EE help needed with current monitor circuit ...

[wquiles]

Thanks to a sugestions by Newbie, I am using the TI/Burr-Brown INA196 Current Shunt Monitor, along with a highly accurate Ohmite CS3 Series, 4-terminal 0.01 Ohm current sense resistor (1% tolerance). I am using a standard 9V alkaline cell (the small 5mm RED LED draws less than 2mA if I recall) and this drives a standard 7805 to get a nice and stable 5V to the INA196.

The circuit that I build it shown here:

Here are some pictures of the actual pieces and project:

EDIT: Pinout for the INA196 in the diagram is not correct - I am re-using a part in DesignWorks Lite . Correct pinout is:
1 - out
2 - gnd
3 - V+ (supply)
4 - Vin (-)
5 - Vin (+)

The problem is that when I turn the circuit "ON", I get a "nominal" 26.7mV on the output, even when I have no current flowing through the LED. When I set my bench supply to turn on the LED at various currents (100mA, 200mA, etc.), the output voltage does get proportionally higher, but I always have the same "offset", plus the gain is all messed up (I get a factor of 7.1 times the mV reading in the voltmeter to match the accurate output current reading on my bench supply). According to the data sheet for the INA196 there should be no offet - voltage is zero when the input current is zero, and it should have a gain of about 20.

Any ideas that is going on? Suggestions?

Will

 

[evan9162]

put a 100k resistor from the output to ground as a weak pull-down?

 

[CM]

What is the expected maximum current through the sense resistor?

 

[wquiles]

put a 100k resistor from the output to ground as a weak pull-down?

I will try tonight to see if it helps any.

What is the expected maximum current through the sense resistor?

It is a 3Watt 0.01 Ohm, so it can take quite a bit of current :naughty: . I tested up to 1 Amp so far.


I also just added some photos to the main post so that you can "see" what I am trying to do 😉

Will

 

[Handlobraesing]

If you aren't going to push more than an amp or two, 0.1 ohm is small enough to be reasonably transparent to the circuit and gives you ten times better resolution.

Most DMMs can measure to 0.1mV and a 0.1 ohm would give you 100mV/amp, or 0.1mV/1mA.

Get a 100mV/amp shunt on eBay. and you don't have anything to calibrate.

 

[wquiles]

If you aren't going to push more than an amp or two, 0.1 ohm is small enough to be reasonably transparent to the circuit and gives you ten times better resolution.

Most DMMs can measure to 0.1mV and a 0.1 ohm would give you 100mV/amp, or 0.1mV/1mA.

Get a 100mV/amp shunt on eBay. and you don't have anything to calibrate.

This is basically the shunt you mentioned, but to the next level in accuracy and isolation, assuming I can get it to work 😉

Will

 

[vanalog]

This looks like a really fun problem.

Looking at the INA196 spec sheet confused me even more:

The input offset range is from .5 mV (typical) to 3 mV max.

The gain is 20 V/V. In my simple Op amp conception, output offset = input offset X gain, giving a range of 10 to 60 mV. What you are seeing falls in this range.

However, they also spec a total ouput error of 2.2% max, stating: "Total output error includes effects of gain error and VOS". Pretty incredible performance!

Looking at the Output Error vs. Vsense graph, this level of error occurs at a Vsense of about 10 mV.

So I am not sure what spec to believe...

If you were measuring 700 mA, you would see 7 mV across the shunt, and 140 mV out. I agree that using a .1 ohm shunt would get you further out of the noise floor, whatever is causing it.

I do think your circuit needs bypass caps at the sense amp power pin: a 2 uF. tantalum and a .01 ceramic. I would look at the output with a scope to see if it is oscillating.

I have also seen some big offsets from thermoelectric effects: if the input nodes are at different temps, dissimilar alloys, etc. A shot of freeze spray can exaggerate these. It is tough to measure a few mV accurately...

 

[CM]

I will try tonight to see if it helps any.



It is a 3Watt 0.01 Ohm, so it can take quite a bit of current :naughty: . I tested up to 1 Amp so far.


I also just added some photos to the main post so that you can "see" what I am trying to do 😉

Will

You should increase the sense resistor value by an order of magnitude to diminish offset voltage effects. Be liberal with de-coupling. Linear regulators can be noisy.

