The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> CMFB loopbandwidth
https://designers-guide.org/forum/YaBB.pl?num=1254307398

Message started by raja.cedt on Sep 30^th, 2009, 3:43am

Title: CMFB loopbandwidth
Post by raja.cedt on Sep 30^th, 2009, 3:43am

hi,
can any body please explain on what basis cmfb loop bandwidth has to decide? is there any relation between op amp loop bandwidth and CMFB loop bandwidth?

Thanks,
rajasekhar.

Title: Re: CMFB loopbandwidth
Post by loose-electron on Oct 2^nd, 2009, 10:08am

In a theoretical sense, the two can be totally uncorrelated. You could (in the world of math at least) design a system that has two different bandwidths, one for the op-amp and one for the CMFB system.

The practical reality is something else - the circuits are shared and exhibit BW characteristics that track each other.

Where the difference comes in is the setting of the op-amp gain/phase dominant pole, and the setting of the CMFB dominant pole.

Academic wisdom is that you set the CMFB BW as high as you can go to get good CMFB rejection over a wide frequency range. However in some applications, CMFB gets used to "keep the operating point centered" and can use a lower BW. Depends on your input signals and what you are trying to get done.

Jerry

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 4^th, 2009, 10:56am

hi,

CMFB Loop works in parallel with the opamp loop...Although CMFB loop shares the stages of the opamp...But you can design CMFB loop characterstics irrespective of the differential Loop through the extra CMFB Gain Stage.
Normal CMFB Loop Bandwidth depends on expected imbalance/error in output common mode and common mode transients in your circuit. Ideally, CMFB BW should be as high as opamp BW, but general trend is to keep CMFB BW atleast 30 % of the OPAMP Gain BW.

Also, if you want to alter the CMFB bandwidth, you can always compensate it by placing passive R-C or just C compensation after the averaging node.

This is a generic post...Lemme know a bit more specifically about the application of your ckt.

regards,
Mayank.

Title: Re: CMFB loopbandwidth
Post by raja.cedt on Oct 4^th, 2009, 10:52pm

hi,
thanks for your reply, but generally from what spec people used derive CMFB bandwidth...i think from supply profile and input signal frequency, is it true?

Thanks,
Rajasekhar.

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 5^th, 2009, 12:03am

Hi,
Yeah, you are correct. As i said, CMFB BW is generally decided based on common mode transients in your ckt. Supply & Substrate Noise are the main contributors to CM transients.
I am still not sure how it is connected to input signal freq. Maybe, [in connection to your query on diffamp] , mismatch in load transistors might generate a CM variation at o/p node which is a direct function of i/p signal freq. Is this interpretation correct ??

-- Mayank.

Title: Re: CMFB loopbandwidth
Post by raja.cedt on Oct 5^th, 2009, 8:54pm

hi Mayank,
output common mode should settle much faster than the signal, other wise you will see differential mode gain variation, This is how i understand. You can correct me if am wrong.

Thanks,
Rajasekhar.

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 7^th, 2009, 8:04am

Hi Raja,
I thought over it for quite some time. The way I see it is that till the time your o/p CM Voltage is within your OCMR range of the OTA/Opamp, (which is defined by the limits for MOSes in o/p branch to be in saturation region), the variation in Differential Gain is negligible.
Hence, your o/p CM can vary by a slight amount within your OCMR range for some time until it finally settles to the CM reference you have set up in your CMFB Loop.

The effect you are talking about will be visible only if CM transient variation violates OCMR of the OTA/opamp.

Always do a transient simulation by giving a Common Mode Step and see the CM transient Response of your circuit. CMFB BW of around 30% of Differential BW covers around 95 % settling of CM value at o/p.

Plz. Tell me if the explanation is satisfactory/faulty.

thanx,
Mayank.

Title: Re: CMFB loopbandwidth
Post by Berti on Oct 11^th, 2009, 11:42pm

CM variations can result in differential signals due the CM-to-DM gain of the amplifier (caused by mismatch).
Ideally, the CM bandwidth of the amplifier is therefore comparable to the DM bandwidth. However, a high CM bandwidth is often related to higher power consumption. This results in a trade-off between power consumption/circuit complexity and robustness of your amplifier against CM variations. The 30% Mayank claims seem to be arbitrarily ... I would rather suggest to evaluate this trade-offs for your particular design.

