The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Design >> Analog Design >> bandgap stability issue https://designers-guide.org/forum/YaBB.pl?num=1222077588 Message started by trashbox on Sep 22nd, 2008, 2:59am |
Title: bandgap stability issue Post by trashbox on Sep 22nd, 2008, 2:59am Hi all, Refer to the bandgap in attachment, I break the loop at 2 and find the frequency respondence is strange: phase decreases so fast and even cross 0 degree. How to explain it in theory? How do you analyze the bandgap top stability issue? I break at 1 instead of 2, the result is resonable. Thanks very much! Trashbox |
Title: Re: bandgap stability issue Post by Frank Wiedmann on Sep 22nd, 2008, 5:29am How do you break the loop? Are you using stb analysis or a similar method? See the current discussion at http://www.designers-guide.org/Forum/YaBB.pl?num=1220974790 for possible problems of traditional methods for opening the loop. |
Title: Re: bandgap stability issue Post by thechopper on Sep 22nd, 2008, 6:50am Hi, Be careful! ...you have to loops (one is positive feedback, the other one is negative feedback: in order to be stable neg feedback gain must be larger than positive one) here with one node in common. In order for the feedback theory in multi-loop systems to be valid you need to analyze the loop by breaking it in such common node (one break open all the loops at the same time). Therefore breakpoint 2 should be the correct one. Break point 1 only accounts for one feedback loop. Having said this, it is not recommendable to break the loop since ac loads at the break point are never replicated accurately when the loop is open and therefore might change the overall stability analysis. As Frank suggested you should use stb analysis or a similar one in order to not affect the actual loop. Tosei |
Title: Re: bandgap stability issue Post by Tlaloc on Sep 22nd, 2008, 7:46am In order to see a frequency response that is similar to what you are used to, you should break the loop at 1. However, as Tosei pointed out, you have two feedback paths; one positive, and negative. Breaking at 1 verifies the negative feedback path, but you should also verify the positive feedback path by breaking the loop at the negative terminal of the amp. Then you have to show that the negative feedback loop gain is greater than the positive feedback loop gain at all frequencies. If that condition doesn't hold, you have an oscillator. Those two AC analyzes should be run at all PVT to verify stability. Of course, the other way is just to use stb at node 2. |
Title: Re: bandgap stability issue Post by trashbox on Sep 22nd, 2008, 11:41pm Hi Tlaloc, thechopper and Frank Wiedmann! Thanks for your comments. Also I read Frank's blog and it's very good! I have no spectre and run into problems when using eldo .lstb command. As Tlaloc said, I checked two feedback path (break amplifier's positive and negative input at the same time): phase margin, gain difference. The results are good. Hope it woks. Thanks again. Trashbox |
Title: Re: bandgap stability issue Post by Frank Wiedmann on Sep 23rd, 2008, 1:50am What kind of problems do you have with Eldo's .lstb analysis? Please note that unlike for Spectre's stb analysis, the orientation of the zero voltage source (loop gain probe) is important for Eldo's .lstb analysis. Have you tried connecting the zero voltage source in the opposite direction? |
Title: Re: bandgap stability issue Post by trashbox on Sep 23rd, 2008, 7:16pm Frank Wiedmann wrote on Sep 23rd, 2008, 1:50am:
Hi Frank! I set up a simple test circuit as attachment. Then write stimulus according eldo manual: vstb v1 v2 dc 0 ac 1 .lstb vstb .op .ac dec 100 0.1 1g .plot ac lstb_db .plot ac lstb_p However I found the dc operation point is not correct.My eldo version is ELDO v6.9_1.1. Furthermore, I changed the vstb's orientation into: vstb v2 v1 dc 0 ac 1, this problem still exists. Would you have any suggestion? Thanks! |
Title: Re: bandgap stability issue Post by Frank Wiedmann on Sep 23rd, 2008, 11:57pm I do not use Eldo, so I cannot help you much here. A voltage source with dc 0 should not change the operating point. Do you get the correct operating point when you connect V1 and V2 directly with a wire? I noticed that the example posted by gafsos at http://www.edaboard.com/ftopic265486.html has ac 0 instead of ac 1 for the voltage source. This does not change the operating point, however, and is probably not the cause of your problem. |
Title: Re: bandgap stability issue Post by holddreams on Sep 24th, 2008, 5:51am Hi, all, I have a question here: Instead of breaking the loop at point 2, how will it be when just breaking the loop at point 1 and the negative input of the opamp at the same time? Is the frequency-phase the same? Thanks. |
Title: Re: bandgap stability issue Post by trashbox on Sep 24th, 2008, 7:20pm holddreams wrote on Sep 24th, 2008, 5:51am:
Hi holddreams, If you break the loop at point 1 and the negative input of opamp at the same time, this open-loop is a simple two stage sytem if you use one-stage opamp such as one stage folded-cascode. The dominant pole is opamp's output because its large output impedance.Please refer to the attachment. Yes, just like other guys said, in this case, the ac output impedance is not the same situation with close-loop because in real application, the feedback will affect the output impedance. However I think, even output ac impedance is affected by feedback, Rout=Ro/(1+A0), here R0 is open loop output impedance, A0 is open loop gain, the dominant pole will not change and the second pole is pushed farther so that PM will be better. Right? |
Title: Re: bandgap stability issue Post by trashbox on Sep 24th, 2008, 7:28pm Hi Frank! Thanks, I also noticed this post. But this post run ac and tran at the same time, I'm afraid this guy does not check it carefully. By the way, your blog at geocities is really good! Trashbox Frank Wiedmann wrote on Sep 23rd, 2008, 11:57pm:
|
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |