1A current source but still not stable

[dmm2018]

Hello...

After watching Dave's videos: Precision 1A Current Source (Part 1 & Part 2), I've been wishing to build one. I have just completed one few days ago and wish to share the results.

R1, R2, R3, R4 and R5 are all 10 kohm (Vishay S102C, 5 ppm/degC, sourced from hifi-szjxic)

RM is VPR221 (0.5 ohm, 2 ppm/degC, sourced from hifi-szjxic)

RP consists of two resistors connected in series
-   RP1 = 200 ohm (TE, 15 ppm/degC, sourced from element14)
-   RPT = 50 ohm trimmer (Bourns 3296, 100 ppm/degC, sourced from element14)

Q1 is RFP12N10L (Fairchild Semiconductor, sourced from RS)

AD780AN is a voltage reference (configured to give 2.5 V, salvaged from a scrapped 15-year old DAQ board). The AD780 is powered by a +5 V, generated by LM7805, supplied by a battery.

AD708JN is an op-amp (salvaged from a scrapped 15-year old DAQ board). The AD708 is powered by dual supplies (+/- 12 V, batteries).

Two DMMs (connected in loop) were used to measure the current:
- Keysight 34470A (N) - calibrated 5 months ago;
- Keysight 34470A (L) - calibrated 54 months ago;

The DMMs were warmed up more than 1 hour.

The current source with 2 different VS, i.e. 6V and 12V, were tested. The recording started after the current source was powered up for 1 min. The current source with VS = 12V was tested first, then stopped 10 mins before continuing with the test with VS = 6V. Each record took 30 mins.

The tests were conducted in an air-conditioned room (the aircond temperature setting was 18 degC), felt a little air-flow.

With VS=12V, the current kept drifting (descreasing) in the 30-min duration. With VS = 6V, the current seems to settle faster, but still have fluctuation, probably due to the not so stable room temperature or airflow. Still thinking why with VS = 12V the current kept decreasing and not stable...

Please comment and suggest how could further improve the stability of the current source.

Thanks.

[Zero999]

It's oscillating. Try looking at the output with an oscilloscope.

An additional RC circuit is required to stabilise the circuit.

https://www.electronicsweekly.com/blogs/engineer-in-wonderland/current-sink-stability-2015-10/
https://www.eevblog.com/forum/projects/compensation-methods-in-current-sink-circuits/msg735672/#msg735672

[Rerouter]

Easiest way to get it more stable, Add a generic 10K resistor to the inverting leg (this one can be 5%, its just to give the feedback capacitance some room to work), and from the inverting pin to the output pin of the op amp, add something in the order of 1nF, this will decrease the response speed of the loop, but make it more stable (when its oscillating, its correcting faster than the mosfet can react)

Next up is add a generic 10K resistor from the op amp output to ground (again can be 5% cheap resistor), this keeps the output of the op amp biased on, and helps it deal with the capacitive mosfet a bit nicer.

[strawberry]

https://www.analog.com/media/en/technical-documentation/application-notes/an133f.pdf
for generic OP is hard to drive capacitative(MOSFET gate) loads

[dmm2018]

Thanks Zero999, Rerouter and strawberry.

When the circuit was on breadboard, we did measure the voltage across the RM (0.5 ohm) using a Tek scope and the voltage waveform looked stable. However, after the circuit was on a PCB, we have not measured the voltage waveform across the RM yet. We will check the waveform again to confirm whether there is oscillation. We will also add resistors/capacitors to further enhance the performance of the circuit.

[dmm2018]

The current source were tested using two different VS, i.e. 6V and 12V.

With VS = 6V, the current (as can be seen in the figure below) seems to drift less (compared to that when VS = 12V). The current fluctuated within a span of approximately 10uA~11.7uA.

With VS = 12V, during the 30-min interval, the current drifted (decreased) approximately 93.9uA~95.1uA (nearly 100ppm), and we believe, if we kept logging the data, the current is expected to drift/drop more.

With VS = 12V, the heat dissipated by the power MOSFET is theoretically double than that with VS = 6V. Did this extra heat dissipation cause the VREF (7 ppm/degC max), R1~R5 (5ppm/degC) and RM (2ppm/degC) to drift and hence the huge 100ppm drift? Did thermal EMF play any role in this 100ppm drift?

The RP was used to trim the current so it's near to 1A.

Kindly advise.

[Rerouter]

R1-R5 all have the same temperature coefficient, so as long as they are kept at a similar temperature, the relative change in ratio should be almost 0ppm. (this is why low power dividers don't care about absolute accuracy of the resistances, only that the ratio is accurate)

So this leaves the Op amp's input offset voltage, that likely will change with temperature, some possible leakage depending on the type of feedback capacitor you chose, and ground loop resistance, (copper's PPM is not so nice)

If you think about it from a control loop perspective, R5 and the voltage references ground should be tied as close as possible to ground of the sense resistor as possible, Ideally with separate traces for both to prevent one from influencing the other. This way the control loop is referenced to only the resistor, with almost no copper resistance contributing. Its due to coppers influence that you find things like 4 leaded sense resistors, with dedicated voltage sense points, for your breadboard design, just keep them close.

