Designer’s Guide Community

HdrChopper

Community Fellow

Offline

Posts: 493

Weird stability analysis results in a bandgap circuit
Jul 14^th, 2009, 6:00am

All,

I'm analyzing the stability of a band-gap reference circuit in a BICMOS process. The gain of loop circuit is well known and
equal to the product of [GM(Q0)/gm(M0)] and [gm(M1)/gm(Q1)]; where GM(Q0) is the composite transconductance of Q0 and the 2K resistor.
When running the stability analysis with a probe located in A I get the expected low freq gain of ~ -8.2dB, as shown in the figure.
However when I move the probe to B the DC gain and the high frequency response both change drastically.
As far as I know the location of the probe for running stability analysis is not important as long as I place it in a single loop that
at least can include local feedback loops, like this is the case (beta helper of Q1 is another feedback loop).

Any hints or clues about what is going on? Anything I might be missing?
Your comments are more than welcome.

Regards
Tosei

Back to top

BG.jpg

Keep it simple

IP Logged

HdrChopper Community Fellow Offline Posts: 493	Re: Weird stability analysis results in a bandgap circuit Reply #1 - Aug 2^nd, 2009, 11:11am Nobody can help on this one?
Back to top	Keep it simple IP Logged

wave Senior Member Offline Posts: 117 Silicon Valley	Re: Weird stability analysis results in a bandgap circuit Reply #2 - Aug 3^rd, 2009, 4:21pm HdrChopper wrote on Aug 2^nd, 2009, 11:11am: Nobody can help on this one? I have a hard time believing they are the same loop. However, I can't quite convince myself why. None the less, if the simulator indicates they are different, that is very likely the reason. Wave
Back to top	IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: Weird stability analysis results in a bandgap circuit
Reply #3 - Aug 4^th, 2009, 5:49pm

wave wrote on Aug 3^rd, 2009, 4:21pm:

I have a hard time believing they are the *same* loop.
However, I can't quite convince myself why.

Hi Wave,

Thanks for your reply. Why aren�t you convinced the loop is not the same?
Both source A and B are in "series" with the signal.
In any case if you believe they are not the same, which one would you choose for analyzing stability for the band-gap? and why? (you already know which one is correct. What I cannot get is why the other one is not.)

Thanks
Tosei

Back to top

Keep it simple

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: Weird stability analysis results in a bandgap circuit Reply #4 - Aug 5^th, 2009, 9:24pm Hi Tosei, My simulation result is different from yours. Both A and B proves the same loop gain magnitude and phase response. The low-frequency loop gain is -8.37dB. By the way, it seems that the 100 kOhm resistor can be removed. That has nothing effect on the function. Best Regards, Yawei
Back to top	tb_CurCellStb_CaseAB.JPG IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Weird stability analysis results in a bandgap circuit Reply #5 - Aug 6^th, 2009, 7:44am hi, @Hi Tosei, i simulated the ckt what you have given in hspice ...it is giving more or less same results in both cases @Yawei, how come you got same without 100k, i feel because of this resistor npn transistors are getting biased, correct me if am wrong Thanks, Rajasekhar.
Back to top	IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: Weird stability analysis results in a bandgap circuit
Reply #6 - Aug 6^th, 2009, 7:12pm

Hi Yawei / Rajasekhar,

Thanks both for your inputs. Based on what you got I should assume there must be something wrong with my simulation set up, I guess?

Rajasekhar, I understand both cases should give exactly the same result, which is what Yawei got and I was expecting. However you said "more or less" the same result. What is the difference you observed?

Thanks and Regards
Tosei

Back to top

Keep it simple

IP Logged

raja.cedt

Senior Fellow

Offline

Posts: 1516
Germany

Re: Weird stability analysis results in a bandgap circuit
Reply #7 - Aug 6^th, 2009, 9:17pm

hi tosei,
i did this simulation in hspise...so i used to put LC filter at the point where i want break...so what i feel is while doing this some loading we have to take care (generally i keep total ckt wherever i want to adjust loading )..but this will not give much difference for this ckt....i got minute difference in the magnitude and no difference in the phase plot. In fact this expected because at A both sides are high impedance nodes..but at B one side is low impedance ........Let me know if any issue and thanks a lot because i learned a lot through this discussion

thanks,
Rajasekhar.

Back to top

IP Logged

ywguo

Community Fellow

Offline

Posts: 943
Shanghai, PRC

Re: Weird stability analysis results in a bandgap circuit
Reply #8 - Aug 9^th, 2009, 1:58am

raja.cedt wrote on Aug 6^th, 2009, 9:17pm:

hi tosei,
� � i did this simulation in hspise...so i used to put LC filter at the point where i want break...so what i feel is while doing this �some loading we have to take care �(generally i keep total ckt wherever i want to adjust loading )..but this will not give much difference �for this ckt....i got minute difference in the magnitude and no difference in the phase plot. In fact this expected because at A both sides are high impedance nodes..but at B one side is low impedance ........Let me know if any issue and thanks a lot because i learned a lot through this discussion

thanks,
Rajasekhar.

Hi Rajasekhar, you are right. So spectre has a convenient way, stb analysis, to check the loop gain/phase response without breaking the loop. It adopts the method named return ratio in fact. Please search IEEE or Paul Gray's textbook for return ratio.

Best regards,
Yawei

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: Weird stability analysis results in a bandgap circuit
Reply #9 - Aug 9^th, 2009, 3:50pm

Hi Rajasekhar,

Actually the metod Yawei pointed out is the one I used for analyzing the stability of this circuit. My point is that according to this methodology theory the port which is used for computing the return ratio could be placed at any point in the loop. No loading effects has to be considered since the loop is not broken (therefore the loading conditions are intact).
However I got different results when placed at A or B.

Regards
Tosei

Back to top

Keep it simple

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: Weird stability analysis results in a bandgap circuit
Reply #10 - Aug 10^th, 2009, 12:52am

ywguo wrote on Aug 9^th, 2009, 1:58am:

Hi Rajasekhar, you are right. So spectre has a convenient way, stb analysis, to check the loop gain/phase response without breaking the loop. It adopts the method named return ratio in fact. Please search IEEE or Paul Gray's textbook for return ratio.

Best regards,
Yawei

You can find the simulation method used by the stb analysis in http://www.kenkundert.com/docs/cd2001-01.pdf. Some more information about loop gain simulation is available on my webpage http://sites.google.com/site/frankwiedmann/loopgain.

Regarding the differences between Tosei's and Yawei's results: It might be helpful if you would post your netlists so that you can see if there are any differences between your simulation setups.

Back to top

IP Logged

nobody Community Member Offline Posts: 75	Re: Weird stability analysis results in a bandgap circuit Reply #11 - Aug 10^th, 2009, 3:55am Frank proposes a good way to solve this problem. I believe Yawei simulated that ckt using Spectre and it is complicated to do loop gain simulation by using Hspice.
Back to top	IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: Weird stability analysis results in a bandgap circuit Reply #12 - Aug 10^th, 2009, 6:56am hi, i am using hspice in office and may be i can use middle broke method in LT spice...if any body using hspice can you explain how you to do stability analysis in h spice......for my office work also i am using like LC method only Thanks, Rajasekhar.
Back to top	IP Logged