitry...digital offset and gain adjust registers per channel… guaranteed monotonicity, integral nonlinearity (INL) of ±1 LSB, low noise, and 10 μs settling time. “. . Others experiments with this: https://www.laserlance.com/projects/arduino-dac-library-and-shield/ (shared PCB at oshpark: https://oshpark.com/profiles/Laser-Lance. ±2 ppm FSR/°C max.
AD5765 Complete Quad, 16-Bit, High Accuracy, Serial Input, ±5 V DAC. More or less same as AD5764 except only 5v ????
AD5763 2 channel 16 bit bipolar
Try this parametric search at analog.com: 16bit <2LSBINL bipolar DAC
AD5781 True 18-Bit, Voltage Output DAC ±0.5 LSB INL, ±0.5 LSB DNL (recommended for new design)
AD5791 20 bit. Arduino shield: http://41j.com/blog/2015/10/ad5791-board-rev4/. Evaluation kit: http://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/EVAL-AD5791.html. Note that there are two revisions of that evaluation board. Last post in https://www.eevblog.com/forum/metrology/an-arbitrary-output-voltage-divider-for-a-precision-voltage-reference/ shows a video of a “slapped-together AD5791 + LTZ1000”
The following are DACs that I have tested:
AD5761 (single channel)
I also tried this 1 channel DAC, mainly because it’s recommended for new designs in the analog webpage and I got some samples from Analog Devices. Wasn’t too difficult to get it up’n running. In my crude setup I found that it drifts some quite a few uVs. Note that the temperature coefficient in bipolar mode is +/-15uV. Might be a bit much for my application.
Code for arduino at Github.
DAC8734 quad bipolar 16 bits dac https://www.youtube.com/watch?v=uYl_jqJeGtA
I really wanted to use this DAC. But I never managed to “tame” it. More about by problems here: DAC8734 drawing too much current from analog supply and gives completely wrong output values. My theory is that I, at one or several stages, has done something wrong and “killed” the device.
How accurate? 16 bits ? Is it possible to “control” drift etc by measuring with a 24bit AD converter and (slowly) compensate drift in software, such as described in A Standards Lab Grade 20-Bit DAC with 0.1ppm/°C Drift (2001)
What about LTC2758, Dual Serial 18-Bit SoftSpan IOUTDACs ? Or the one channel version LTC2756 of it ? This one fits nicely into the concept having different ranges with different resolution.
A video presentation of LTC2756 (single channel version).
Comparing LTC2758(18bit) and AD5791(20bit), datasheet indicates:
- 1 ppm resolution 1 ppm
- INL Error ±0.25LSB (max +1 LSB)
- <0.05 ppm/°C temperature drift
- 1 µs settling time
- Wide power supply range up to ±16.5 V
- 20-lead TSSOP package
- INL Error: ±0.5LSB (max ±1 LSB)
- Program or Pin-Strap Six Output Ranges: 0V to 5V, 0V to 10V, –2.5V to 7.5V, ±2.5V, ±5V, ±10V n
- Settling Time: 2.1µs
- Bipolar Zero Temperature Coefficient ±0.2 ppm/°C
- Gain Error Temperature Coefficient ∆Gain/∆Temp ±0.25 ppm/°C
- Gain Error All Output Ranges ±6 (max ±32 LSB)
- DNL Differential Nonlinearity ±0.2LSB (max ±1 LSB)
- 2.7V to 5.5V Single Supply Operation
- Voltage-Controlled Offset and Gain Trims
- 48-Pin 7mm × 7mm LQFP Package (0.5mm between pins)
As we see, the AD5791 has better specs. But is it overkill for the rest of my circuit ? One nice ting is that it accepts high reference voltages so a buffered LM399 or LTZ1000 can drive be used directly without voltage booster. However, it needs both positive and negative reference to reach its full potential. With the LTC2758 I get 2 channels that will cover both voltage output control and current limit control. Will 18 bit be enough?
I’ve briefly tested the LTC2756 with Teensy 3.2. Based on the official Linduino code for LTC2758. Code in github: https://github.com/hellange/ltc2758test
I also ended up using the LTC2756 in my first prototype. It’s easy to program, is bipolar and have several ranges. Seems stable.
Texas Instruments have a new 20 bit DAC11001A Seems it accepts a higher reference voltage that “normal”. This means that one could use an LM399 apprx 7V reference directly without any voltage divider circuitry.
ADC – Analog to digital conversion
I want a bipolar 24 bit ADC. Preferably with as a high voltage as possible to avoid too much voltage scaling around in the circuit. The better ones can typically give real 20bits resolution at low conversion rate.
I’ve not find any good information that indicates if sigma-delta ADCs are better suited for my purpose than SAR ADCs.
The 24 bit AD7175 and AD7176 look interesting. “…fast settling, highly accurate, high resolution, multiplexed Σ-Δ analog-to-digital converter (ADC) for low bandwidth input signals” +/-2.5V or 5V supply. Bipolar.
I’ve briefly tested the AD7176 with arduino: AD7176 meets Arduino (January 2019) and the AD717x seems to be good enough for my application. I’ve also later tested it with external ltc6655 reference and AD8475 buffers that gave even better results. Same configuration as shown in the datasheet, except using LTC6655 reference instead.
An interesting article about what can be accomplished with a 32bit ADC. ADS1262, 32-bit ADC Eval kit from TI seems to provide a starting point for an inexpensive voltage measurement platform: https://xdevs.com/review/ti_ads1262_p1/ https://xdevs.com/review/ti_ads1262_p2/
The article also shows schematics for a stable voltage reference:
Would be nice to have fast ADC even if DC is main priority. With a fast ADC it can be possible to implement a “digitizer” that can be used to inspect fast transitions etc. A poor man’s oscilloscope…
LT2380-24 SAR ADC looks interesting.
In order to obtain accurate values, there is a need for a voltage reference. A nice overview of various voltage references: https://xdevs.com/review/dcvref_table/, for example unheated ones:
What seems to the most used ultra stable voltage references are LTZ1000/LTZ1000A and LM399. XDevs.com has a lot of insight, especially for the LTZ1000, based on the KX-REF project or the later 792x projects.
