ADC and/or Programmable Gain Amplifier Help

[rs20]

SPI is straightforward, you have the data lines to and from the chip, the clock, and the chip select pin. It should be self-evident how these line up between the ADC and the MCU.

Your circuit looks good, except you must add a resistor between the op-amp output and the ADC input. A C by itself will do nothing but stress out the op-amp (and possibly make it go unstable). You want an RC filter.

If you had differential inputs, you'd use a resistor on each line, and a capacitor bridging the two ADC pins (like the 1uF and 500 ohm resistors shown in the diagram you sent). Extra caps (like the 0.1uF caps shown in the diagram) may be useful to, but less so that the differential capacitor.

[tyguy2]

So this is what a differential input would look like? (I'm using the 1uf cap (c7) for decoupling)

[rs20]

You've got the 1uF cap on the wrong side of the 500 ohm resistors. It goes input, resistor, capacitor || output.

[nctnico]

I'd look for an ADC with an internal PGA. TI has several of these (but not cheap). Using external opamps will most likely cause all kinds of problems with offsets and noise.

[tyguy2]

I'd look for an ADC with an internal PGA. TI has several of these (but not cheap). Using external opamps will most likely cause all kinds of problems with offsets and noise.

The amplitude of the input signal will range between 0v and 1mv
The device needs to be as accurate as possible (little noise)
The resolution can be anything above 10 bits
The sample rate is trivial, my device will only sample at 500hz

The MCP3421 is a wonderful part, I've used it with great success in the past, but the bit depth/sample rate tradeoff doesn't work for your specs. 12-bit mode only samples as 240 SPS, and the LSB at 8x gain and 12-bit mode is 125uV -- which means that with a maximum signal swing of 1mV, you'll only have three "actual" bits. The MCP3421 can't operated at 500 Hz under any circumstances.

18-bit mode could give you ~2uV LSB, which corresponds to 9 actual bits in your usecase, but only gives 3.75 SPS (yes, less than four readings per second).

To clarify, when you say the resolution must be about 10 bits, do you mean you want the LSB to be 1uV or finer? It's a little clearer if you state resolution as a minimum step size.

For the record, I'd endorse an analog stage with a bandwidth of 1 kHz and a gain of 1000 or so (using some sort of chopper amp or similar). Fortunately it's not at all difficult to get a gain of 1000 at such a low bandwidth. At that stage you have a 1V signal that you want to sample with 1mV resolution, which is totally easy with all manner of different ADC technologies.

We already went over this. The analog solution can give me a higher resolution for cheaper. If you can find something that has a high resolution at 1mv, please tell me, because it would be way easier.

[tyguy2]

Perhaps something like this, but smaller? http://www.ti.com/lit/ds/sbas661b/sbas661b.pdf

[tyguy2]

Also, if I'm measuring negative voltages, would a simple DC offset work (ADC doesn't do negative voltage, and I don't want to create a negative voltage ref), or would that introduce too much noise?

[danadak]

You are basically asking for 1 uV resolution or better, and I assume you have a
common mode environment to deal with.

Here are some ap notes dealing with noise and basics of PCB issues -

https://www.dropbox.com/s/2h96beh1fbvz4e2/noise_notes.zip?dl=0


https://www.dropbox.com/s/ruaf9booe17jk8n/PCB%20Layout.zip?dl=0

My first instincts to do this would be to examine low noise instrumentation amps.

Noise tools -

http://www.analog.com/en/design-center/interactive-design-tools/adi-diffampcalc.html



Regards, Dana.

[tyguy2]

So what about the DC offset? How would I measure these negative voltages.
Edit:
Nevermind, it looks like the chip has the integrated in it.

[danadak]

What are you measuring ? A bridge type sensor, a ground referenced sensor ?


Regards, Dana.

[tyguy2]

I'm measuring the human heat beat/ muscle signals.

[danadak]

Then for sure this is a differential measurement because of the high CM environment. This tool
also helps to pick an IA for the application in terms of input CM range and output saturation.

http://www.analog.com/en/products/amplifiers/instrumentation-amplifiers.html

Use the analog tool to evaluate the noise previously.


Regards, Dana.

[danadak]

Also at analog.com look at the 24 bit DelSig A/Ds. Some of them have a cm range that includes
50 mV below ground.


Regards, Dana.

[tyguy2]

So what would you recommend? Should I use the Instrumentation Amplifiers to amplify the signal going into the ADC? I don't see any ADC's on Analog.com that can detect 50 mV below ground. All of the products they offer are just modulators. Also, all of these chips use CMOS to communicate, not SPI or I2C. Maybe I'm missing something...

[andy1]

You mentioned that the application is ECG/EMG  so have you looked at TI ADS1298 (or even ADS1299 for lower noise) they are designed just for this application and it has 16/24b adcs , frontend PGA:s etc integrated. Samplerates go up to kiloherz range and default is actually 500. I am using the 8 channel ADS1298 for ECG capturing and it works really well, there are smaller models with less channels also but similar performance etc.

[danadak]

Delta Sigma selection guide -


http://www.analog.com/en/parametricsearch/10825?mtuid=bd62016050314188bdabc76d71314342#/p1746=1.25|10000000&p4364=Sigma-Delta


Regards, Dana.