[SOLVED] simulating the closed loop stability of the common mode amplifier loop

[Junus2012]

Dear friends,

How can I simulate the stability common mode feedback loop in fully differential amplifier.

I need to measure the PM, GBW, and the closed loop gain of the CMFB loop.

You can use the attached image for your explanation please,

my idea is to apply an ac component to the VCM terminal and run the ac simulation to measure the transfer function from Vo1 or Vo2 then from there I get the needed parameters,

Please correct me if I am wrong or if there is better method


Best Regards

 

[vivekroy]

I would use the test bench as shown in the attached image.
Note that in common mode, X and Y move together and no current flows through the resistors and therefore the resistors do not come in the common mode circuit. They however load the circuit in the differential mode of operation.
'dcfeed' is a component present in analogLib of Cadence which is a short for small signal simulations. Similarly, 'dcblock' is another component which is an open for small signal simulations. These are used to set the proper DC operating points for the transistors.

 

[Junus2012]

I am sorry but this method is not clear how it works

 

[FvM]

PM and GBW, use Middlebrook's method
closed loop gain, apply stimulus to Vcm input

 

[Junus2012]

Dear Fvm

Thank you for your reply, I am not familiar with this method, also I tried to find some expalanation about it from internet but I didn't understand it, can you please suggest me different method,

what about if I connect the two inputs of the fully differential amplifier to the common mode voltage.

under the open loop gain of the differential amplifier i apply an ac signal to the VCM of the CMFB amplifier then I check the response from the output to this point,, what you think about it

 

[vivekroy]

Make sure your devices see the same DC bias as the devices in the original amplifier would. Breaking any closed loop system means that that the DC operating points will go haywire. This is why I used the elaborate scheme to make sure that all the devices see the same DC operating points.
The VCVS are used to bias the devices properly. The load that the CMFB amplifier sees also needs to be proper else its pole will not be at the correct point which can change the phase margin of the common mode loop.

 

[frankrose]

I assume you have Cadence. The easiest method is to use diffstbprobe from analogLib. You can simulate common and differential mode stability too with it.
Setup (page 32-33): https://www.lumerink.com/courses/ece515/Handouts/Fully-Differential Opamps.pdf
Analysis setup is on page 42-43.

 

[Junus2012]

Dear frankrose, thank you very much for your kind help,

With unknown reason my professor needs me to implement the simulation by using traditional analyses, once I showed him the stability simulation setup he didn't want this methods,

Therefore I need the method using like ac or transient analyses. I found the method from Martin book and I attached it below please, I didn't understand the part when he said I must break the part of Vcntrl like he is not including the CMFB amplifier in the test, kindly have a look on it please

 

[FvM]

Breaking the loop is the usual method to measure loop gain, but it's important not to corrupt the bias point. Middlebrooks method has the advantage of keeping the loop virtually closed. You can refer to the loopgain example provided in LTspice educational folder. It's particularly simple implemented in a circuit branch with high load impedance like Vcms-Vcmc where Middlebrook loop gain reduces to pure voltage gain V(x)/V, lower part of the example.

 

[sutapanaki]

Junus2012,

To get the stability of your CMFB you'll have to break the loop and not put stimuli at VCM as you initially suggested. If you don't work with Cadence, then your best bet is to use the Middlebrook's method as FvM suggested. You have to keep your main amplifier biased well in the condition it will normally work.
In the picture from Johns/Martin book if you break the loop at Vcntrl you break the CMFB loop and you can test its stability by injecting a signal in the loop. Just as in any other negative feedback configuration.
Here is a word of caution. If your main amplifier is fully differential and has two stages and you have a single CMFB looking at the common mode of the output and feeding back to the 1st stage of the main amplifier, then you have to be extra careful because the main amplifier may latch in common-mode.

 

[Junus2012]

Thank you FvM for your explanation,

when you tell about the "Breaking the loop is the usual method to measure loop gain'' is like I am simulating the loop gain without including the CMFB amplifier (also cal it common sense amplifier) as we break the loop from vcntrl,, Such kind of simulation I have done from the DC analyses by sweeping the Vcntrl (or the Vcms or Vcmc) and recording the output two voltages under the open loop condition of the main differential amplifier while the two inputs at the common mode voltage, after deriving the output voltage I found the open loop gain is sufficiently high that I could use the diode connected CMFB amplifier. and as you noted I was certain to include the required biasing voltage (Vcntrl) in my sweep range.

However, as I am interested now with the stability of the closed CMFB loop including the common sense amplifier, I would go for your suggested method of Middlebrooks method, I will do it like this way

I will connect an ac source with zero DC voltage in front of the common sense amplifier, ( at the Vcntrl point), then I will run the AC simulation to find the transfer function between the two points before and after the AC source (Vx/Vy). I will keep the main amplifier under open loop condition with inputs connected at the common mode voltage.

Please correct me if I am wrong,
Thank you very much once again

- - - Updated - - -

- - - Updated - - -

- - - Updated - - -

Junus2012,

To get the stability of your CMFB you'll have to break the loop and not put stimuli at VCM as you initially suggested. If you don't work with Cadence, then your best bet is to use the Middlebrook's method as FvM suggested. You have to keep your main amplifier biased well in the condition it will normally work.
In the picture from Johns/Martin book if you break the loop at Vcntrl you break the CMFB loop and you can test its stability by injecting a signal in the loop. Just as in any other negative feedback configuration.
Here is a word of caution. If your main amplifier is fully differential and has two stages and you have a single CMFB looking at the common mode of the output and feeding back to the 1st stage of the main amplifier, then you have to be extra careful because the main amplifier may latch in common-mode.

Thank you Suta for your reply,

I appreciate your help very much,

yes it is important to monitor my amplifier is biased correctly under this test

- - - Updated - - -

Dear my Mentors,

After reading your comments many times I now realized why it is NOT right setup to apply my signal at VCM, this is the reason you are telling me to inject the signal in the loop,

then I went back to the gray book and snapped this part below,

according to the picture and your simulation setup explanation, I am going to measure the acms*(-acmc)

While if I apply the signal at VCM then it basically mean I am measuring the ACMFB according the image below, which should always give a value of one if the acms*(-acmc) is high enough



Thank you guys very much

 

[sutapanaki]

See, CMFB loops don't really differ much from any other negative feedback loop. A voltage regulator from feedback point of view is very much similar to what the CMFB is doing. So, everything you know already about negative feedback, its stability and criteria for stability, how to test for it, it is all applicable for CMFB too.