Designer’s Guide Community

carlgrace Senior Member Offline Posts: 231 Berkeley, CA	Deep Nwell needed for ADC? Dec 17^th, 2010, 3:16pm Does anyone have any experience with deep nwell? Does it really help much for, say, a 12b 80MSPS pipelined ADC? I built one years ago in 0.25um CMOS without deep nwell. Now I'm building another in 65nm but I would prefer not to use deep nwell so I can reduce layout size. Do you guys think this is a mistake, or can I get away with it?
Back to top	IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Deep Nwell needed for ADC?
Reply #1 - Dec 20^th, 2010, 7:11am

carlgrace wrote on Dec 17^th, 2010, 3:16pm:

Does anyone have any experience with deep nwell? �Does it really help much for, say, a 12b 80MSPS pipelined ADC? �I built one years ago in 0.25um CMOS without deep nwell. �Now I'm building another in 65nm but I would prefer not to use deep nwell so I can reduce layout size. �Do you guys think this is a mistake, or can I get away with it?

it's difficult to say. there are 2 points which come to my mind:

1. Improved isolation against substrate noise: This is probably the big advantage, although I don't know how good the isolation remains as you go faster and faster. Obviously, you need to use good shielding.

2. Impact ionization: If you are planning on using folded cascode, maybe with gain boosting to cut down on power in the amps, then you may run into issues with impact ionization in the N-cascode. Deep Nwell technology allows you to use a local P-well for the NMOS, permitting you to tie B-S together and avoid the Rout degradation, but depending on the topology you choose, this may or may not be relevant.

You need to consider the cost-performance tradeoff. The only times I used Deep-Nwell was when I anyway needed it for some other reason, like HV support, or when the technology used anyway happened to have it.

Vivek

Back to top

IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Deep Nwell needed for ADC?
Reply #2 - Jan 22^nd, 2011, 1:02pm

The deep N-Well concept (10um well) was introduced as a concept on IBM foundry processes. Buried inside IBM there was a study done (I initiated it about 8 years back) on how much these things help with noise isolation,.

Bottom line - its unique to the design and it helps a few dB but its not the savior of noise. With and without deep nwell isolation were used and the noise coupling differences on some test circuits was like 2-4 dB difference and sometimes it made no difference.

Not needed, use good design practices and be done with it.

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

Berti Community Fellow Offline Posts: 356	Re: Deep Nwell needed for ADC? Reply #3 - Jan 25^th, 2011, 3:08am Hi Jerry, Are those 2-4dBs from the substrate isolation only? I made the experience that coupling is often dominated by coupling between metal wires, pad-ring etc., and not the substrate. Regards
Back to top	IP Logged

loose-electron

Senior Fellow

Offline

Best Design Tool =
Capable Designers

Posts: 1638
San Diego California

Re: Deep Nwell needed for ADC?
Reply #4 - Jan 27^th, 2011, 3:01pm

Berti wrote on Jan 25^th, 2011, 3:08am:

Hi Jerry,

Are those 2-4dBs from the substrate isolation only?

I made the experience that coupling is often dominated by coupling between metal wires, pad-ring etc., and not the substrate.

Regards

Yes - to pull all the other variables out of the equation the layouts were exact same everything - spacing routing, circuits etc, the ones with/without Deep Nwell tubs were just with/without the tubs, and bare substrate was left in place on some test cases, and then regeular depth Nwell was used in the third case (there was a bunch of different situations that all got tested.)

The report was not too widely published because at the time IBM thought they had something great in the deep nwell concept, and the reality of the matter was more along the lines of "it helps a few dB, but thats it" which is not what they wanted to hear.

I totally agree with you that there are som many other coupling paths that have not been accounted for.

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

Berti Community Fellow Offline Posts: 356	Re: Deep Nwell needed for ADC? Reply #5 - Jan 28^th, 2011, 4:35am Well, but that means the improvement of 2-4dB �s for "substrate coupling"+"whathever else". The actual improvement which comes from the deep nwell might therefore be slightly higher than 2-4dB (even only slightly), as basically two sources are added ... Regards
Back to top	IP Logged

loose-electron Senior Fellow Offline Best Design Tool = Capable Designers Posts: 1638 San Diego California	Re: Deep Nwell needed for ADC? Reply #6 - Jan 28^th, 2011, 4:46pm the actual improvement seen from the use of Deep N well was 2-4 dB I can not state it any more clearly than that
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