What’s All This LTZ1000 Stuff, Anyway? “Though appearing simple upon first glance, the venerable LTZ1000 voltage reference has more complexities and subtleties than one might think. Paul Rako takes a closer look”
The LM399 is used in http://www.ianjohnston.com/index.php/videos/20-video-blog-022-handheld-precision-digital-voltage-source together with a 18 bit DAC. I guess this means the LM399 should be sufficient for my use. Several sources indicate that it needs some “burn-in” to become stable… According to Ian Johnston, the LM399 “totally transformed the unit“: https://youtu.be/-bkINyHdLG4?t=185 .
Looking at the LM399 datasheet there is a “portable calibrator”
I had problems understanding this circuit because of the 5K resistor between output and the noninverting opamp input. I found the answer here: https://electronics.stackexchange.com/questions/329661/why-is-there-an-extra-resistor-on-the-opamp-in-the-lm399-datasheet
According to an entry in https://www.eevblog.com/forum/metrology/influence-of-resistors-in-lm399-reference-circuit/, this is the schematics for the 34401 using LM399:
The voltage reference circuit is explained in the 34401 service manual: “The instrument precision voltage reference is U403. Resistor R409 provides a stable bias current for the reference zener diode. R408 and CR404 provide a bias to assure that the reference zener biases to +7 V during power up. IC U400A amplifies the voltage reference to +10 V while amplifier U401A inverts the +10 V reference to –10 V. The reference voltages force precision slope currents for the integrating ADC through U102E–R17, R18. Amplifier U401B provides a precise +5 V reference for the U500 on chip ADC.“
I’ve been looking at the LTC6655. However, according to a later discussion in the eevblog forum, it seems to have less that ideal long term stability: https://www.eevblog.com/forum/metrology/ltc6655b-long-term-drift/.
Long term characterization of voltage references (2013) compare AD587UQ, ADR445BRZ, ADR435BRZ, MAX6126AASA50, LTC6655BHMS8-5, MAX6350CSA+ LT1021BCN8-5, VRE305AD for eLisa, a space-based project aiming at detecting gravitational waves(!). According to the test: “The requirements of relative output stability of 1 ppm/√ Hz down to 0.1 mHz were not met by any of the tested devices, but 4 references approaches the objective : the AD587UQ, the MAX6126AASA50, the LT1021-BCN8-5 and the LT6655BHM“.
How to Choose a Voltage Reference (2009) “…to measure voltage, you need a standard to measure against. That standard is a voltage reference. The question for any system designer is not whether he needs a voltage reference, but rather, which one?..“. Example table:
Some general insight/opinions regarding references e.g. here: https://www.eevblog.com/forum/projects/dc-voltage-references-battle-overview-of-market/
Voltage references hold steady (2010) “Decades ago, these references provided initial accuracies of only ±10%, whereas modern reference ICs can provide initial accuracies of 100 ppm, or 0.01%“
Will the right voltage reference stand up? “In a perfect world, you are done as you find the IC chip with the correct output voltage for your ADC. However, in our non-perfect world the voltage reference chip has initial output errors, and an inherent inability to drive the ADC’s reference pin directly….“
build a .01% accurate voltage reference ” After much head scratching, you measure the A/D voltage reference with your trusty DMM and get to thinking. A 12-bit A/D means the input signal is going to be represented by one of 4,096 possible codes; one part in 4,096 is approximately .024%. In other words, if you want to get 12 bits of absolute accuracy, the reference voltage must be within .024% of the desired value. This puts a heavy burden on the test equipment used to measure the reference voltage...The construction part of this article is geared towards voltmeter calibration and/or accuracy verification. However, the following discussion of various reference voltage parameters applies equally well to using a voltage reference as a circuit element…”
Tips and tricks for designing with voltage references (Texas Instruments, 2017) discusses voltage reference essensials, topologies, performance, design and pcb design guidelines
Keithley DMM7510 uses another reference, LT-FLU, according to this teardown: https://youtu.be/uvgJ2zAxgAY?t=2026. Haven’t read too much about that one, but there is a discussion here: https://www.eevblog.com/forum/metrology/the-ltflu-(aka-sza263)-reference-zener-diode-circuit/.
Avoid REFERENCE PITFALLS (from http://www.analog.com/media/en/technical-documentation/application-notes/an82f.pdf):
- Current-Hungry Loads “…best performance is not obtained by running the reference at maximum current“
- Board Leakage “…board leakage caused by the residues of water-soluble flux… …tightly packed circuit board may leave no choice but to agglomerate incompatible traces. In this case, use a guard ring to eliminate reference shift“
- “NC” Pins “…For some ICs, “NC” means “this pin is floating, you can hook it up to whatever you want.” In the case of a reference, it means “don’t connect anything to this pin.” That includes ESD and board leakage, as well as intentional connections. External connections will, at best, cause output voltage shifts and, at worst, permanently shift the output voltage out of spec.”
