Designer’s Guide Community

reactgary New Member Offline Posts: 9	Noise Budget in Pipelined ADC May 21^st, 2006, 8:35pm Hi all, What total input refered noise level should we keep in a ADC design, should it be less than 1/2 of the Qutanizational noise of the adc, or it should be ok if it is less than Qnoise of the ADC? Thanks Gary
Back to top	IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Noise Budget in Pipelined ADC Reply #1 - May 22^nd, 2006, 9:36pm Gary, The budget depends on how much SNR or SNDR you want. Best regards, Yawei
Back to top	IP Logged

vivkr Community Fellow Offline Posts: 780	Re: Noise Budget in Pipelined ADC Reply #2 - May 22^nd, 2006, 11:57pm Hi Gary, Another important point to reckon with is the power you are willing to burn. If you want higher SNR, then you will inevitably need much more power. You have to decide the acceptable power consumption along with acceptable SNR levels. Regards Vivek
Back to top	IP Logged

reactgary

New Member

Offline

Posts: 9

Re: Noise Budget in Pipelined ADC
Reply #3 - May 23^rd, 2006, 8:23am

Thanks All,

I think I know the key concept about the tradeoff between SNR / Power.
I want to know more about the inherent noise budget in a pipelined ADC.
For example, if I am going to make a 10bit ADC, with 1.5bit / stage,
The Qnoise of the second stage would be (VFS/512)^2/12 <<<<< let it be Qa
So, what is a reasonable choice of the noise budget of the first stage
Someone told me that we should keep it half of Qa.
Is it reasonable? Thanks for all the opinions of you guys!

Gary

Back to top

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Noise Budget in Pipelined ADC Reply #4 - May 23^rd, 2006, 9:24pm Gary, The Quantization error is (Vfs/1024)^2/12 for a 10 bit ADC. The key point is to keep the overall thermal noise not more than the quantization error. Best regards, Yawei
Back to top	IP Logged

xaviar New Member Offline Posts: 3	Re: Noise Budget in Pipelined ADC Reply #5 - Jul 12^th, 2006, 4:59am Gray, you can check in some PH.D dissertation in UC berkerly,in the most case Qnoise is set to be (1/6)LSB,it will give SNR 1~2dB buget. If you locate Qnoise nearly to (1/2)LSB,it will make your SNR lower.
Back to top	IP Logged

TT

New Member

Offline

Posts: 7

Re: Noise Budget in Pipelined ADC
Reply #6 - Jul 12^th, 2006, 7:18pm

Gary.

if you do 1.5bit/stage then the noise of the second and third stage may not be neglectible, while having more bits
at the input will reduce the noise requirement of the subsequent stage, though increasing a bit the complexity of
the first stage. Also do not forget to add contribution of the voltage reference which are in practice REALLY not neglectible
and usually requires external caps or high current.

TT

Back to top

IP Logged

Sid

Community Member

Offline

Posts: 35

Re: Noise Budget in Pipelined ADC
Reply #7 - Jul 15^th, 2006, 6:08pm

Hi,

It is not power efficient to be limited by Quantization Noise, especially at HIGH resolution/SNR. For example, you will design a 14-bit converter even if you want ENOB = 12 (i.e. SNR = 72 dB). This is because you can reduce the quantization noise to the 14-bit level and allow more thermal noise (hence the caps can be smaller thereby saving power). Of course, if you do this, you will need a higher gain*bandwidth 1st stage OTA to settle to 14-bit level and have 14-bit gain-error. So, the design choice for "how many bits?" is basically made based on whether you save more power by having smaller caps or less "accurate" OTAs?

For the 10-bit case, I guess having quantization noise about equal to thermal noise is probably the way to go.

-Sid

Back to top

IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Noise Budget in Pipelined ADC
Reply #8 - Aug 3^rd, 2006, 3:16pm

A little thinking here. this is about SNR at each stage, as it relates to the residue of each stage in the pipeline.

More specific, it is the IRN (input referred noise) at each stage.

That said, you might want to build a math model of this. (or even just an Excel spreadsheet with some equations in it.) Keep track of the signal IRN, the residue, and look at the compounding noise vs. gain through each conversion stage.

If you put in the gains, noise, of each stage and the resulting SNR's at each output, you can define an aceptable SNR, and gain a better understanding of the noise down the chain.

Back to top

Jerry Twomey
www.effectiveelectrons.com
Read My Electronic Design Column Here
Contract IC-PCB-System Design - Analog, Mixed Signal, RF & Medical

IP Logged