The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> stb and ac analysis do not match?
https://designers-guide.org/forum/YaBB.pl?num=1352294950

Message started by xianweng on Nov 7^th, 2012, 5:28am

Title: stb and ac analysis do not match?
Post by xianweng on Nov 7^th, 2012, 5:28am

Hi,Guys,I have some problem to understand why stb and ac do not match.When I run stb analyis in a loop,the phase starts from -180 and it increses to a positive value and gain curve is normal! When I run ac analysis ,it seems like that phase and gain curve are normal .Could someone tell me which one simulation is right?

Thanks any help in advacne

Title: Re: stb and ac analysis do not match?
Post by tzg6sa on Nov 7^th, 2012, 6:13am

I don't see how do you break the loop in the AC case. It seems to me that you forces one node to VAC=1, but I don't see at which point do you check the result.

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 7^th, 2012, 7:17am

Besides the fact that your drawing is very hard to read (what is the part at the right side from C=1F ?), I assume that your ac analysis is run open loop, correct? In this case, the result shows that the dc feedback is positive (phase zero for s=0). A circuit with positive dc feedback cannot work because of an unstable bias point. Thus, something is wrong.

Title: Re: stb and ac analysis do not match?
Post by wave on Nov 7^th, 2012, 2:00pm

A couple comments.

First, I would expect STB and AC to be different by definition - the nature of their injections.
Research Michael Tian's paper (with Ken Kundert), Middlebrook, Paul Hurst, etc.

Second, I have seen cases where the DC starting point of phase with STB seems "off" by 180deg, and you basically look at the change in Phase. -- with Integrators in particular.
I've never seen a full explanation of this.

;)
Wave

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 8^th, 2012, 12:18am

Xianweng, in your circuit diagrams the signal source is shown.
Question: Are you aware that it must be switched-off (Vin=0) during loop gain analysis?
As already mentioned, at dc you need negative feedback (phase=-180 deg).
But also the results of STB analysis are weird. The gain drops with rising frequency - and the phase function rises?
This is NOT correct.
Thus: Check the bias point for normal operation and don`t forget to ground the opamp input during open-loop analysis.

wave wrote on Nov 7^th, 2012, 2:00pm:

First, I would expect STB and AC to be different by definition - the nature of their injections.

No, I don`t think so. As far as I know, the only difference is that the STB analysis also includes bi-directional signal travelling.
Thus, both should give identical results for uni-directional applications (as in our case).

Title: Re: stb and ac analysis do not match?
Post by Ken Kundert on Nov 8^th, 2012, 12:24am

No, they are not the same. The AC analysis approach you describe does not give accurate results and should never be used.

-Ken

Title: Re: stb and ac analysis do not match?
Post by Frank Wiedmann on Nov 8^th, 2012, 6:14am

wave wrote on Nov 7^th, 2012, 2:00pm:

Second, I have seen cases where the DC starting point of phase with STB seems "off" by 180deg, and you basically look at the change in Phase. -- with Integrators in particular.
I've never seen a full explanation of this.

;)
Wave

You might want to take a look at http://www.designers-guide.org/Forum/YaBB.pl?num=1294178255 (and maybe http://www.designers-guide.org/Forum/YaBB.pl?num=1124688329).

Title: Re: stb and ac analysis do not match?
Post by xianweng on Nov 8^th, 2012, 6:16am

Horror Vacui wrote on Nov 7^th, 2012, 6:13am:

I don't see how do you break the loop in the AC case. It seems to me that you forces one node to VAC=1, but I don't see at which point do you check the result.

The right half is ac analysis
1.I break the loop with a big resistor(100G) and a big capacitor(1F)
2.I check the reulst at point A

Title: Re: stb and ac analysis do not match?
Post by xianweng on Nov 8^th, 2012, 6:20am

buddypoor wrote on Nov 7^th, 2012, 7:17am:

1.the right side capacitor is 1F
2.the right side mos is pmos ,so it is negative feedback

Title: Re: stb and ac analysis do not match?
Post by xianweng on Nov 8^th, 2012, 6:32am

buddypoor wrote on Nov 8^th, 2012, 12:18am:

First, I would expect STB and AC to be different by definition - the nature of their injections.

1.Vin is not zero ,as I drawed ,it is constant DC voltage ,1.86,to provide bias
2.yes ,the results of STB analysis are weird. The gain drops with rising frequency and the phase function rises
3.the bias point is normal.I do stb and dc analysis simultaneous,the dc result show that the positive input of op and the negative input of op are equal .It means bias point is normal

Title: Re: stb and ac analysis do not match?
Post by xianweng on Nov 8^th, 2012, 6:42am

Ken Kundert wrote on Nov 8^th, 2012, 12:24am:

No, they are not the same. The AC analysis approach you describe does not give accurate results and should never be used.

-Ken

I know the AC analysis approach does not give accurate reults ,but it could not give a wrong answer. And stb is more accurate.The ac analysis approach I used is describe in the book:designing_analog_chips.I attached the picture that the book used

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 8^th, 2012, 7:29am

Ken Kundert wrote on Nov 8^th, 2012, 12:24am:

No, they are not the same. The AC analysis approach you describe does not give accurate results and should never be used.
-Ken

Hi Ken,

I am not quite sure to whom your reply is adressed and which ac analysis you are referring to.
Nevertheless, since more than 20 years I have used such an ac analysis approach (the correct one, not the L-C method !) without any problems - knowing that bi-directional signal movement is not covered.
Therefore, some problems on my side regarding your statement "should never be used".
I would rather say: Should be used only if the application allows its use (uni-directional propagation, restauration of dc bias point and loading).