- Trim-Induced Temperature Drift “…Trimming may be necessary to calibrate the system, but it can also adversely affect the tempco of the reference…”
- Burn-In “…relieves stresses built into the reference and circuit board during assembly and it ages the reference beyond the highest long-term drift region, which occurs when power is first applied to the part“
- Board Stress “Circuit boards are not perfectly elastic, so bending forces may cause permanent deformation and a permanent step-change in reference output voltage…“
- Temperature-Induced Noise “Even though references operate on very meager supply currents, dissipation in the reference is enough to cause small temperature gradients in the package leads“
Would also add humidity. Even packaging is relevant. Here is a graph showing long term drift for LTC6655 (from https://www.eevblog.com/forum/metrology/ltc6655b-long-term-drift/25/ :
What kind of voltage regulators should I use to power opamps etc. ? Are standard 780x ok, or should I go for some more advanced…
A Guide to Choosing the Right Ultra – Low IQ Low Dropout Linear Voltage Regulators“The following paper discusses the tradeoffs between achieving low [quiescent current] and good dynamic performance when choosing an LDO…“
How to pick a linear regulator for noise-sensitive applications “A lownoise power solution is essential to preserving signal accuracy and integrity. This article addresses criteria and parameters to consider in designing such a power solution, including important specifications for picking a linear regulator“
AN159: Measuring 2nV/√Hz Noise and 120dB Supply Rejection on Linear Regulators – The Quest for Quiet“Low noise amplifiers and analog-to-digital converters (ADCs) do not have infinite supply rejection and the cleaner the regulator output is, the higher their performance. These are just a few applications where linear regulators are required to provide quiet power supply rails, but how does one ensure that the regulator is performing as advertised?“
Power Management for Precision Analog (TI) using linear 36V 150mA/200mA negative and positive regulators TPS7A30 and TPS7A49. “The TPS7A30 family is designed using bipolar technology, and is ideal for high-accuracy, high-precision instrumentation applications where clean voltage rails are critical to maximize system performance. This design makes the device an excellent choice to power operational amplifiers, analog-to-digital converters (ADCs), digital-to-analog converters (DACs), and other high-performance analog circuitry”.”…TPS7A30/49 development kits remove switching noise and increase the performance of data converters, operational amplifiers, clocks and other signal chain devices…” (http://embedded-infos.blogspot.no/2011/02/texas-instruments-introduces-new.html)
TPS7A33 /TPS7A470 –36V & 36V 1A Ultralow-Noise Negative Voltage Regulator. Noise:16 μ/4μVRMS, PSRR:72 dB”…ideal to power operational amplifiers, analog-to-digital converters (ADCs), digital-to-analog converters (DACs), and other high- APPLICATIONS performance analog circuitry“
ADP7118 20 V, 200 mA, Low Noise, CMOS LDO Linear Regulator. Noise:11uV, PSRR of 88 dB at 10 kHz. “This high input voltage LDO is ideal for the regulation of high performance analog and mixed-signal circuits operating from 20 V down to 1.2 V rails.“
ADP7142 40V, 200mA
ADP7182 –28 V, −200 mA, Low Noise, Linear Regulator “This high input voltage LDO is ideal for regulation of high performance analog and mixed signal circuits operating from −27 V down to −1.2 V rails.“
ADP5070 switchmode preregulator ++ ( http://www.analog.com/en/design-center/evaluation-hardware-and-software/evaluation-boards-kits/eval-adp5070.html, “For more details about the dc-to-dc converters, refer to the ADP5070 and ADP5071 data sheets. For further information on the LDO regulators, refer to the ADP7142 and ADP7182 data sheets. These data sheets must be used in conjunction with this user guide when using the evaluation board. “) together with ADP7142 and ADP7182
LTC3265 dual supply: https://www.digikey.com/en/product-highlight/l/linear-tech/ltc3265-low-noise-and-high-voltage-dual-supply-with-boost-inverting-charge-pumps. “LTC3265 is a low-noise dual polarity output power supply, which includes a boost charge pump, an inverting charge pump, and two low-noise positive and negative LDO post regulators. The boost charge pump powers the positive LDO post regulator while the inverting charge pump powers the negative LDO regulator. Each LDO can provide up to 50 mA of output current. The LDO output voltages can be adjusted using external resistor dividers“
LT3042 ” high performance low dropout linear
regulator featuring LTC’s ultralow noise and ultrahigh PSRR
architecture for powering noise sensitive RF applications” (solution:http://www.linear.com/solutions/5762)
LT3094/LT3045 “LT3094 (2018), an ultralow noise, ultrahigh power supply ripple rejection (PSRR), low dropout voltage, 500mA negative linear regulator that targets applications requiring the lowest noise performance… …The LT3094 is a complementary negative version of the 20V, 500mA, 0.8µVRMSLT3045 ultralow noise LDO and the two devices when used together combine for an ultralow noise bipolar power supply solution…“
Voltage regulator noise experiments: Simple circuits reduce regulator noise floor (EDN)
Do I need to control/reduce the slew rate ? http://www.ti.com/lit/ug/tidu026/tidu026.pdf
I believe it matters, even if the main goal is relatively low bandwidth. I have experimented with high precision opamps with relatively low slew rate (apprx. 1.4V/us). I get into problems if the voltage rises to high too fast, such as in a square wave signal. In an earlier design i tried to keep the voltage level 1:1 thoughout the whole control loop using a ADA4522 high voltage opamp. The idea was to avoid expensive precision voltage dividers and gain stages. Testing with square wave input showed that ADA4522 could not keep up with the speed (read: high rise time). I ended up dividing the output with a factor of 10 in the feedback loop. That fixed the problem. Added benefit is that I can now use lower voltage in the control loop which gives me even more (and better) alternative precision opamps. The ADA4522 is nice, though. Cons: Added cost for precision divider/gain stages.
I also ended up replacing ADA4522 in the current limit stage with faster (but lower precision) opamp with good results. The limiting shall happen as fast as possible.
Stability and avoid oscillation
Noise, phase shift, capacitive load and its affect on stability and oscillation is discussed alot of places. I’ve experienced it myself. I probably need to read up on phase margin, bode plots etc. (but note the “slew rate” section that the slew rate of the opamp also seems very relevant).
Phase Relations in Active Filters “…but in some applications, the phase response of the filter is important. An example of this might be where a filter is an element of a process control loop. Here the total phase shift is of concern, since it may affect loop stability…“
AN-1148 Linear Regulators: Theory of Operation and Compensation“...All voltage regulators use a feedback loop to hold the output voltage constant. The feedback signal experiences changes in both gain and phase as it goes through the loop, and the amount of phase shift which has occurred at the unity gain (0 dB) frequency determines stability...”
Why Op Amps Oscillate—an intuitive look at two frequent causes (2012) and Taming the Oscillating Op Amp (2012) and Taming Oscillations—the capacitive load problem (2012) “A simple non-inverting amplifier can be unstable or have excessive overshoot and ringing if the phase shift or delay created by the op amp’s input capacitance (plus some stray capacitance) reacting with the feedback network resistance is too great…”
Do-it-yourself: Three ways to stabilize op amp capacitive loads (2017) “Many resources present basic stability theory in great detail, including TI Precision Labs’ videos on op amp stability. There are different compensation circuits which allow the op amp to remain stable while driving the capacitive load. In this blog post, I’ll review three common compensation circuits that can be designed and tested using the do-it-yourself amplifier evaluation module (DIYAMP-EVM).”
Practical Techniques to Avoid Instability Due to Capacitive Loading (Analog Dialogue) “To avoid sacrificing performance with light loads, most amplifiers are not heavily compensated internally for substantial capacitive loads, so external compensation techniques must be used to optimize those applications in which a large capacitive load at the output of the op amp must be handled.“
Op amps with capacitive load capabilities, ex: lm8261(single +/-30V), LM8272(dual, +/-24V). More about them here: Unlimited Capacitive Load Drive Op Amp Takes Guesswork Out Of Design
LTC2057, or other high voltage opamps from linear: http://cds.linear.com/docs/en/product-selector-card/2PB_6015f.pdf ?
