The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Basic ADC query
https://designers-guide.org/forum/YaBB.pl?num=1177655417

Message started by adesign on Apr 26^th, 2007, 11:30pm

Title: Basic ADC query
Post by adesign on Apr 26^th, 2007, 11:30pm

Dear all,

Why can't we get 20-bit resolution with pipeline or SAR ADCs. What would be the issues if we increase the bumber of bits in Nyquist ADCs?

Regards,

Title: Re: Basic ADC query
Post by ywguo on Apr 28^th, 2007, 3:38am

1. Device mismatch.
2. thermal noise and flicker knoise.

I think those are two basic limits.

Title: Re: Basic ADC query
Post by sheldon on Apr 28^th, 2007, 6:20am

Adesign,

Just to expand on Yawei's comments about device matching, typically
mismatch is on the order of 1:1000. So to achieve 20 bits of accuracy
would require 10 bits of segmentation in the first stage. This level of
segmentation limits performance and yield.

In addition at this level of resolution there are capacitor effects
that need to be accounted for including, capacitor non-linearity and
memory effects.

The effect of capacitor non-linearity is discussed in the following
paper,
"Fully Differential ADC with Common-Mode Range and
Rail-to-Rail Nonlinear Capacitor Compensation", Hester,
et al., JSSC, Feb. 1990

Designing for 20 bits of resolution is difficult because accuracy
is sensitive to higher order effects including material properties.

Best Regards,

Sheldon

Title: Re: Basic ADC query
Post by adesign on Apr 29^th, 2007, 8:35pm

Thanks all for your valuable comments....

Title: Re: Basic ADC query
Post by krishnap on Apr 30^th, 2007, 11:40pm

Speed of operation will also be limited, with the High resolution.

Title: Re: Basic ADC query
Post by loose-electron on May 23^rd, 2007, 6:15pm

Noise will become the one that will bite you.

I have done 14/16 bit conversion, but you need to get very fussy about thermal noise, linearity and offsets, and due to sampling accuracy your sampling rate has to go down. There are some low convert rate devices out there at 20 bits I believe, but they are in KHz convert rates.

After you get around the problems with thermal/flicker noise, you can do some digital averaging to reduce the noise some more, but again, your conversion rates have gone down some more.

The prior comments by others are very valid also - some can be solved (the non-linear capacitor can be corrected for with digital curve compensation for example) but there are a large number of subtle things that become an issue.

An antenna, with a 28 bit converter connected to it, converting at 6GHz is not going to happen in the forseeable future.
Never say never!

;)

Title: Re: Basic ADC query
Post by RobG on May 25^th, 2007, 9:44am

adesign wrote on Apr 26^th, 2007, 11:30pm:

Dear all,

Why can't we get 20-bit resolution with pipeline or SAR ADCs. What would be the issues if we increase the bumber of bits in Nyquist ADCs?

Regards,

Dr. Temes asked me this the other day. Not because I knew, but because he was grilling me for an exam :-). Like LE alluded to, mismatch and other issues (finite opamp gain, offset, etc) can be calibrated out, but thermal noise cannot. It is for this reason that oversampling (noise averaging) is more practical for higher resolution.

rg

Title: Re: Basic ADC query
Post by mg777 on May 25^th, 2007, 10:26am

I was intrigued enough by this question to go and re-read (perhaps for the twentieth time) Feynman's timeless chapter on the Brownian ratchet. Afterwards I was doodling and came up with an interesting observation:

kT/q is a voltage & so is sqrt(kT/C). So I idly equate the two at room temp and get C = 6.4 aF (a=atto) which is interestingly close to the parallel plate Cox of a W=1u MOS transistor. I wonder what that means (there's a resistor in there somewhere).

On a more airy note: there's much physical richness in the behavior of atoms, and oxides are already a few atoms thick - so I'd say that the analog designer of 10-15 years from now (I hope I'm not completely senile by then) will be working with completely different paradigms. A celebrated analog design text would have expanded to 16 authors (8 of whom are from the physics department) and an SC filter would refer to Schrodinger's Cat.

M.G.Rajan
www.eecalc.com

Title: Re: Basic ADC query
Post by loose-electron on May 25^th, 2007, 6:46pm

Quote:

so I'd say that the analog designer of 10-15 years from now (I hope I'm not completely senile by then) will be working with completely different paradigms.

Hm... Well, the methods used have not really changed in the last 27 years that I have been involved.

Smaller sizes, smaller voltages, higher frequencies, lower currents, more transistors, but the methods are still pretty close to being the same. More dependency on EDA is the big one I have noticed.

Jerry

Title: Re: Basic ADC query
Post by mg777 on May 26^th, 2007, 8:14am

Here are some interesting references - I realize I'm getting off-topic, so I'll leave it here.

1. An Ising model description of sigma-delta data converters
http://adsabs.harvard.edu/abs/2000APS..MARY25008M

2. Novel Analog-to-Digital Conversion Architecture Using Electron Waveguides
http://www-mtl.mit.edu/~alamo/pdf/1993/RJ-59.pdf

3. From analog to digital: exploring cell dynamics with single quantum dots.
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=16284775&dopt=Abstract