The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> How to improve op's noise performance
https://designers-guide.org/forum/YaBB.pl?num=1199959344

Message started by trashbox on Jan 10^th, 2008, 2:02am

Title: How to improve op's noise performance
Post by trashbox on Jan 10^th, 2008, 2:02am

For a folded-cascod op, I want to improve its noise performance. In generally, increase W*L and W/L can decrease flick noise and thermal noise, however there is no enough layout area margin.

In theory, increasing the current can improve it, however the simulation shows the noise becomes a little larger with the current increase with UMC model. It is strange.

Is there any other method to improve op's noise performance without increasing area obviously? Thanks.

Title: Re: How to improve op's noise performance
Post by thechopper on Jan 10^th, 2008, 8:06am

Hi Trashbox,

Concerning flicker I guess the only way out is the one you mentioned. Concerning thermal, you can play around with the input gm1 by either putting more tail current or increase W/L as you pointed out. However, from the output integrated noise this will not be useful, since the BW under which you have to integrate that noise is proportional to the input gm (and inversely proportional to the compensation capacitor. Therefore the effect of gm gets canceled out.
Bottom line, no matter what you do with your input gm, the one setting the output integrated noise level will be the compensation capacitor (the larger the better for reducing noise). This is true assuming the input gm is much larger than the gm of the folded cascode circuit, which is valid for most of the cases.
The output integrated thermal noise (noise power) will result equal to (4/3)KT/Co, where Co is the compensation cap.

Reducing noise is always about increasing power or area.
Hope this helps

Tosei

Title: Re: How to improve op's noise performance
Post by vivkr on Jan 11^th, 2008, 4:02am

Hi,

A few more points in addition to the excellent ones from Tosei:

1/f noise may also be easily modulated out of band with chopping or correlated double sampling (CDS) at the cost of additional complexity and in
case of CDS, extra noise and power consumption.

An important point which is often overlooked is the possibility of actually improving the noise performance of your amplifier itself. The (4/3)kT/Co
limit is the limit on the total noise power in the entire relevant bandwidth of the opamp. However, this assumes

(a) that the noise comes only from the input pair. An assumption which is unfortunately not true. The other transistors will also add noise,
and depending on the necessary swing and the quality of design, the additional noise from these devices might become significant. In a folded
cascode amp, there are noise contributions from the P and N current mirrors in addition to the input pair, if we ignore the cascode noise contribution.
The gm of these mirror devices must be significantly smaller than the gm of the input pair to suppress their noise. In a poorly optimized design, this
is not the case. You should try to quantify the total noise power in band vs. the noise power due to the input pair as a quality factor. A good value
to aim for is to get the total noise power of the amp to be between 20% - 30% greater than the contribution of the input pair. If you can achieve this
goal, then you can be happy.

(b) Remember that although the total noise contribution is limited by C0, the noise floor will be reduced with higher gm. This helps if you are
aiming for a continuous-time application where the noise is of interest in a small bandwidth only, or where you have some stages later on which
limit the bandwidth.

(c) Finally, you should seriously consider using another opamp topology. Folded cascode opamps are the noisiest from a simple analysis
perspective because you have 4 additional transistors (apart from the input pair) adding their noise.

Regards
Vivek