“So whereas an IC designer would say, “An output capacitor puts a pole in the ac response,” [Bob] Pease would use the time-domain equivalent by saying, “An output capacitor puts a lag in the ac response.” So when you realize the capacitor is putting in a delay, you can start to see why the system will tend to oscillate.” – http://www.electronicdesign.com/power/what-s-all-capacitive-loading-stuff-anyhow
Ask The Applications Engineer-25: Op Amps Driving Capacitive Loads describes noise-gain manipulation, out-of-loop compensation, in-loop compensation
A good discussion of how to obtain stability in an electronic load. https://www.eevblog.com/forum/projects/dynamic-electronic-load-project/25/. Should also be relevant for power supply (especially a 4 quadrant power supply!)
https://electronics.stackexchange.com/questions/400522/constant-current-source-with-mosfet-opamp-and-instrument-amp “After plugging everything back in and switching it on again the oscillation had vanished! There is still some small oscillation on the op amp output but it is almost non-existent. I will upload the updated schematic and comment further if I make further improvements. Thank you all for the help!“
Frequency compensation types
Introduction to Control Algorithms in Switching Regulators mentions three types of error-amplifier frequency response tailoring or “compensation” types. Even if the article is about switching regulators, I believe it’s relevant for linear regulators as well…
Type 1: “…would keep some loops stable, but with very poor loop bandwidth…“
Type 2: “…current-mode converters… …the effect of this is to extend the useful bandwidth of the loop...”
Type 3: “…used for voltage-mode control… …are placed to keep gain high and minimize phase delay to as high a frequency as possible. With so many variables to play with, different pole-zero placement schemes can be devised to optimize results under differing conditions…“
Current limiting / clamp
https://www.eevblog.com/forum/projects/current-limiting-stability-problems/ “…my voltage control works fine, but I have problems with my current control. I use the schematics of David’s power supply with an LT3083, but with an ASC712-5 current sensor. Everything works fine…but when it limits the current, the voltage goes down and the current as well….an oscillating current control loop.“
ADD CURRENT LIMIT TO THE BUF634 (2000) “…For stability, the bandwidth of A2 must be less than approximately one-fourth the bandwidth of A3, and R1 C1/(2π) must be less than approximately one-fourth the bandwidth of A2…“
https://www.eevblog.com/forum/beginners/supply-lm350t-instead-of-lt3080/ Some info about how to “OR” the current and voltage control signal: “……The problem with the transistor as shown is that it adds voltage gain to the error amplifier for the current control loop making it much more difficult to frequency compensate…. Usually this is done with a pair of diodes but in the past when they were more available like the 2N404A, PNP emitter followers with high Vbe breakdown could be used. Or use diodes and PNP emitter followers… “
http://wahz.blogspot.nl/2015/06/lab-power-supply-current_13.html and related posts discusses various approaches to current limiting. Trial and errors.
https://www.eevblog.com/forum/projects/limiting-op-amp-output/ discusses various “clamping”/”limiting” methods. Also mentioning the AD8036/37 Voltage Feedback Clamp Amps. Here is one of the proposed/discussed circuits:
Precision rectifier/clamp works near 0V for single-supply circuits “…the transistor has put a lot of extra gain inside the loop, so, to avoid oscillation, I would put a 100pF capacitor between the op-amp’s output and its inverting input – compensation experts might have a more elegant solution.“
Output stage / current booster
Can we get away with an op amp driving a MOSFET output stage directly? What about stability, ringing, overshoot, gate capacitance…
How to Make Linear Mode Work “Theoretically, linear mode operation is very easy. Simply bias the gate to deliver some desired amount of current or power, and stay within the manufacturer’s Forward Safe Operating Area (FSOA)… The reality however is that linear mode operation is one of the trickiest power applications of all, turning many “simple” designs into a reliability nightmare“
AN-272 Op Amp Booster Designs (2013) “This application report describes the output “booster,” or post amplifier, designs required to achieve needed voltage or current gain in applications that require substantially greater output voltage swing or current (or both) than IC amplifiers can deliver“
How to Buffer an Op-Amp Output for Higher Current (2016)“There is a special category of high-output-current amplifiers, with current capability approaching or even exceeding 1000 mA. If a high-output-current part is compatible with your application, by all means use it. But there are a few reasons why you might prefer to buffer the output of a more general-purpose amplifier….“
Designing a *linear* MOSFET driver stage (discussion) “I’m looking for a MOSFET driver circuit that can be placed between an op-amp and a power MOSFET to operate the transistor as a linear amplifier… …I suspect a biased push-pull BJT output driver would work fine- maybe 4 small BJTs (2 connected as diodes) a couple bias resistors plus maybe a couple ohms each of emitter degeneration… ” Use lm8261 (Single RRIO High Output Current & Unlimited Cap Load Op Amp) to drive a pushpull mosfet output stage ?
ADA4870 High Speed, High Voltage, 1 A Output Drive Amplifier.It’s a current feedback construction. What does that mean in my app, can it still be used? “By placing the ADA4870 inside the feedback loop of the ADA4637-1, the composite amplifier provides the high output current of the ADA4870 while preserving the dc precision of the ADA4637-1. “:
More about composite ADA4870 in https://www.analog.com/en/analog-dialogue/articles/a-large-current-source-with-high-accuracy-and-fast-settling.html
ADA4700-1 is an interesting high voltage opamp. Can it be used to drive MOSFETs directly ? The datasheet discusses various compensation that can be used for driving capacitive load: “Although the ADA4700-1 behaves well when driving capacitive
loads, CL, as seen in Figure 27 to Figure 30, extra compensation
can improve the response when large capacitances need to be
What about APEX amps ? Such as APEX PA12. Not sure if offset and tempco is good enough… What about combining precision op amp with APEX booster stage such as APEX PB50. APEX application examples: https://www.apexanalog.com/resources/appnotes/an14u.pdf: “The programmable power supply (PPS) application is useful to demonstrate the versatility of the PB series boosters. Along with the need to supply high voltages and currents, programmable power supplies often need high accuracy and low drift, while at other times they may need to be fast-responding. The PB series allows the designer to optimise the circuit for these choices…“. Problem is that they are extremely expensive: >$100 !!! I tried to ask for a sample but my request was rejected 😦 . For a one-of prototype it could potentially save design time and reduce complexity….
