Designer’s Guide Community

New Member

Offline

Posts: 2

Common mode feedback compensation in single-stage folded cascode
Apr 29^th, 2011, 6:12am

Hello there,

I am designing what will be my first switched capacitor circuit. I started designing a fully differential folded cascode OTA. For some reason I had the very naive impression that by choosing a single-stage OTA I would have no concern with instability. Big mistake! Actually, I checked my specs and a single-stage is not a bad idea after all, but of course, the CMFB instability is a concern now.

After reading some textbooks and papers, my conclusion is that there isn't enough discussion about this issue and how to address it easily. I eventually found a home-made solution by adding some compensation capacitors as shown in the figures below (see the schematics of the OTA). I have probably reinvented the wheel somehow, but it seems to work. Look at the step responses �of the common mode bias with and without the compensation (figures).

Now, I wonder if there is a better solution than the one I found. Have anyone used that kind of compensation before? Is there a better solution?

In some textbooks they suggest to lower the CMFB gain, but I kind of disliked the concept. There should be more papers on the subject I think.

Any suggestions? please.

Cheers.

Back to top

folded_cascode_with_CMFB_circuit.jpg

IP Logged

Lex

Senior Member

Offline

Posts: 201
Eindhoven, Holland

Re: Common mode feedback compensation in single-state folded cascode
Reply #1 - Apr 29^th, 2011, 7:07am

Well your OTA may be a single stage, but looking at your CMFB loop, I see two stages. Since it is your first experience with CMFB regulation, I would advise you to draw the CMFB loop only, simplify it by making it look like single-ended, work out a frequency compensation of your choice, and then go back to the original scheme, and apply it. I think the method u used now is called capacitive narrowbanding, but i could be mistaking.

Back to top

IP Logged

wave

Senior Member

Offline

Posts: 117
Silicon Valley

Re: Common mode feedback compensation in single-stage folded cascode
Reply #2 - Apr 29^th, 2011, 10:41am

Stability of the CMFB is almost always harder than the main amp.

I'd check your NMOS current sources. �Your main amp does not seem to have an independent bias, but completely feed from the CMFB loop.
You could have startup and/or stability issues with that.

Also I think the Diode connected NMOS should be the FB point, not the drain of the other mirror.

Wave
Cheesy

Back to top

IP Logged

orient_bis

New Member

Offline

Posts: 2

Re: Common mode feedback compensation in single-stage folded cascode
Reply #3 - May 5^th, 2011, 5:05am

Hi there,

Sorry for the delay, I was out of office for some days.

@Alexandar: Yes, two stages indeed. I will follow your suggestion and analyze the CMFB loop separately. I will try to break the CMFB loop somewhere to apply some open-loop phase margin analysis. Thanks.

@Wave: Yes, it is now very clear to me that the CMFB loop will likely have one additional stage w.r.t. the diff amp. However, I am a bit confused with your comments. The main amp already has independent PMOS current sources. If I added another current source at the NMOS end, the circuit would have one bias source too many. I have seen somewhere that another independent current source may be added in parallel with the NMOS current sources driven by the CMFB amplifier. While that reduces the gain of the CMFB loop, which, in turn, improves the stability, I never thought of start-up issues with this circuit. Would you please elaborate your concern with start-up issues?

I am also confused with your second suggestion. I think you are suggesting to use diode-connected loads for the CMFB amp, which also reduces its DC gain quite significantly. Is my understanding right ? Thanks.

I think that having a high DC-gain but correctly compensated CMFB amplifier is a more appropriate design approach. However, if anyone already sees a flaw in my reasoning (and schematic), please let me know. I don't want have bad surprises when my silicon is back 6 months from now... Huh

Back to top

IP Logged

Lex Senior Member Offline Posts: 201 Eindhoven, Holland	Re: Common mode feedback compensation in single-stage folded cascode Reply #4 - May 5^th, 2011, 8:06am wave wrote on Apr 29^th, 2011, 10:41am: ... Also I think the Diode connected NMOS should be the FB point, not the drain of the other mirror. I don't think so. That would be a positive feedback loop.
Back to top	IP Logged

wave

Senior Member

Offline

Posts: 117
Silicon Valley

Re: Common mode feedback compensation in single-stage folded cascode
Reply #5 - May 5^th, 2011, 10:27am

Visualize this:
Vop=Von, and both happen to initialize to Vdd.
That means the current in 1/2 your CMFB amp is OFF.
That means your Nmos diode connected mirror is OFF. �(image is too blurry to read the label.)
That means your CMFB look collapses.

Your node VBCM is high Impedance at all points.
There is no Low Z control point.

Wave
Wink

Back to top

IP Logged

orient

New Member

Offline

Posts: 6

Re: Common mode feedback compensation in single-stage folded cascode
Reply #6 - May 5^th, 2011, 6:02pm

Hi Wave,

Thanks for your comments. Well, in the situation you just described, because VCM is a reference voltage and will be set at start-up, the CMFB amp's output �(VBCM) will be pulled up to Vdd by the branch that isn't shut off. �This will bring Vop and Von down towards VCM, which, in turn, will fully enable the CMFB amplifier (the diode-connected branch will start conducting).

Alternatively, if at start-up, both Vop and Von are grounded, the opposite just occurs. The diode-connected branch will make sure Vop and Von are pulled up towards VCM.

Do you see my point? Does it make sense for you?

I think the situation you described occurs in other circuit topologies, perhaps if the CMFB was also a folded-cascode or something like that, but I am just speculating here.

Cheers.

Back to top

IP Logged

zwtang Junior Member Offline Posts: 11 Shanghai, P.R.China	Re: Common mode feedback compensation in single-stage folded cascode Reply #7 - May 14^th, 2011, 7:06am Hi, orient_bis Since OTA is used in switched-capacitor circuits, why not use a switched-capacitor CMFB circuit? I suggest you read this paper: TCASII, No.12, Dec. 2003, "Analysis of Switched-Capacitor Common-Mode Feedback Circuit". zwtang 2011/5/14
Back to top	IP Logged

orient

New Member

Offline

Posts: 6

Re: Common mode feedback compensation in single-stage folded cascode
Reply #8 - May 15^th, 2011, 5:58pm

Hi zwtang,

Thanks for the advice, indeed I also noticed that switched-capacitor CMFB is very popular too.

This will be my first switched-capacitor circuit, so I was wishing to use conventional CMFB first, then perhaps in the future I can look into it.

To be honest, at this stage, power consumption is not a big problem for me. Perhaps settling-time is more important.

Is there any penalty in settling time if I use a switched-capacitor feedback?

Don't feel forced to answer, I can look into the paper you suggested later on. Thanks.

Cheers.

Back to top

IP Logged