[Rerouter]

Something like this, There are times where star grounding doesnt matter, and times where it hurts more than it helps, but in this case its about keeping the same reference point.

[spec]

Hi dmm2018

In addition to the other member's points, here are a few more:

The AD708 is not too happy with a single supply rail of just 5V. +-5V is an OK working minimum and +-15V is optimum. +-12V would be fine too.

With your present arrangement you are going below the common mode input voltage range for an AD708 running on a 5V or 6V or 12V single supply rail.

I would recommend that you use another battery and 7905 regulator to give a -5V supply line. This will make the AD708 very happy.

But, I dont think that a 7805 will be able to generate a 5v supply line from a 6V battery: the 7805 dropout voltage is 2V average and the spec sheet gives no maximum. Neither would the 7805 be capable of supplying 1A (drop out and thermal limitations).

Hate to say this, but the power supplies for this circuit would appear to need a bit of a make-over. Don't forget that with a circuit like this, which has high current and prescission electronics, good solid power lines and 0V line are absolutely essential to get good performance, and the AD709, with its low input offset voltage of 30uV max and reasonable input bias current of 1nA max, has the potential for excellent performance. But you are throwing away at least 10uV of precision with such a high value for R5 (10k) 1k, would be a much better choice.

A simple way to solve many of the above problems would be to get another opamp that would work with a single 5V supply line and has an input voltage range that includes the 0V line. It would be better to get an opamp with a much lower input bias current too. The OPA191 would be a good choice. As this opaamp has a rail to rail output capability it would also resolve the issue between the AD708 minimum output voltage which may not go low enough to turn the NMOSFET off, so there will be an ambiguity when generating low value constant currents.

The circuit has no decoupling capacitors, so I would recommend that you connect 100nF X7R ceramic through-hole capacitors as follows:
from the +Vs pin(5V) of the AD708 to the 0V reference of the AD708
from the -Vs pin(-5V) of the AD708 to the 0V reference of the AD708
from the 0V reference of the NMOSFET to the 5V supply close to the NMOSFET.

To inhibit possible parasitic oscillations of the NMOSFET (typically 500KHz to 20MHz), it is best to connect a 47R resistor directly to the gate pin of the NMOSFET, keeping the resistor lead to the NMOSFET gate as short as possible. There should be no other connections to the NMOSFET gate pin. The 47R resistor is a gate stopper and  has nothing to do with the vital loop stability issue that Zero999 mentioned in Reply #1.

It may also be a good idea to connect 47uF, or larger, electrolytic capacitors across the +-  supply lines as close to the NMOSFET as possible.

Also, keep your circuit out of draughts and away from any sources of heat, like the sun's rays for example. In fact, if you put the circuit in a box it would be the best approach (but don't forget cooling for the NMOSFET).

Finally (at last  ) I would like to emphasis how important Rerouter's comments are, in Reply #7, about star points for the wiring.

http://www.ti.com/lit/ds/symlink/lm340.pdf

http://www.ti.com/lit/ds/symlink/lm79.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/ad708.pdf

https://www.analog.com/media/en/technical-documentation/data-sheets/ad708.pdf

https://www.onsemi.com/pub/Collateral/RFP12N10L-D.pdf

http://www.ti.com/lit/ds/symlink/opa191.pdf

[SiliconWizard]

Wondering whether you have let it run for much longer than 30min and see what you get.

The worst I can see from your captures is something around 94ppm for 12V. That doesn't seem too bad for such a simple circuit. Of course if it keeps drifting in the same direction forever it will be a problem, but does it? Would be interesting to have a measurement of the local temperature (very close to the circuit) at the same time as well. The transistor will heat up quite a bit @1A.

Heck, for 6V, you get a ~12ppm variation (min to max). That's very low. What kind of stability exactly are you after? 

There are also a bit too many resistors for my taste (even if they all have a low TC, that adds up...)

[Dr. Frank]

This circuit usually is extremely stable, I have designed exactly that topology, up to 100mA.
See here: https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

The stability @ 6V supply seems to be reasonable, whereas at 12V, it should be the same stability, as this circuit does not depend on the supply voltage of the current path.

This excessive drift does not fit to the parameters (T.C.) of the high quality components.
Make sure, that the FET does not heat up the VPR221. Therefore separate both heatsinks.

The different batteries used, and their inter connection is also not precisely drawn in your schematic. Please provide an excact schematic, how these 3 / 4 batteries are connected to the circuit.
The reference and the OpAmp and the GND sense side should best be supplied by the same battery, to avoid any ground loops.

Please check, if you have used a correct Kelvin / 4 wire connection for the current shunt VPR 221. See datasheet of the shunt.
The current connections I+ I- should be connected the FET, and to - of the Battery VS / 6V, whereas the sense lines S+, S-should be connected to the + of the OpAmp and to GND of the Opamp supply.
There must be no other connection between GND and the - of the VS battery!

Then the OpAmp should be checked, if it's still ok.
If I remember right, a similar OpAmp, AD706 is used also in hp 34401 DMMs, and that caused a lot of failures after years, in the Ohm range, due to breakdown of the input transistors.
That was a systematic error, happening on many instruments.

So I wouldn't trust this salvaged part at all .. check if the reference voltage @ + and feedback voltage @  - are stable and always identical to a few µV, or if the difference voltage drifts.. that would imply an excessive bias current.

Maybe you choose another Opamp which has rail-to-rail voltage characteristics, or use a chopper OpAmp.


Concerning both 34470A, before you start measurements, without current, let them warm up 1h at least, and do an ACAL on both, then NULL both instruments in the 1A range.
That might improve the accuracy.

Current measurement is not so precise by specification, anyhow, so the absolute difference in reading seems to be all right.

If one of both instruments already has FW 3.0, you can use the DIG feature for DCI @ 20µs to check if the current output is oscillating.
I've also implemented a stabilizing capacitor C6 of 10nF between - and the direct output of the OpAmp, and this is very important for stability over different loads, and compliance voltages by the FET. 

Frank   

[spec]

This circuit usually is extremely stable, I have designed exactly that topology, up to 100mA.
See here: https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

The OPs circuit is almost guaranteed to oscillate because it has two unit slopes active at 0dB. Your circuit has compensation to sort out this problem.

[dmm2018]

The different batteries used, and their inter connection is also not precisely drawn in your schematic. Please provide an excact schematic, how these 3 / 4 batteries are connected to the circuit.

The reference and the OpAmp and the GND sense side should best be supplied by the same battery, to avoid any ground loops.

Please check, if you have used a correct Kelvin / 4 wire connection for the current shunt VPR 221. See datasheet of the shunt.

The current connections I+ I- should be connected the FET, and to - of the Battery VS / 6V, whereas the sense lines S+, S-should be connected to the + of the OpAmp and to GND of the Opamp supply.

There must be no other connection between GND and the - of the VS battery!

Thank you Dr. Frank. The schematic and the PCB layout are attached. Three batteries are used, i.e. Battery A (12V), Battery B (12V) and Battery C (6V or 12V).

For a VPR221 with Kelvin / 4-wire connection, I'm not sure whether it's correct to connect the trimming circuit to Pin-2 and Pin-3 of the VPR221 as shown in the schematic.

Kindly advise.

[dmm2018]

Hello Zero999, Rerouter, strawberry and spec,

We have checked the voltage across the VPR221 (between Pin-2 and Pin-3). Several waveforms with different timebases (10ms/div, 1ms/div, 100us/div, 10us/div, 1us/div and 100ns/div) were recorded (using Tekronix TPS2014, AC coupling). The measured waveforms are quite stable (except the one with 10us/div).

Kindly advise.

[Doctorandus_P]

I find it a bit strange that you're "complaining" about what looks like a bit of drift in the last 3 digits of a 7 1/2 digit EUR 1500 meter with such a simple circuit.

Is your MOSfet designed for lineair applications, or is it optimised for fast switching, which most are?
This can be a source of instability (local heating in the FET) and even destruction.

I think power transistors are easier to tame for a current sink.

For better stability the circuit mentioned by Zero999 looks good.

From memory:
I think Davey Jones put some beefy capacitors over the shunt resistors to improve stability. But because the shunt resistors are so small you would need some relatively big caps (Few 100 uF) to get a usable time constant.

Also:
It is not possible to check stability with a DMM. You need an oscilloscope for that, and preferably some rapidly changing setpoints to measure step response. Opamp's do not like the capacitive loading of power MOSfet's and step responce is probably lousy without further compensation.
For these measurements it also does not matter if you just turned your fancy DMM on, or if it has been accumulating dust or warming up for months.

Opamps are not ideal gain blocks you can use randomly.
There are thousands of different opamps, and they all have different limitations and flaws, and it is the job of a decent engineer to select the right opamp and work around it's limitations to make a circuit with it that works reliably and to specifications.

Talking about specifications:
What kind of figures are you expecting for?
Such a wide open circuit will probably drift a few miles if you touch it or blow in its direction if you measure it with a 7  1/2 digit DMMM.
(P.S. Don't take it badly, i'm jealous, but please learn to interpret what you see on such a meter more accurately).