Could I simply use a LTC6090 140V (50mARMS) Op Amp with a audio amplifier MOSFET output as indicated in the datasheet ? Using LT1166 as booster ? “The LT1166 is ideally suited for driving power MOSFET devices because it eliminates all quiescent current adjustments and critical transistor matching. Multiple output stages using the LT1166 can be paralleled to obtain higher output current” (LT1166 is also used in Keithley 2601)
AN-1645 LM4702 Driving a MOSFET Output Stage , page 14:
The article also compares driving a mosfet output stage directly and with an added driver stage.
What about a high power opamp such as OPA549 ?
Do I need to bias to reduce crossover problems, or is it not relevant for a DC supply ? Learn more !
https://electronics.stackexchange.com/questions/42927/purpose-of-these-two-transistors shows this configuration (for a headphone amp):
Fundamentals of MOSFET and IGBT Gate Driver Circuits explains a lot about mosfets, driver circuit, optimize turn off, turn on etc. Here is one of the diagrams:
Using LT1166 “…bias generating system for controlling class AB output current in high powered amplifiers… The LT1166 is ideally suited for driving power MOSFET devices because it eliminates all quiescent current adjustments and critical transistor matching…”
In Increase Amplifier Output Drive Using a Push-Pull Amplifier Stage they simply add a 1k resistor to improve linearity…
” The 1k resistor between the transistor bases and emitters allows the LT6020 to provide the first 600µA of load current before the NPN and PNP start providing additional current. This resistor improves linearity in the middle of the range as the transistors turn off and on.“
Which precision opamp
I need some precision opamp in the control loop (error amplifier, current limit etc.). Focus is DC accuracy (low offset and drift). As long as the loop is stable and have reasonably good transient response, I don’t need high bandwidth in the first revision. I have a gut-feeling that it’s difficult to obtain everything at the same time…
ADA4522 5uV offset, 55V – looks promising ! Also “recommended for new designs”. Is also available as single or quad versions! Extremely low offset voltage drift: 22 nV/°C maximum. Slewerate (1.4V/s) is not that great.
ADA4528 2.5uV offset. Low offset voltage drift: 0.015 μV/°C maximum (better initial offset, slower slew rate (0,45V/us) than ADA4522, only 5.5V. Can work as error amplifier if gain is somewhere else)
LTC2057HV 4uV, 0.45V/μs, 60V (single opamp)
Is slew rate relevant ? It’s a DC supply, but it should handle transients…
Precision opamp (Vos ≤1mV & TCVos ≤2uV/C) Selector guide from Analog Devices: Precision Op Amps (Vos ≤1mV & TCVos ≤2uV/C)
Opamp to drive the output stage:
LT1636 ” Unlike most micropower op amps, the LT1636 can drive heavy loads; its rail-to-rail output drives 18mA. The LT1636 is unity-gain stable into all capacitive loads up to 10,000pF “
How to choose an operational amplifier (Nuts & >Volts Magazine) “…so why not call [input offset voltage (Vos) ] the output voltage error? This is because the gain of the circuit will affect the output. A circuit with a gain of 100 will increase this error by a factor of 100. That is why this is referred to the input rather than to the output…“
Which transistor in the output stage
I’ve seen both bipolar and mosfets used in output stages. I’ve decided to focus on mosfets transistors. Without any particularly good reason. Main problem is which to chose when operated in linear mode…
https://www.allaboutcircuits.com/technical-articles/how-to-buffer-an-op-amp-output-for-higher-current-part-1 contains some discussions about pros and cons: “I often wish that there were a concise, definitive answer to the old question, “Which are better, BJTs or MOSFETs?” But as you probably know, this is like asking, “Which are better, cars or trucks?” With both questions, there is no universal answer; rather, the details of each situation determine which option is preferable. In this article we will explore the use of MOSFETs in the specific context of buffering an op-amp output for higher current, and in the process we will be able to form a general idea of when MOSFETs might be preferable to BJTs, and vice versa.“
AN-1645 LM4702 Driving a MOSFET Output Stage (2013) contains info about distortion in different mosfet combos. Also a bit on output stage in general including gate resistor choice and snubber (section 7).
MOSFETs Withstand Stress of Linear-Mode Operation “Die temperature variations can trigger catastropic failure in linear-mode operation…“
Linear MOSFET and Its Use in Electronic Load (Kerry Wong, 2016) “I tested a few of the IRFP150N’s individually in linear mode and they all failed when the power dissipation reached between 45W to 60W… There is a class of so called linear MOSFET (e.g. IXYS‘ Linear L2™ MOSFETs) which is specifically designed to operate in linear region with an extended FBSOA…“
https://hackaday.io/project/21029-low-cost-electronic-load briefly discusses the use of special (expensive) mosfets for linear operation: “…The datasheet states that this MOSFET is specifically designed for linear operation and that it has a “Guaranteed FBSOA”, or Forward-Bias Safe Operating Area, at 75 degrees Celsius. This makes it a perfect fit for an electronic load… However, these features do come at a cost: compared to the price of a single IRF540 ($1, Digikey), these special linear-MOSFETS from IXYS are pricey ($12, Digikey)… …references to other electronic load circuits (see EEVblog, Kerry Wong 1, Kerry Wong 2, Mightwatt)“
Some mosfet data
If I want to drive a mosfet with an opamp, should the gate capacitance be as low as possible ?
From https://electronics.stackexchange.com/questions/187563/designing-a-linear-mosfet-driver-stage: “…since appearing in earlier questions I have replaced the MOSFET with the smallest capacitance device I was able to find (IRF530N -> IRFZ24N)”
- IRFZ24N input capacitance 370pF
- IRF530N input capacitance 920pF
- IRF9Z24N input capacitance 570pF
- IRF9530N input capacitance 860pF
Seems to be differences between the N and non-N versions…
What about “linear” mosfets ? IXTK90N25L2 ? High gate capacitance ? EXPENSIVE !
Currently I use the IRFZ24N / IRF9Z24N combo in a prototype (However, replacing them with others with higher capacitance does not seem to be a problem in my circuit. I need to investigate if there are other parts of my circuit that forces me to limit the bandwidth).
