sigma delta oversampling rate.

[jjm2]

I have a couple of questions on sigma delta converters. I hope someone can help.

If I need 10 bits of resolution, what oversampling rate do I need? I've seen in the literature that 64x should give 10 bits. Does this mean a total of 640 bits from the analog portion of the converter? I've seen it implied that 64 bits from the analog portion should give 10 bits. This doesn't make any sense to me, as if I put in the voltage equivalent to a digital 1, it takes 512 samples for the integrator to even detect the signal.
Any help or insight would be appreciated.

JJM2

 

[IRstuff]

see:

http://www.analog.com/Analog_Root/s...s/interactiveTools/sdtutorial/sdtutorial.html

http://www.analog.com/en/prodRes/0,2889,AD7724_43,00.html

TTFN

 

[logbook]

Your question is a bit vague. Are you taking an 8 bit converter and oversampling to give higher resolution or are you just taking a "1-bit" converter and seeing what is possible?

 

[jjm2]

hi logbook,

I am trying to get 10 bits of accuracy from a single bit, 1st
order sigma delta. My real question relates as much to the digital portion as to the analog.

How many samples from the 1 bit converter are required to get 10 bits of resolution after filtering and decimation?

Some of the literature I have read implies that you can use 64 bits from the analog portion to get 10 bits of output resolution through the filtering/decimation process. It appears to me that the minimum number of samples to get "10" bits of resolution is 513. The very first sample, assuming the sigma delta is cleared every conversion point gives you the MSB. You then need 512 more to get the value to the LSB. This doesn't seem quite right either though, since you can just use an adder to calculate the output. (essentially an LPF w/ 512 taps)

When I ran the AD simulator (linked above) It also took the 512 samples to measure a digital 0000000001 equivelent input.

Any thoughts or insight?

Thanks

 

[felixc]

What is the bandwidth of the signal that you want to read? I have used a 20-bit sigma-delta converter quite a while ago (AD7710 maybe?) and sampling speed was a tradeoff between the bandwidth of the signal to be measured and the precision of the conversion. The higher the bandwidth the lower effective bits, because of the higher samling rate incurred.

 

[jjm2]

The bandwidth I need is about 5kHz. So I'd like to get 10 bits at 10kHz.

Thanks,
JJM

 

[IRstuff]

Is this for school? There is little commercial reason to design your own ADC when Analog devices sells a 16-bit device for $27.06:

http://www.analog.com/en/prod/0,,AD7723,00.html

TTFN

 

[itsmoked]

And Linear sells 24bit devices for $5.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com