24bit ADC +/-10v

Hi,

I need to read in a signal to my arduino uno, the signal is a +/- 10v and Ideally I’d like 24bit resolution. Does anyone know of an ADC which can do this? Or how can I get around this problem?

Thanks

Jim

Analog Devices AD7713, AD7732, AD7734
All available at Digikey.

24 bits is asking a lot. That is about 1.2uV per step.

Ever hear CDs sampled at 24-bit, 96 KHz? Sounds super clean. I have a couple from Rhino records that I got for little money thru Audio magazine sibscription, part of the Up Close series from several years ago. Even ripped to MP3 at 256K sampling they sound great.

Are you responding to me? What is your point?

Yes. Just saying.

Oh. They do that with big, fancy, well-engineered equipment, of course. I think for the average goomba, 16 bits is not an easy target to hit and get meaningful results. So we have to strive to be above-average goombas.

Well it's true that to take advantage of all 24 bits of such a ADC the board layout is critical with lots of attention to physically separating (not electrically of course) analog ground from digital ground for the lower 4 or 5 bits to actually represent anything meaningful.

Don't expect 24 bit measurements from an arduino.
The analog inputs are 10 bits resolution, so how many years would you have to be averaging for before you get the 2^14 records?
Summing four analog reads to get 12-bit resolution is very reasonable.

what do you need such fine measurement for?

That's why the external 24-bit ADC.

Thanks for the info, that ADC looks perfect

You never said how fast you need it to be.

ideally, id like to read each channel (I fancy the 4 channel version) every 1/10th of a second, but 1/100th would be even better. Will the AD7734 do this?

retrolefty:
Well it's true that to take advantage of all 24 bits of such a ADC the board layout is critical with lots of attention to physically separating (not electrically of course) analog ground from digital ground for the lower 4 or 5 bits to actually represent anything meaningful.

Don't forget guard rings and the shape of the edges of the traces.

Typically you can do a lot of smoothing in software, but for 24bit, you would want to address every possible source of noise on the input.

one question is the actual bit count after all the overhead in the chip. the Teensy has a 16bit ADC but has 13bit output after the overhead.

ad2049q:
Summing four analog reads to get 12-bit resolution is very reasonable.

No, really it isn’t.

To get ‘n’ bits of additional precision, you need 22n samples.
With a 10 bit converter, n = 2, so you’d need 22x2 = 24 = 16 samples.

Effective Resolution, aka Effective Number of Bits or ENOB, is always less than the stated resolution.

Page 10 seems to indicate 21 bits of effective resolution at best.

AVR121: Enhancing ADC resolution by oversampling
http://www.atmel.com/images/doc8003.pdf

Note that this relies on noise equal to about 1 LSB to be present on the signal. I have also seen this done as a linear triangle wave sync'd to the sample, where the time for one ramp up is equal to the number of extra times you'll be sampling. So oversampling x4, the ramp goes up during 4 samples, then back down during 4 samples.

Effective Number of Bits:

Noise, versus Effective Resolution, versus Effective Number of Bits:

I apologize for earlier stating that Effective Resolution and Effective Number of Bits are the same. They are not.

ideally, id like to read each channel (I fancy the 4 channel version) every 1/10th of a second, but 1/100th would be even better. Will the AD7734 do this?

1/100 = 100 Hz.
Data sheet says:
Optimized for fast channel switching
18-bit p-p resolution (21 bits effective) at 500 Hz
16-bit p-p resolution (19 bits effective) at 2 kHz
14-bit p-p resolution (18 bits effective) at 15 kHz

500 >100, so I'd say yes.

Keeping in mind that you’ll never get to 24 bits of effective number of bits without some chicanery.