What about the “logic level” versions of them? Is the only difference a lower GS voltage ? Will that help me in any way ? Investigate…
What about IRFP240/IRFP9240 pairs? I’ve seen several reference to them specially in audio application. They come in TO-247 packages. Should be better that TO-220 from heat transfer point of view. But they have higher capacitance that the IRFZs. The IRFP250 is used in the Rigol DL3021 electronic load as indicated the the eevblog teardown @17:17 and BK Precision 8500 as indicated in the eevblog teardown @15:36. IRFP250 are rated a bit higher from an effect perspective but has double the capacitance of IRFP240. Also, the IRFP240 has a p-channel counterpart IRFP9240. The 250 does not ? Siglent SPD3303x power supply gets away with a IRFP150 as indicated the eevblog teardown @20:29 . According to this teardown the Keithley 236/237 SMUs uses IRF630/IRF9630, but those SMUs only go up to 100mA. Keithley 2302 battery simulator uses IRFP240/IRFP9140 pair as indicated in the eevblog teardown @16:17. Siglent SPD1168X uses IRFP150N in teardown @32:10.
I might try to test and compare the IRFP240/IRFP9240 and IRFZ24N/IRF9Z24N and see if the input capacitance difference matters in my application… Or the 140 instead of 240. I do not need 200V rated MOSFETS…
In general I find it hard to find good recommendations/tests of which mosfets to in linear mode. I guess I just have to try different types in my specific application. I’ve read some places that older mosfets are better suited for linear operation. Newer generations are tailored towards switching. I don’t know…
SOA (Safe operating area) and thermal resistance ?
To make a reliable product which does not break down after a short while, maybe I should focus more on linear mode SOA and thermal resistance when selecting transistor. Need to explore this more !! Just comparing IRFP240 with IRFZ24N seems to give some indications:
IRFP240 to the left and IRFZ24N to the right. Looks like the IRFP240 is much better. Also the IRFP240 has a junction case coefficient of max 3.3 vs IRFZ24N max 0.83. Maybe comparing the two is not fair. They have different specs and even different casing. But it shows the importance of taking these specifications into account when selecting transistors. However: Maybe it’s not a problem at 20watt (20V@1A)…
Also not that the SOA spec does NOT include DC operation… I would avoided such devices if it wasn’t for the fact that several commercial power supplies/loads actually uses them (at least the IRFP2x0) !
Parallel mosfets in linear mode
I was originally trying to use only one N and one P mosfet at the output in order to keep things simple. I found, however, that a single transistor might get too hot. Sure it’s possible to use several in parallel to increase current capabilities and distribute effect, but it add complexity especially when operating in linear mode.
AN11599 Using power MOSFETs in parallel (2015) From section 5: “If a group of MOSFETs must operate in linear (partially enhanced) mode, great caution is needed. MOSFETs simply paralleled together as they are for fully enhanced conduction are very unlikely to share power or current well.“
I’ve briefly tested two in parallel. I used source resistors for each. My initial attempt was without separate base resistors. That was unstable. Adding base resistor fixed that problem. That corresponds to what’s recommended in the figure above. I’ve seen some suggests separate drivers for each base as well, but I used a single opamp driver. I might end up with two transistors in parallel.
As mentioned, I’ll probably go for MOSFET. However, note that in the eevblog teardown of Atten PPS3205T-3S the transistor bipolar 2SD1047 is supposed to be specifically made for these kinds of applications with “wide ASO because of on-chip ballast resistance”.
Also, the new Rohde & Schwarz uses two? TIP147/TIP142 pair in their NGM202 two quadrant supply according to https://goughlui.com/2019/08/20/rs-ngm202-in-depth-ch6-teardown/. Hmmm…. maybe I should reconsider and at least test an output stage with those darlingtons…
In order to waste less power through the output stage, linear power supplies often use so-called tracking preregulators. The idea is that the voltage supplied to the output stage will track the actual output voltage, typically a few voltages higher. This way the transistor will waste less power.
I have a feeling that a tracking regulator will complicate the design significantly, so I will not prioritise that in the first revision. Expect some huge heatsinks and a fan for now…
Or could I use “multilevel output stage supply” such as in this PSL -3604 project ? “To minimize dissipation power output stage uses 4-level supply. It provides high dynamics, while conventional circuits with transformer tap switching or preregulator need time for bulk capacitors charging“
Heatsinks and fans
Info about possible cooling solutions ? For example such as the ones that seems to use CPU cooler:
I haven’t fully understood all the transistor datasheet info, heat transfer etc. that is involved to make sure transistors operate in a safe area. But I liked the explaination given in Scullcom Hobby Electronics #46 – Electronic DC Load Part 2 where he explains some of it: http://Scullcom Hobby Electronics #46 – Electronic DC Load Part 2
One of the things that has surprised me when looking at the construction of various power supplies and SMUs is the heatsinks and fan placements. Actually even more so on expensive equipment that cheap equipment. Also, it seems that the noise level does not have high priority.
For my project, the noise level has high priority. If I keep the circuit design simple with no preregulator, the output transistors will generate a lot of heat at high currents and low output voltages. I’ll not get away without a fan. I’ll probably go for a heatsink tunnel solution with a quality low noise fan blowing air though it. I’ve tested a 60x60x100 heatsink with a Noctua silent fan. That solution has no problem with 2A (40W) and almost silent. Example of tunnel solutions is the Agilent 663x/66332 series
That is a 100W power supply, but I’m not sure if it uses a preregulator to keep the voltages over the transistors lower at lower voltage outputs. Rumours say the power supply is noisy, but it’s easy to change the fan to one with lower noise.
I’m not sure how hot the output transistors are “allowed” to be. According to Daves teardown of Siglent SPD3303 its output transistor maxes out at a around 100C
Dave’s saying “that’s pushing it”. So I guess I’ll try to get the temperature below 100 degreeC. 80 ?
During my “research” one of the most surprising learning is how important resistors are! One of the problem is that change value over time and temperature. That causes major problems for high precision circuits. Absolute value is not so important because inaccuracies can be calibrated out. The stability, however, is important.
I plan to have a PCB where it’s possible to use different types of resistors. Then I can use expensive parts to test precision. And use cheap parts for throwaway experiments.
I need to learn and experiment more with these things. Good resistors are extremely expensive so you should know where the stability matter where it does not.
Seems that Vishay is “the place to go” for precision resistor. Checkout Design and Selector Guide for High-Precision Resistors Vishay Foil Resistors
Using resistors to set opamp gain typically requires a resistor divider. In that case, the specifications for each resistor is not critical. It’s the tracking specifications that are important. 4 resistor version examples: Vishay AORN 5ppm/C tracking, Vishay MORN: 1ppm/C tracking, Analog LT5400: 0.2ppm/C (note that Analog states that: This parameter is not 100% tested !).