Cheers

Title: Re: CMFB loopbandwidth
Post by raja.cedt on Oct 11^th, 2009, 11:54pm

hi berthi,
tx for your reply, actually that is the i am asking from starting onwards. I donno how to evaluate that.. i wanted to apply some bump in the supply and check weather it is effecting differential gain through common mode change at the output.

Thanks,
Rajasekhar.

Title: Re: CMFB loopbandwidth
Post by Berti on Oct 13^th, 2009, 10:18pm

Hi Rajasekhar,

Most of the time I work with switched-capacitor CMFB. In this case I usually sweep the CMFB capacitor at the ouput of the OTA and plot it versus DM and CM bandwidth. With increasing capacitor, the CM bandwidth gets larger while the DM mode bandwidth is compromised. Unless the application is very critical concerning CMRR/PSRR I choose the capacitor so that I don't compromise to much of the DM bandwidth (10-20%) and still get reasonable CM bandwidth.

I guess such a procedure would also be possible for other CMFB circuits.
Other opinions are welcome!!!

Regards

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 14^th, 2009, 7:23pm

Hi,
CMFB Loop Compensation can also be done by placing a capacitor to ground/vdd after the averaging node. This will modify CMFB BW while having minimal effect on the Differential Mode BW.
This way you can keep the CMFB S-C at OTA o/p less & still get good CMFB BW through this kind of compensation.

thanx,
Mayank.

Title: Re: CMFB loopbandwidth
Post by Berti on Oct 14^th, 2009, 10:40pm

Hi Mayank,

Please correct me if I am wrong, but placing a capacitor as you describe in you post indeed increases the phase margin as you kill the gain of the CMFB, but it doesn't improve the common-mode bandwidth. I think that this solution therefore gives you a very poor results.

By the way, I was not talking about stability/compensation in my previous post but just selecting the capacitor of the CMFB large enough to achieve sufficient common-mode bandwidth.

Cheers

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 15^th, 2009, 12:14am

Hi,

Oh my Bad !! I thought we were moving on to compensation of CMFB Loop which should be independent of Diff. Loop. I didnt say it would increase CMFB BW, I wanted to say you could alter it without much effect on Diff Loop BW so as to make your CMFB Loop stable.

regards n sorry,
Mayank.

Title: Re: CMFB loopbandwidth
Post by duncandu on Oct 24^th, 2009, 10:19am

raja.cedt wrote on Sep 30^th, 2009, 3:43am:

hi,
can any body please explain on what basis cmfb loop bandwidth has to decide? is there any relation between op amp loop bandwidth and CMFB loop bandwidth?

Thanks,
rajasekhar.

No relationship for continuous-time filter, I think. Just be careful, sometimes, CMFB is also a two-pole system and you should care common-mode oscillation.
For switched filter, the settling of CMFB is also a concern, but settling time for CMFB is not directly related to its bandwidth. You have to figure it out by simulation.

Title: Re: CMFB loopbandwidth
Post by Berti on Oct 26^th, 2009, 4:27am

Duncandu, I think your post is full of flaws:

Quote:

No relationship for continuous-time filter, ....

Often the CMFB has some parts in common with the main Opamp. Hence, there is typically a relationship between CM and DM bandwidth.

Quote:

For switched filter, the settling of CMFB is also a concern, but settling time for CMFB is not directly related to its bandwidth.

What determines the settling time if not the bandwidth ????

Cheers

Title: Re: CMFB loopbandwidth
Post by Mayank on Oct 26^th, 2009, 6:34am

Hi Ducandu,
i beg to differ. Settling Time is always INVERSELY PROPORTIONAL to the bandwidth of the system...Higher the CMFB BW, faster is the o/p CM settling.
Further CM Oscillations (i suppose you are referring to the ringing during CM settling) depends directly on Phase Margin of the CMFB Loop.

@ Berti : say, for example, We consider a generic 2-stage Rz-Cc compensated OTA with a CMFB Control Loop. Generally CMFB employs one extra Gain Stage and returns the feebback to some point inside the first stage of the OTA(which in most cases is the gate voltage of the Load Transistors in the 1st stage). This extra gain stage of CMFB Loop introduces another Pole into the CMFB Loop, while the tf of the 1st & 2nd stage will retain it's poles and zeros (with slight change in gain coz now we are looking at CM instead of DM) .
BottomLine : You are correct. There is a certain relationship b/w CM & DM BW.
My Question : Without any extra compensation for CMFB Loop(like placing RC compensation post the averaging node), Is CMFB BW always lesser than DM bandwidth ??

regards,
Mayank.