 

[wquiles]

I did another experiment last night and this time I had the ceiling FL lights off. I did not seem to have the offset that time!. I might have a simple noise problem 😡

The data sheet talks about some possible input filtering on page 9, but I don't want to yet go and desolder what I have connected to the input pins of the very tiny surface mount part. I am going to try some small caps right across the shunt resistor and/or from each leg of the resistor to ground first and see what I find. I will keep you guys posted 😉

Will

 

[evan9162]

I'm using a 3 watt Ohmite 0.01 ohm resistor as a current shunt, but the difference is that I simply have it inline on a banana->alligator clip test lead (no amplifier circuit). I then connected two extra wires to the leads of the shunt (kelvin contacts) which I connect to my DMM. Using the 200mV scale, it's simple to remember to move the decimal place over (10mV = 1A). It measures within 1% of what my 3 other DMMs read, all the way up to 10A.

Might be something to consider if the monitor circuit doesn't work out.

 

[wquiles]

I'm using a 3 watt Ohmite 0.01 ohm resistor as a current shunt, but the difference is that I simply have it inline on a banana->alligator clip test lead (no amplifier circuit). I then connected two extra wires to the leads of the shunt (kelvin contacts) which I connect to my DMM. Using the 200mV scale, it's simple to remember to move the decimal place over (10mV = 1A). It measures within 1% of what my 3 other DMMs read, all the way up to 10A.

Might be something to consider if the monitor circuit doesn't work out.

Absolutely - this is the fallback plan since the Ohmite 3W 0.01 Ohm resistors (same ones you are using!) come in a pack of two, so I have another one I can use that way 😉

Will

 

[Handlobraesing]

I'm using a 3 watt Ohmite 0.01 ohm resistor as a current shunt, but the difference is that I simply have it inline on a banana->alligator clip test lead (no amplifier circuit). I then connected two extra wires to the leads of the shunt (kelvin contacts) which I connect to my DMM. Using the 200mV scale, it's simple to remember to move the decimal place over (10mV = 1A). It measures within 1% of what my 3 other DMMs read, all the way up to 10A.

Might be something to consider if the monitor circuit doesn't work out.

Most factory calibrated <$200 DMMs are limited to maximum resolution of 0.1mV and the error of 0.25% reading + 3-4 LSD. (This means that at exactly 0.1mV input, the input is allowed 0.25% error +/- 3-4 counts of the least significant digit) which leads to real life error of 40-50mA.

As a rule, you should have ten times the resolution than what you're trying to resolve. If you want to measure accurate within 10mA, then the resolution should be 1mA. By using a 0.1 ohm resistor, your resolution would increase ten times and error would be within 0.25%+3-4mA.(+ shunt error)

For a current less than 1A, I don't think 0.01 ohm is warranted unless you're dealing with exceptionally low system voltage where 0.1 ohm would introduce a significant relative change in system resistance.

 

[MrAl]

Hello there,

I built a current monitor a while back but it was for AC current so for a while i used
a home made 0.01 ohm shunt (a brass strip with leads soldered to it) and used an
ac DMM to measure the voltage across it...no amp. Then i decided to use an analog
AC meter so i set up an op amp amplifier with a gain over 100 so i could read on the
3v scale if i rem right. The amp was simple and i used a common op amp, but for
this i was going to measure AC so the dc offset didnt matter...i could even capacitor
couple the signal out to get rid of ANY dc offset, and this would work fine as long
as the dc offset didnt cause the amp to saturate.

In any case, you are working with DC so you need to get rid of your offset.
There are many ways to do this, the simplest is to use a pot on the output and
bias the negative lead of the meter to a plus or minus voltage to null out the meter
(to zero volts). This will get rid of your offset, but you also need to consider the
temperature coefficient of the input offset just in case it changes a lot with
temperature...in which case your offset would change as the temperature changes,
unless of course you can keep the circuit at room temperature or something.
They do make very good op amps though just for this purpose that have extremely
low input offset. They are sometimes referred to as 'chopper stabilized op amps'
and can be found on the National web site. These op amps have internal circuitry
to periodically measure the output offset dc and make a correction if needed.
Cant beat that, huh?

As some others have pointed out, if you increase your sense resistor to 0.100 ohms
you will automatically increase your signal-to-garbage ratio by 10 times! If you
can stand the extra loss, it's wise to do it. I've made several current controllers
for Lux 1 watters that use 0.1 ohms and they work just fine with a common op amp,
even given the sometimes large offset spec.

BTW, with a 1 amp current you'll lose about 2.5 percent efficiency driving
a Lux at about 3.5 volts because of the 0.1 ohm resistor.
Driving a higher voltage Lux you'll loose even less.


Take care,
Al