Designer’s Guide Community

dandelion

Community Member

Offline

Posts: 98

A strange phenominon on the open loop analysis
Aug 23^rd, 2007, 2:50am

Hi，
I have a puzzle on the loop gain analysis.

Pls. see the attached files which is still about the loop gain analysis. The amplifier A in the diagram is the simple single stage diff pair stracture.From the file we can see that the different breaking loop point leads to different siumlation results. As my understanding, the loop gain analysis with different breaking point in the loop should have the same results if the loading effect can be avoided.

Would anyone pls. shed some light on it?

Thanks a lot!

Back to top

Stability_issues.pdf

IP Logged

Frank Wiedmann

Community Fellow

Offline

Posts: 678
Munich, Germany

Re: A strange phenominon on the open loop analysis
Reply #1 - Aug 23^rd, 2007, 4:58am

Your method preserves the dc operating point of the circuit, but it strongly changes the small-signal ac impedances at the breaking point. This explains the different behavior you are seeing. If you are using Spectre, I would strongly suggest that you use the stb analysis for simulating loop gain. The algorithm used by this analysis does not change the impedances so that it really does not matter where you insert the probe into the loop. The algorithm used by the stb analysis is described in http://www.kenkundert.com/docs/cd2001-01.pdf.

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: A strange phenominon on the open loop analysis
Reply #2 - Aug 23^rd, 2007, 7:14am

I agree 100% with Frank. Here is another piece of info to further clarify this thinG:

By breaking the loop or by using the stb analysis (as Frank suggested) you are actually measuring the loop's return ratio. Formally that is not the same as the loop gain A*f. The loop gain is defined as the gain (in the forward amplifier) times the feedback gain. This loop gain DOES NOT include feedback effects in your forward amplifier and forward effects in your feedback network, which in reality they exist.
Return ratio analysis (either by breaking the loop or by using stb analysis) does consider such effects.

Hope this helps
tosei

Back to top

Keep it simple

IP Logged

mahno

New Member

Offline

Posts: 2

Re: A strange phenominon on the open loop analysis
Reply #3 - Aug 23^rd, 2007, 8:11am

Hello,

Here is a nice paper which shows how to correctly break the loop.

The first method you use is O.K. (very close to the real thing as you would see from this paper)

The second one is wrong.

First the compensation cap should be left in its place from gate to drain of the fet.

the big L and C associated with the ac test source should be killed

place the a source in series with the gate of the fet and with the output of the opamp.

let's say the + of this source will be on the gate and the - on the output of the opamp.

the dc value of tis ac test source should be 0

Now the output to be monitored will be the output of the opamp with the positive reference being the output to ground (for the reference of the ac source given above)

good luck

Back to top

img_5151247_0001.pdf

IP Logged

ywguo Community Fellow Offline Posts: 943 Shanghai, PRC	Re: A strange phenominon on the open loop analysis Reply #4 - Sep 2^nd, 2007, 2:00am Hi Mahno, I read your attachment. Would you please explain the weird things happened to Ti and Tv in figure 8-11 and 8-12? Thanks Yawei
Back to top	IP Logged

Frank Wiedmann Community Fellow Offline Posts: 678 Munich, Germany	Re: A strange phenominon on the open loop analysis Reply #5 - Sep 2^nd, 2007, 11:22pm The reason are the impedances seen from the point where you insert the test sources. Please see the formulas at the end of http://groups.yahoo.com/group/Design-Oriented_Analysis_D-OA/message/41 (to understand what they mean, you will probably need to read that thread from the beginning).
Back to top	IP Logged

Monkeybad

Junior Member

Offline

Posts: 31

Re: A strange phenominon on the open loop analysis
Reply #6 - Sep 2^nd, 2007, 11:42pm

Basically I think this two method will get the same result, because it represent the loop-gain.
The difference between method 1 and method 2 is that the method 2 its unit-gain BW seems faster.
When you break the loop to measure the return ratio, you need to consider the output loading effect.
In method 2, add some cap in the OP to slow down the unit agin BW maybe it will get the similar PM in method 1.
This added cap is the output loading effect seen from the OP in returned signal.
This just my thought to explain this phenomenon.
I hope it help to find the correct reason.

BEST REGARD

Back to top

IP Logged