There are also two resistor networks. I noticed that the LTZ1000 daughter board for AD5791 uses a Vishay Y1685V0001TT9 with tracking specified to 0.2ppm ! But it’s expensive. In that application it’s used to create -10V from 10V. Other cheaper individual resistors are used for the same in the LT6655 based daughter board.
Other two resistor networks examples: MAX5490/1/2 and Vishay MPM (2ppm/C tracking). They seem quite similar. The Vishay spec shows better 2000h stability.
Current shunt resistor selection
There are several ways to measure current, but I’ll go for the most basic approach: measure voltage over a shunt resistor.
Using resistors for current sensing: It’s more that just I=V/R : “Ideally, the sensing resistor value should be relatively large so that the resultant voltage drop will also be large, thus minimizing effects of circuit and system noise on the sensed voltage, as well as maximizing its dynamic range. However, a larger value at a given current also means there is less voltage—and thus less available power—for the load due to IR drop, as well as I2R resistor self-heating, wasted power, and added thermal load. It’s clearly a tradeoff and compromise situation…. Note that using a snip of copper wire or PC-board track might seem like a good way to get a milliohm-valued sense resistor at nearly zero cost. However, the TCR of copper is around 4000 ppm/°C (0.4%/°C), which is orders of magnitude inferior to a low-TCR sense resistor…. In some cases, a viable tactic for reducing the temperature rise due to self-heating is to use a larger-wattage, which will be less affected by self-heating. But these, too, have a somewhat higher component cost and larger footprint. The designer must do a careful analysis of the current, the dissipation, the effects of TCR, and any derating needed for long-term reliability and performance “
When using low value current shunt to measure large currents, even the solder join makes a difference. That brings you to kelvin connections. Optimize High-Current Sensing Accuracy by Improving Pad Layout of Low-Value Shunt Resistors : “…note the 22% error associated with the solder resistance without using Kelvin sensing. This is an equivalent solder resistance of about 0.144 mΩ.”
Selecting shunt resistor can be a daunting task. But you don’t have to tests all of them yourself, instead dive into this impressive work: Study of temperature coefficient on 260 precision resistors
Also take a look at Vishays selector guide (Design and Selector Guide for High-Precision Resistors Vishay Foil Resistors). Seems Vishay is the place to go. Have also read nice things about Dale resistors.
Do I have to care about “burden” voltage is is that only relevant for “external” measurement ?
Seems 0.1ohm is can be a suitable value to used for 1-2A range. Large currents will heat up the resistor. I’m not sure if the tempco is more important that high wattage. A high wattage resistor will heat up less that a low wattage I believe. But if the tempco is low enough, that’s not so important. Where is the balance ? Don’t know. Can it be calculated or must it be tested ?
Dave Jones explains current shunt and how to select values for gain etc. for his uSupply (direct link). It uses CSM2512 Bulk Metal® Foil high-precision current sensing resistor with resistance of 0.01 Ω and tolerance of 0.1%
The Keithley DMM6500 also seem to use a 0.1ohm resistor, but from Isabellenhuette according to the teardown at lygte.info
Seems that the same is also used in Wavetek/Datron 4808 multifunction calibrator as shown in one of the picture in xdevs teardown. Prema 6047 uses a precision 100mohm resistor from https://www.isabellenhuette.de/en/precision-and-power-resistors/ as indicated in the teardown @18:35. I had never heard about Isabellenhuette before, but is a german company known for high current shunts with low temperature drift. Prema 6047 also uses several good Vishay resistor (500, 5k, 50k ohm) for other current ranges? as well. Rohde & Schwarz also uses a shunt resistor PHB-R100 from Isabellenhuette in their newer power supplies as indicated in this teardown @6:21. Thats not as good as the one used in Prema. Agilent 66312A (2A) uses a 0.25ohm Dale resistor as indicated in the Kerry Wong teardown @11:58 (text review here). The 66332 (5A) uses as low as 0.05ohm resistor for the 5A range according to the schematics.
Keithley 2400 SMU uses four Dale resistors in a parallel series configuration according to the teardown @18:34. Not sure for what range. If not sure the 2400 should be used as reference for high currents (1A) anyway. It’s current source accuracy in the 1A range is only 0.27% according to the datasheet. In addition, the datasheet footnote states that “Specifications valid for continuous output currents below 105mA. For operation above 105mA continuous for >1 minute, derate accuracy 10%/35mA above 105mA”. It’s also worth noting that the sink capabilities are even worse according to the datasheet.
Keithley 2602 uses VCS301T 0.1ohm according to https://www.eevblog.com/forum/repair/keithley-2602-dual-smu-repair/. According to the datasheet, the 1A source range accuracy is around 0.05%. Keithley 2425 uses VPR221T 1.0ohm (100mA range?) and VCS301? 0.1ohm (1A range?) as shown in a xdev teardown/repair @20:00
This means that in the “older” Keithley 24xx series of SMUs, the (famous) 2400 might not be the right choice if you want high accuracy current measurement above 100mA. Note, however, that higher current versions typically lose the lowest current range found in 2400 and are more expensive.
I’ve used a Vishay VCS1625 in my first prototype. Should be good, but at current >500mA I’m not happy with the drift. Does it drift because of self heating ? Could be other reasons but I want to try a VCS301/2 or VPR221Z (which Dave Jones uses in his 1A current reference). I’ve testet VPR221Z 0.5ohm (lowest value for that type) and it seems very promising (but expensive!). It gets a temperature apprx 50degreeC @1A. But the temperature increases rapidly @1.5-2A. At 2A it reaches >90degreeC and would need heatsink. Ideally I would have a lower value resistor. 0.25ohm might be the sweetspot if I aim for 2A…
Candidates: UPDATE: I need to update this. There are differences between Z versions and non-Z versions ! Be aware of that also when looking at prices !
For 10mA range, I believe there are many options. More research needed. In my prototype I use a relatively cheap SMD resistor with a relatively decent tempco (25ppm). Current source stability in that range has not been a problem so far. But I guess self heating is not a big problem here.
