Help with current sense for electronic load

[lukaq]

Hello,

I have no idea where to go next with my project, so I seek help here.

What I have: DAC that sets output current, OP amp which looks at current sense voltage and DAC voltage and drives FET. Current sense is 10mR and voltage across it is read with INA199A2 (for OP amp too read)

I have problem with current sense for electronic load, where I set current, but output current drifts from set current because as soon as sense resistor, heats up just a little bit and i'm talking <1°, which has 20ppm/°C and base resistance of 10mR.
From picture you can see what I mean..

How do you prevent this? I don't see any way to compensate for that

Is it possible that when FET heats up, that somehow affects current reading?

[exoticelectron]

How tightly are you trying to control the current?

You could use a current measurement method that doesnt use a series resistor, like one of those flux gap devices.

You could measure the temperature of the resistor and compensate (microcontroller).

[exoticelectron]

Uhhh..are you sure you have those calculations right?

For instance you show at a deltaC of the resistor of 10 degrees, the resistor changes by 2%?

Shouldnt that be 0.00002(20ppm/C) * 10deg * 0.01ohm  =  a change of 0.000002ohms or 0.02%?

[lukaq]

How tightly are you trying to control the current?

That is a good question. I don't know, I have no data and have not seen for any diy or commercial unit like BK or HP/Agilent/... and similar how good of a regulation and setting capability they have.

If I set 10mA, I do want to have 10mA, that would be my goal.

So until my expectations get crushed, I aim to stay in 1mA range. I kinda want high voltage and high current unit (but not high power). With this being said, that would be great for low current application like testing battery, yet unit as whole, I would like to aim at 20-30A full scale range. I can have up to 7 FETS so no more than 4.5A per FET. Than again if I have problem with 1 FET changing current so much, having all 7 could be 7x worse, right?

And BK has 0.1R per fet AND input current shunts, I would assume they do high side reading, on that resistors. 0.1R are per FET and for regulating each FET + current sharing, is that so?

What's the difference between high side sensing and low side sensing, there should be one right? I have low side, HP used it also.

You could use a current measurement method that doesn't use a series resistor, like one of those flux gap devices.

Something like acs712? But as I remember, they are noisy and how would I use them? Read total current or per FET and have it in (FET, OP, current sense) loop

You could measure the temperature of the resistor and compensate (microcontroller).

Uhh, that would be very complicated?
This is the resistor
https://docs.google.com/viewer?url=http%3A%2F%2Fwww.farnell.com%2Fdatasheets%2F2034700.pdf


I do ask myself if it would be better to go with local fet regulation and use one current sense for reading current. But I have no idea again, where to get that kinda shunt (in size and performance like BK has)

[exoticelectron]

Uhhh..are you sure you have those calculations right?

For instance you show at a deltaC of the resistor of 10 degrees, the resistor changes by 2%?

Shouldnt that be 0.00002(20ppm/C) * 10deg * 0.01ohm  =  a change of 0.000002ohms or 0.02%?

Did you see this?

Also, "If I set 10ma I want 10ma" is meaningless, literally. You have to engineer to a tolerance even if that tolerance is "I dont care".  If you are trying to duplicate the precision of a current source made by Agilent, using a MCU to compensate for a resistor tempco to achieve better than 20ppm is not at all out of the question.

[lukaq]

Uhhh..are you sure you have those calculations right?

For instance you show at a deltaC of the resistor of 10 degrees, the resistor changes by 2%?

Shouldnt that be 0.00002(20ppm/C) * 10deg * 0.01ohm  =  a change of 0.000002ohms or 0.02%?

Let me check that again

0.00002 * 10 = 0.0002
0.0002 *0.01 = 0.000002
0.000002 + 0.01 = 0.010002
0.010002 / 0.01 = 1.0002
1.0002 - 1 = 0.0002
0.0002 *100 = 0.02%

So you are right, something went wrong in excel

now this leaves me puzzled, If resistance change is so small, why does my current drift >1ma for some 40° FET temperature increase (not the shunt, it "far" away)

[lukaq]

Uhhh..are you sure you have those calculations right?

For instance you show at a deltaC of the resistor of 10 degrees, the resistor changes by 2%?

Shouldnt that be 0.00002(20ppm/C) * 10deg * 0.01ohm  =  a change of 0.000002ohms or 0.02%?

Did you see this?

Also, "If I set 10ma I want 10ma" is meaningless, literally. You have to engineer to a tolerance even if that tolerance is "I dont care".  If you are trying to duplicate the precision of a current source made by Agilent, using a MCU to compensate for a resistor tempco to achieve better than 20ppm is not at all out of the question.

Tolerance of INA199 and with gain of 100 adds far more than shunt with it's 20ppm it would seem
And i'm sure it's temperature related drift

[macboy]

How do you know that the output current is changing?
If you are monitoring the current with a multimeter, have you considered that it also has a temperature coefficient, and that it might be even worse than that of your circuit?

[lukaq]

I did measure output, but most of the time, I look at voltage on output of INA or on shunt itself, they do agree with each other
For multimeter I use 34401a. And measuring voltage is temperature independent from circuit under test.

If you have a specific way to measure, do tell, I'll try that

[exoticelectron]

I did measure output, but most of the time, I look at voltage on output of INA or on shunt itself, they do agree with each other
For multimeter I use 34401a. And measuring voltage is temperature independent from circuit under test.

If you have a specific way to measure, do tell, I'll try that

Okay lets start over:

-what are your measurements
-can we see schematic of circuit
-please annotate schematic with nodes that correspond to your measurements so we can see what you've measured where

If we eliminate tempco of current sense resistor outright, because it must be very small, then we are left with other possibilities. What is offset voltage of INA? How would that affect the output current at a gain of 100?

[lukaq]

I don't have exact schematic, I went with one of them and I experiment with different components

But in a nutshell, it's like this:


There is only one stage, and one fet per stage.
Current sense R1 is 10mR with INA199A2 as in datasheet. It's output then goes to 1k resistor to - pin of IC1B
What is different is placement of that capacitor, I have 110nF going from output to - pin of IC1B
LM324 was replaced with OPA4188 (FET driver = IC1B, will be replaced with something that can have higher supply later), DAC buffer is not OPA4188, but it is OPA340 (IC1A)

INA temp offset: 0.1uV/°C typ. up to 0.5uV/°C Offset Drift (Maximum)
INA voltage offset: ±5uV typ. up to ±150 uV max

I have no idea what happened the other day, but now current is in +/-1mA. Shunt itself seems to live up to the specs, as does INA.
But I found that I have some problem with OPA4188, when I change voltage across FET from <10v to 30v, and I want current of 1A, I get ~975mA which then goes up to 1A as voltage increases past let's say 12v I believe it was. Then the current is constant.

I've checked voltages on 4188, + is constant ~1.035v, - changes as described, but does match at higher voltages, and output is around 4V, far from 4.9v it can output, so fet is driver right

[Damianos]

From the description and the schematic that is presented here, I conclude that someone has thrown some components on a breadboard and he is measuring something, with unknown meaning, in an attempt to achieve a "good" result! 

Perhaps I misunderstood the concept due to the language or missing imagination or knowledge!

[lukaq]

I did a bit of video watching on this topic, I have seen same problems as I have and now I know where to probe and check.

This is what was new, OPA4340 was replaced with OPA4188, far better op amp for this application. Dac has big offset, nothing I can do with this specific one for now, so I have turned to current sense op amp.
It seems that it's not as good as if I would use 4188. I will have to test this. Here is simulation:

Test current is just under 1A, so looking at waveforms, it's clear that 4188 has far better offset than INA