The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Stability analysis of a continuous time CMFB
https://designers-guide.org/forum/YaBB.pl?num=1492096904

Message started by polyam on Apr 13^th, 2017, 8:21am

Title: Stability analysis of a continuous time CMFB
Post by polyam on Apr 13^th, 2017, 8:21am

Hi,

I am trying to check the stability of my CMFB. To do so, I am using the test bench (TB) shown in the screenshot. The gm cell is a telescopic OTA and the schematic of the CMFB is also shown. I have several problems and doubts.

1- Am I using a proper TB to examine the stability of the CMFB? I am using cmdmprobe in analoglibe to break the loop. (I tried to follow this tutorial: https://pdfs.semanticscholar.org/c1dc/91d09c4112cf9aab8baefcc345bd2141fcdd.pdf ).
When CMDM -->1 it measures common mode response
When CMDM -->-1 it measures differential mode response.

2- I don't know how to set the value of the resistors. The way to choose the value of the resistors on the TB is not clear to me.

3- The results I get from the stb analysis is not rational to me. I cannot interpret the results. Please see the stb responses when CMDM is 1 and -1. Is this the way I have to follow to simulate the phase margin of the CMFB?

Thank you

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 13^th, 2017, 8:22am

The schematic of the gm-cell

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 13^th, 2017, 8:23am

The schematic of the CMFB.

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 13^th, 2017, 8:24am

STB response when CMDM is 1.

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 13^th, 2017, 8:24am

STB response when CMDM is -1.

Title: Re: Stability analysis of a continuous time CMFB
Post by Ken Kundert on Apr 13^th, 2017, 12:28pm

Quote:

1- Am I using a proper TB to examine the stability of the CMFB? I am using cmdmprobe in analoglibe to break the loop.

The cmdmprobe has been deprecated. It is inaccurate in certain situations. You should use back-to-back ideal baluns. Alternately, you can use the diffstbprobe.

Quote:

2- I don't know how to set the value of the resistors. The way to choose the value of the resistors on the TB is not clear to me.

Your testbench should mimic the circuit in which your amplifier resides. You don't want to know the stability of your amplifier, you want to know the stability of the circuit that the amplifier is embedded in. Asking the stability of the amplifier is the wrong question.

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 14^th, 2017, 11:24am

Hi Ken,

Thank you very much for replying my question. I checked "diffstbprobe" and it seems to be a nice solution. Let me be more clear on my problem and the reason behind my previous question. I am designing a multi-stage (3-stages in my case) amplifier with no capacitor feedforward compensation. Please see the screenshot. Three CMFBs are used to control the CM level. CMFB 1 for gm1, CMFB2 for gm2 and gmp2 and finally CMFB3 for gm3 and gmp3. The op-amp is going to be used in a CT delta-sigma modulator with the active-RC integrator. I performed ac analysis and the system satisfies my specs. I attached the ac response of the op-amp.
I also simulated my CT-DSM with the op-amp and it didn't work. I suspected that there was something wrong with my CMFBs (CMFB1,2,3). That's why I performed transient analysis on my op-amp and the op-amp gets unstable. Please see the step response.

If I break my op-amp into three parts namely

gm1+CMFB1 ( I already showed this part in my last post)
gm2+gmp2+CMFB2
And gm3+gmp3+CMFB3

I would be able to check the stability of each part individually. (That's what's happening in my head, I don't know if I am truly right).

Having had above explanation, several questions came to me:
1- Am I thinking in a right way? (Checking the stability of individual parts)
If so, What would be a proper test bench and proper analysis?

2- Could you please let me know what I have to do in such a case?

3- Should I put my entire op-amp in a proper TB (in my case an active RC-integrator) and then check the stability of the CMFBs?

Sorry for a bunch of questions!
Many thanks

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 14^th, 2017, 11:25am

AC-response of the 3-stage op-amp

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 14^th, 2017, 11:25am

Step response of the op-amp

Title: Re: Stability analysis of a continuous time CMFB
Post by Ken Kundert on Apr 14^th, 2017, 8:18pm

I see no feedback at all.

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 16^th, 2017, 9:21am

Ken,

I attached all screenshots of my op-amp in a single ZIP file.

