Precision current Sensing? (100s of microamps)

[hex4def6]

Hi all,

I'm trying to integrate some current sensing into a project. I've got about 20 shunt resistors that I want to use to measure current into various portions of my circuitry.

Now, my plan is to create a daughterboard that can be attached to the mainboard to measure these currents. The design is pretty crowded right now, with a lot of digital stuff going on. At most, I can use 2 extra layers for my shunt resistor monitoring.

I want to use a MAX9934 to convert the voltages across the shunt resistors to single ended voltages, and run those to the daughterboard through a connector.

I'm going to use my two layer allowance in the following way: first = power plane + MAX9934 outputs, second = analog ground.

Finally, on the daughterboard, I want a high - resolution, multi channel ADC + uC.

Any pitfalls I should look for? I want to be able to measure both sub-milliamp to 100s of milliamps. This results in looking for 100's of microvolts to ~100 millivolts, which is quite a large range...

Any recommendations on ADC / practicality?

 

[MagicSmoker]

You're not asking for anything too difficult, here. My only suggestion would be to scale the gain of the '9934 so its output at max current matches the reference for the ADC. And as for a multichannel ADC+uC, I highly recommend using an AVR. Most of them are 10 bits which covers your dynamic range requirement. I don't know if any of them have 20+ channels, but it's easy enough to daisy chain them together via an SPI bus. I did just that with two ATMEGA48 on a board recently. Works great (easy enough for me to say - someone else writes the software for this stuff!)

HTH

 

[hex4def6]

Yeah, that's what I figured as well. I think I'm going to get the high gain version of the 9934, and scale the shunts accordingly. I figure I'll get a better SNR that way.

By the way, why doesn't the datasheet speak in terms of gain, rather than saying "25uA/mV" or whatever? Since the input is a voltage sense, not current...

 

[IRstuff]

Huh? 25uA/mW IS a gain, and is referenced to a voltage input. And ignoring the fact that the datasheet DOES described the voltage gain on page 15.

I don't necessarily see that your gain choice will result in the dynamic range you claim as a requirement. Moreover, the high-gain device has a poorer frequency response, and poorer common-mode rejection ratio.

TTFN
faq731-376

 

[itsmoked]

An alternative scheme would have Board to Board pins rising from each shunt up to the daughter board. No more layers would be required and those tiny signals won't be run all over the main board.

Or flip that and have the shunts on the daughter board.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[hex4def6]

I didn't see that table.. thanks.

Can I use that amp as a current output device (rather than sticking the resistor on the output), and connect it to current-input ADC, such as:

http://www.ti.com/lit/ds/symlink/ddc232.pdf

?

What are the pro/cons of doing that versus using it in that manner versus described in the datasheet (their example of using a 10k res, output hooked to SAR ADC)?

 

[hex4def6]

Also:

Ok, so what is the correct way of doing it?

My constraints on Rsense are:
A) Power Consumption
B) Rail tolerance
C) Max / Min Vsense scale -> ADC to maximize dynamic range.

I had assumed it would be better to crank the gain up on the amp, as long as I didn't saturate the ADC.

Are you saying that for (say) a vsense of 100uV, it's better to have a gain of (say) 50 at the amp, rather than 250 in terms of measurement accuracy?

I understand for the case where I have flexibility for Rsense, that I should probably use the largest value I can, and use a smaller gain.

Assuming I use a 16 bit ADC, with Vref = 500mV, I will get 0.0076mV/count. That gives me 327 counts for 0.1mA over a .1ohm res, but only 65 counts at gain = 50.

My amps are going to be right at the sense res, so I figure that there mode be too much induced noise between the amp and resistor. On the other hand, the output of the amp will have to travel across a noisy board to get to the ADC. Doesn't that suggest I should sacrifice CMRR for higher gain?

 

[PowerAmpGuy]

To all,

I think increasing the gain is a good idea. The key is to use a low offset and low offset drift amplifier. Some amplifiers called “zero drift” use an internal calibration circuit to correct for offset and drift. Below are links to a good article series on the topic of current monitoring. It includes info on many of the tradeoffs, error sources, and PCB layout stuff. I hope this helps.

http://www.eetimes.com/design/indus...Current-Sensing-Tutorial-Part-1--Fundamentals

http://www.eetimes.com/design/industrial-control/4236380/A-Current-Sensing-Tutorial-Part-II--Devices

http://www.eetimes.com/design/indus...A-Current-Sensing-Tutorial-Part-III--Accuracy

http://www.eetimes.com/design/indus...art-IV--Layout-and-Troubleshooting-Guidelines