[dmm2018]

I find it a bit strange that you're "complaining" about what looks like a bit of drift in the last 3 digits of a 7 1/2 digit EUR 1500 meter with such a simple circuit.

From memory:
I think Davey Jones put some beefy capacitors over the shunt resistors to improve stability. But because the shunt resistors are so small you would need some relatively big caps (Few 100 uF) to get a usable time constant.

Talking about specifications:
What kind of figures are you expecting for?
Such a wide open circuit will probably drift a few miles if you touch it or blow in its direction if you measure it with a 7  1/2 digit DMMM.
(P.S. Don't take it badly, i'm jealous, but please learn to interpret what you see on such a meter more accurately).

Thank you very much and we really appreciate for the comments/suggestions. We're impressed with the stability of Dave Jones Precision 1A Current Source (shown in Part 2) which was constructed on a breadboard and without heatsinks. We hope to achieve similar stability. 

Our aim is to have a stable current source that will complement the 34470A in low ohmic resistance (~1mΩ) measurement. We note that the Keysight 34470A does not do so well in low ohmic resistance measurement, e.g. for 1mΩ, the uncertainty is 400%. If not mistaken, DMM6500 can do better than 34470A in low ohmic resistance measurement (e.g. for 1mΩ, the uncertainty is 20%).

Is your MOSfet designed for lineair applications, or is it optimised for fast switching, which most are?
This can be a source of instability (local heating in the FET) and even destruction.

The power MOSFET we used is RFP12N10L, which is not designed for linear applications. We used the same power MOSFET in a DC electronic load project (5A max) and several were damaged so far... 

[dmm2018]

We have done several tests to evaluate the thermal distribution of the current source prototype. PROSKIT MT-4612 infrared thermometer was used.

A small heatsink was placed approximately 1m away from the current source prototype. The temperature of the small heatsink serves as the ambient temperature (Tamb).

The temperatures of seven (7) points on the current source prototype were measured (see figure below). The temperatures were recorded every 5 minutes, and it took approximately 1 min to record all the Tamb and T1~T7.

Several observations:

(a) The ambient temperature Tamb is quite consistent over the 60-min duration.

(b) The temperatures of AD780 and voltage divider resistive network (R1~R5) are very close to each other.

(c) The temperature of AD708 is consistently approximately 2°C higher than the temperature of AD780 and R1~R5.

(d) The temperature difference between the VPR221 (0.5Ω) and the voltage reference AD780 is approximately 3°C to 4°C.

(e) With VS = 12V, the temperature near the power MOSFET (T1) rose to 40.7°C after 60 mins, almost 10°C higher than that when VS = 6V.

The results suggest that the heat generated by the power MOSFET @ VS = 12V has indirectly caused the temperature of other components to raise about 2°C to 3°C. This probably explains why the drift when VS = 12V is more significant than that when VS = 6V.

Still figuring out how these temperature variation "contribute" to such a huge 100ppm drift when VS = 12V... 

Kindly advise.

[dmm2018]

Just took some thermal images of the current source prototype using Testo Thermal Imager. The current source prototype was tested under three different VS (3V, 6V, 12V, VS is the main voltage source that supplying the ~1A current, see first thread). Thermal image was taken every 2 mins for a duration of 1 hour.

[Kleinstein]

When just using it to test meters there is no need to have so much voltage turned to heat at the MOSFETs. This also makes the selection of the MOSFET easier. At low voltage it is less important to have a MOSFET that is specified for linear operation.

If the FET gets really hot, the gate current might get significant, and this could add to the current.  To avoid this one could use a small FET and a BJT combined as a kind of IGBT replacement circuit. So the FET driving the BJT base current.

The adjustment of the 1 A value is probably better done at the divider, not at the shunt.

For less higher frequency noise if could help to add some filtering cap to the voltage divider.

The AD708 may not be the best choice: it's high precision for relatively high impedance source - the shunt is more like low impedance. So other OPs may perform better.  If only DC is important one could consider an ADA4828 or similar.

For the stability one must also take into account that the DMM is not that good at 1 A - the DMM internal shunt and amplifier may be lower quality.

[dmm2018]

Plan to replace the AD780AN with LTC6655BHMS8-1.25. However, found out that the short term drift of the LTC6655BHMS8 is quite bad and is worse than the AD780AN.

A basic circuit of LTC6655BHMS8 was built (BasicConnectionDonutBoard.png). The output voltage of the LTC6655BHMS8 was logged (15 sec after the circuit was powered up) for 1 hour (PLC=10, sampling interval 500ms). The test was repeated several times and the short-term drift was quite consistent and was worse than the AD780AN. For example, as shown in 6655DB1.png, the output drifted 46uV. The trend shows that the output would have drifted further after 1 hour.