I kindly ask you to comment on this.
Thank you.

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 8^th, 2012, 7:32am

xianweng wrote on Nov 8^th, 2012, 6:16am:

1.I break the loop with a big resistor(100G) and a big capacitor(1F)

This resistor does not enable a correct operating point. Instead you must use a big inductor allowing dc feedback.

Title: Re: stb and ac analysis do not match?
Post by Ken Kundert on Nov 8^th, 2012, 9:23pm

Buddypoor,
I don't know what you are referring to when you say "the correct one". Does it involve 1F capacitors?

-Ken

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 9^th, 2012, 1:20am

Ken Kundert wrote on Nov 8^th, 2012, 9:23pm:

Buddypoor,
I don't know what you are referring to when you say "the correct one". Does it involve 1F capacitors?
-Ken

OK agreed, I should have been more specific.
As mentioned somewhat later in my former posting - I consider a loop gain simulation as "correct" under the following pre-conditions:
*restauration of the correct dc bias point
*proper loading at the breakpoint (assuming uni-directional signal propagation).

I think, these conditions are fulfilled by Middlebrook´s double-injection technique.
However, for many opamp applications I have learned that an opening at the opamp output and injection of a series ac signal (without correction using a current source in an second step) gives results, which are exact enough.

Let me define "exact enough":
Knowing that

* each amplifier simulation model never can reflect real conditions
* in reality each passive part has tolerances
* each hardware realization suffers from parasitic influences

I think, the procedure to find the loop gain for a circuit with feedback may allow for a small systematic error - if this error remains within the limits set by the aforementioned uncertainties (other error sources).
As an engineer I follow the rule: A design as well as a measurement procedure must not be as exact as possible but as exact as necessary
(but I know that in some cases it is not easy to specify what really is "necessary").

I hope now you understand why I ask for justification of your statement "ac analysis approach ...should never be used".
Thank you and regards.
B.

Title: Re: stb and ac analysis do not match?
Post by Frank Wiedmann on Nov 9^th, 2012, 1:57am

The method that injects only a series ac voltage simulates what I like to call the "voltage loop gain". The conditions under which this is a good approximation for the "real" loop gain are given in http://www.omicron-lab.com/fileadmin/assets/customer_examples/Bode_Info_LoopGain_V1_0.pdf.

I compare the results of the different loop gain simulation methods at the end of http://groups.yahoo.com/group/Design-Oriented_Analysis_D-OA/message/41. Additional information about loop gain simulation is available on my webpage https://sites.google.com/site/frankwiedmann/loopgain (where you can also find my definition of the Y parameters used in the comparison, see https://sites.google.com/site/frankwiedmann/loopgain#TOC-What-about-other-definitions-for-loop-gain-).

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 9^th, 2012, 3:16am

Frank, thank you for clarification (and for the 1st referenced document, which I didn`t know yet).
Of course, I agree with everything and - if I understood you and the linked documents right - you are supporting my doubts regarding the strict and general statement that an "ac analysis approach should never be used" for loop gain determination, right?
Thank you.
B.

Title: Re: stb and ac analysis do not match?
Post by Frank Wiedmann on Nov 9^th, 2012, 4:32am

In my opinion, simulation of voltage loop gain can be adequate in many situations, but you should know what you are doing. The other methods presented on my webpage have the advantage that they always give the same result, no matter where in the loop you put the probe.

For Spectre users, I see no reason to use ac analysis to simulate voltage loop gain instead of using stb analysis (which corresponds to Tian's Method as described on my webpage), because the results of stb analysis are more accurate and it is at least as easy to set up as ac analysis.

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 9^th, 2012, 4:57am

Fine - thank you.
Your statements - and , in particular, your sentence "...you should know what you are doing" - are in full agreement with my way of "engineering thinking".

Title: Re: stb and ac analysis do not match?
Post by Ken Kundert on Nov 9^th, 2012, 11:00pm

Buddypoor,
I'm with you on the 'correct' method, but not on the 'exact enough'. In my experience attempting to open the loop always changes the loading effects, which changes the measured loop gain at the higher frequencies (near the unity gain frequency), making the results largely useless.

-Ken

Title: Re: stb and ac analysis do not match?
Post by buddypoor on Nov 10^th, 2012, 1:55am

Hi Ken,

I am afraid, I have expressed myself not clear enough.
I think the precondition "I should know what I am doing and why/if it is allowed to do" covers everything.
And - as I have said - I can accept an error if it "remains within the limits set by the aforementioned uncertainties (other error sources)."

To be specific, I am afraid that in many cases the error introduced by the two-pole opamp model (neglecting higher poles as well as parasitic influences) can be in the same order or even larger if compared with the error introduced by the voltage-loop gain method. But, of course, it depends on the particular case and cannot be generalized. Therefore, I have to know what I am doing....
B.

Title: Re: stb and ac analysis do not match?
Post by Frank Wiedmann on Nov 10^th, 2012, 4:23am

Ken Kundert wrote on Nov 9^th, 2012, 11:00pm:

The setup for voltage loop gain simulation does not open the loop. You put the ac voltage source in the same place where you put the loop gain probe for an stb analysis. I agree that opening the loop with large inductors and large capacitors is generally a bad idea.