"3Stage_opamp.png" shows a 3-stage no capacitor feedforward compensation op-amp. gm1, gm2, and gm3 all together make a 3rd order path. gmp2 is located in a 2nd order path and gmp3 is located in a first order path. It shows the topology of the designed op-amp without placing it in any feedback configuration (The transistor level schematics of all above gms are attached).

I also have three CMFBs (please see CMFB.png):

CMFB1 controls the CM level of gm1.

CMFB2 controls the CM level of gm2 and gmp2. (I also attached the schematic of gm2 and gmp2).

CMFB3 controls the CM level of gm3 and gmp3. (Please see the schematic of the gm3 and gmp3 and how they are related together)

Let me review some of my observations:

1- I performed an ac analysis on my op-amp. Please see "ac_response.png". It shows 82 dB DC-gain, 1.8GHz UGBW, and 65-degree phase margin with 3pF load capacitor.

2- I also applied a sin wave to the input of the op-amp. Input amplitude=100uV, input frequency=4MHz.

"in_tran.png" shows the input differential signal.

"gm1_tran.png" shows the tran response at the output of gm1.

"gm2_tran.png" shows the tran response at the output of gm2 (and also gmp2).

"gm3_tran.png" shows the tran response at the output of gm3 (and also gmp3)

The tran response looks logical until 1us and then it seems that the op-amp gets unstable. ( I think there is something wrong with my CMFBs).

Now, I have the following questions:

1- Is what I am guessing correct? CMFBs are not working properly?

2- I want to check the stability of the CMFB (to see the phase margin) no matter what my previous question is right or wrong. To do so, obviously, I have to put the op-amp in a feedback test bench (TB). Could active RC-integrator be a proper test bench since I am going to use this op amp in a CT delta sigma modulator with active RC integrators?

2-1 I think I have to use stb analysis to examine the stability of my CMFBs. So, I have to use three diffstbprobes to break the CM loops while op-amp is placed in a TB asked in question 2. Please put me right if I am wrong.

Many Thanks

Title: Re: Stability analysis of a continuous time CMFB
Post by Ken Kundert on Apr 17^th, 2017, 1:23am

I'm sorry, but I don't open zip files.

Can you identify the feedback loop or loops?

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by ULPAnalog on Apr 17^th, 2017, 2:26am

I took a chance to check the content of the zip. My understanding is that the op-amp itself is feedforward compensated with local feedback around gm stages to set the common mode. I have a couple of questions to polyam.

1. Your CMFB has a structure by name single OTA but I do not see what is inside it. Is it a single stage OTA?
2. Did you verify that each OTA with its CMFB is stable? Instead of attacking the 3 stage opamp directly, you may want to first strip it down to a single OTA with CMFB and check if it stays stable. I wonder if the CMFB loop, itself being a two pole system (one from single OTA and other from the pmos device, the gate of which is being controlled by CMFB loop), is adequately compensated.

Thanks and regards

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 18^th, 2017, 8:23am

Sorry ken, I should have guessed that.
I attached the top-level schematic of the op-amp described earlier.

Loop #1: gm1 and its CMFB1

Loop#2: gmp2 and its CMFB2. CMFB2 sets CM level of gm2 and gmp2.

Loop#3: gmp3 and its CMFB3. CMFB3 sets CM level of gm3 and gmp3.

Thank you

Ken Kundert wrote on Apr 17^th, 2017, 1:23am:

I'm sorry, but I don't open zip files.

Can you identify the feedback loop or loops?

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 18^th, 2017, 8:25am

Ken,
I don't know if you open pdf files or not. But I've inserted all figures in a single pdf file.
Thank you

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 18^th, 2017, 8:44am

ULPAnalog,

Thank you for your reply. Yes! your understanding is absolutely right.

To answer your first question, I am using a differential difference amplifier (DDA) for the firs OTA (gm1) and a so-called 5-transistor amplifier for others (gmp2 and gmp3). Please see the screenshot in CMFB123.pdf.

And about your second question, as a matter of fact, that's exactly what I was asking in my previous posts. I want to check the stability of each gm-cell (OTA) with its corresponding CMFB individually. BUT I don't know how to set up a test bench to do that :'(. That's my problem.

I attached my top-level schematic of the op-amp with "diffstbprobe" to check the stability of each loop. Can you please verify if my test bench to check the stability of the CMFB is right?

Thank you

ULPAnalog wrote on Apr 17^th, 2017, 2:26am:

Title: Re: Stability analysis of a continuous time CMFB
Post by polyam on Apr 18^th, 2017, 8:46am

That's the way I placed "diffstbprobe" on my op-amp to examine the stability of CMFBs. Is that a right TB?

ULPAnalog wrote on Apr 17^th, 2017, 2:26am:

2. Did you verify that each OTA with its CMFB is stable? Instead of attacking the 3 stage opamp directly, you may want to first strip it down to a single OTA with CMFB and check if it stays stable. I wonder if the CMFB loop, itself being a two pole system (one from single OTA and other from the pmos device, the gate of which is being controlled by CMFB loop), is adequately compensated.

Thanks and regards

Title: Re: Stability analysis of a continuous time CMFB
Post by Ken Kundert on Apr 18^th, 2017, 10:29am

I would recommend that as long as the simulation time is not unbearable you run three stb analyses with the amplifiers in place. Then if you want see loop gain, use the loop gain computed by the stb analysis -- do not compute it yourself as you did in your figures.

Finally, the diffstbprobe is not going to work for you. You are interested in checking the stability of the common mode loop, the diffstbprobe ignores common mode signals. Instead you use back to back ideal baluns to decompose the feedback signal into differential and common mode components, and then inject the test source into the common mode path.

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by ULPAnalog on Apr 19^th, 2017, 2:14am

Quote:

Finally, the diffstbprobe is not going to work for you. You are interested in checking the stability of the common mode loop, the diffstbprobe ignores common mode signals. Instead you use back to back ideal baluns to decompose the feedback signal into differential and common mode components, and then inject the test source into the common mode path.

Dear Ken

Isn't diffstbprobe actually made from back to back ideal baluns? My understanding was that the diffstbprobe, based on the choice of common mode stb analysis or diff mode stb analysis, will automatically perform the injection of test source into appropriate signal path. Is that not true?

Title: Re: Stability analysis of a continuous time CMFB
Post by Ken Kundert on Apr 20^th, 2017, 12:40am

Perhaps it is. It was my understanding that diffstbprobe only provided differential stability. My mistake. I'm afraid I never use the diffstbprobe. Cadence has encrypted it, so I am not really sure what is in there. And since they the last version they provided was inaccurate, I figure it is safer simply to role my own.

-Ken

Title: Re: Stability analysis of a continuous time CMFB
Post by Andrew Beckett on Apr 21^st, 2017, 1:44pm

Ken,

It's just back to back baluns (if my memory is correct) and then the choice on the ADE stb analysis from controls whether you are analysing the differential mode or common mode stability by picking the appropriate iprobe.

The earlier cmdmprobe (which wasn't encrypted) was inaccurate - you're right, but diffstbprobe should always have been OK, AFAIK.

Not entirely sure why we encrypted it - that seems more trouble than it's probably worth...

Regards,

Andrew.

Title: Re: Stability analysis of a continuous time CMFB
Post by Frank Wiedmann on Apr 25^th, 2017, 6:41am

Andrew Beckett wrote on Apr 21^st, 2017, 1:44pm:

It's just back to back baluns (if my memory is correct) and then the choice on the ADE stb analysis from controls whether you are analysing the differential mode or common mode stability by picking the appropriate iprobe.

As far as I can tell, your memory is correct (see https://community.cadence.com/cadence_technology_forums/f/33/t/27989).

Title: Re: Stability analysis of a continuous time CMFB
Post by Paulo on Jun 26^th, 2018, 10:12am

Hi polyam,

Did you solve your problem with stb simulations?

I am also working with continuous time sigma-delta modulators and faced similar issues.

Based on your simulation results, your OTA have a CMFB stability problem. I have faced it before and have no success with stb analysis. Thus, I performed a manual analysis to compensate the CMFB loop.

I believe that a DC path from OTA input and OTA output is required for a stb analysis (what is not the case of an active-RC integrator). But, I have never found a tutorial of stb analysis including the case of and active-RC integrator.