Write about selection of capacitor…
Capacitors making audible noise (beeping, “singing”)
When testing the circuit prototype, I often apply square wave signal to the output stage to see how it performs. I then often hear a beeping sound at certain frequencies. I originally though it was the transistors (not sure why I thought that), but could it instead be the capacitor ? I didn’t know it, but it seems there is a well known fact that when an AC voltage in audible frequency range (20 Hz – 20 kHz) is applied to a multilayer capacitor chip, its can vibration and make sound that can be perceived by the human ear. There are even special capacitors made to reduce this phenonomen, such as Murata (https://www.murata.com/en-sg/products/capacitor/mlcc/solution/naki) “The replacement evaluation of low acoustic noise capacitors is demonstrated. The acoustic noise phenomenon occurs in some cases due to the piezoelectric properties of ceramic. In order to solve this, Murata has developed a low acoustic noise capacitor! Click here to experience the effectiveness of the low acoustic noise capacitors of Murata“.
Singing Capacitors,Piezoelectric Effect (from TDK) ” What is a Singing Capacitor? A1. Singing is one of many ways to describe the piezoelectric effect on the capacitor. This “singing” is actually a vibration of the capacitor on the PCB that many occur under specific conditions. “
In Reduce acoustic noise from capacitors they suggest putting slots in the PCB can reduce the problem. And/or use component with a higher voltage rating. Prices for special “low noise!” components are discussed. In my prototype, I can live with a higher cost for a few capacitors if I need to. They also discuss using two capacitors, one on each side of the PCB !
PCB layout considerations
The PCB is a component of op amp design “Most analog designers are familiar with how to use ICs and passive components to implement a design. There is one additional circuit component, however, that must be considered for the design to be a success—the printed circuit board on which the circuit is to be located…“
The basics: How to layout a PCB for an op amp “Applications engineers tend to overlook printed circuit board (PCB) layout during circuit design. It is often the case that a circuit’s schematic is correct, but does not work, or perhaps works with reduced performance. In this post, I will show you how to properly lay out an operational amplifier (op amp) circuit PCB to ensure functionality, performance, and robustness…“
AN 1258 Op Amp Precision Design: PCB Layout Techniques “…techniques for improving the performance, giving more flexibility in solving a given design problem. It demonstrates one important factor necessary to convert a good schematic into a working precision design…“
How to layout a PCB for an op amp“…That’s when I realized that PCB layout isn’t as intuitive as I thought…“
How to layout a PCB for an instrumentation amplifier “… it is easy to make mistakes in the PCB layout that might degrade circuit performance. …shows a PCB layout with three mistakes we at TI commonly see when reviewing INA layouts…“
Layout for precision opamp “The incredible offset and drift performance of modern precision op amps can easily be degraded by poor PCB layout techniques. By utilizing a few simple layout techniques, the performance of the IC can be maintained…“
Isolate analog and digital circuits /opto isolation
Optocoupler as Optical Isolator for SPI Bus System an application note from Vishay showing how opto couplers can be used for SPI
SPIsolator – We’ve Got Your SPI Bus Covered! Analog Device presents their ADUM series of isolators using magnetic coupling
SI866x “…Silicon Lab’s family of ultra-low-power digital isolators are CMOS devices offering substantial data rate, propagation delay, power, size, reliability, and external BOM advantages over legacy isolation technologies…“
MAX14931 vs ISO7241 for ISP communication. Which to use… https://electronics.stackexchange.com/questions/149980/ucontroller-spi-over-optoisolator
Partitioning and layout of a mixed signal PCB tries to explain why separate ground for analog and digital circuitry might not be the correct solution.
ATmega328PB Isolated Application Board An ATmega328PB breakout board with digitally-isolated SPI, I²C, or generic GPIO. $3 on tindie.
The decoupling capacitor…is it really necessary? “…I collected data for about a week, and none of my results matched expectations. I made numerous changes in an attempt to improve performance, but nothing worked. Finally, I decided to add the decoupling capacitor. As you might expect, this solved the issue…“
http://www.analog.com/media/en/technical-documentation/application-notes/an139f.pdf “PC-board layout determines the success or failure of every power supply project. It sets functional, electromagnetic interference (EMI), and thermal behavior. Switching power supply layout is not black magic, but is often overlooked until it is too late in the design process“
Why use guard ring… more info…
Why use isolation slots. From eevblog forum user Bernie in this thread: “There are tons of reasons why you might want to slot a PCB:
– High voltage isolation: Most common and obvious use of slots by preventing arcs creeping over the surface
– Leakage prevention: Sometimes you don’t want surface contamination leaking current in a critical high impedance node, in some cases this can be used instead of or in combination with a guard ring.
– Mechanical isolation: You might have a precision part that works in ppm precision like a reference of precision amp where mechanical stress on the component can cause deviations. Slots prevent force being put on the sensitive area to bend or stretch it.
– Thermal isolation: Sometimes you have precision circuitry close to a power supply or high power amplifier/driver that gets hot. You might want a slot between the two to keep the heat from getting to the sensitive precision circuitry.Alternatively if you are ovenising your precision circuitry then you might want slots to keep the heat in there.
– Ventilation: Sometimes you have a board mount heatsink or large power resistor that gets hot and nothing under it. In that case it might help to slot the pcb under it to allow more fresh air to get to it
– Vibration: Its possible to isolate vibration by slotting a spring shape in the PCB, however this needs careful design to be effective rather than make it worse and might make the PCB easy to snap.
-RF voodoo: Rarely slots might be used in fancy pants RF circuitry because for whatever reason its better to have air as a dielectric there. Only likely makes sense if you worked extensively with microwave for the last 10 years.
– Creative mechanical solutions: Sometimes you simply want to stick something trough the board somewhere, or perhaps something more creative like clamping a magnetic core around a PCB area to create a inductor/transformer(Seen it done before) or something odd like that
Basically mostly just use common sense with slots and don’t overuse them.”
Not sure how the UX should be made. 5″ capacitive LCD with some extra buttons.
The display part is easy, I’ve done a lot of research on 5-7″ capacitive touch displays before (you can read more about that in the ‘display’ categories at weatherhelge.wordpress.com )
Something like Keithley 2450 sourcemeter ?
Keisight sourcemeter and multimeters:
I feel the Keysight sourcemeter has too small digits.
Keithley 2280s power supply:
Maybe a combination of 2280s and 2450…
Statistics / graphs
Live trend graph with limited samples is easy. Trend graph based on ALL samples over long time is more difficult i.e. due to limited sampling memory. Some resources?:
Is decimate the right thing to use ?… probably not with the data I have that can be noisy and irregular ??? Standard algorithm seem to have to much filtering…