How to add the loadings when breaking the feedback loop?

[Ryan1234]

Dear all:
Now I want to use Hspice to analyze the stability of a feedback loop .
Then, in a feedback loop, if I break a node A into B and C, how should I add the loadings on node B and node C ?
Thank your replies.

 

[LvW]

Dear all:
Now I want to use Hspice to analyze the stability of a feedback loop .
Then, in a feedback loop, if I break a node A into B and C, how should I add the loadings on node B and node C ?
Thank your replies.

* As node C will be your input for loop gain simulation there is nothing to add.
* If you know the load, connect it to node B. Or use a replica of the path which you have disconnected. However, don´t forget that the dc loop has also been broken with the consequence of probable bias problems.
In this case you must use the method described by Middlebrook (combination of two separate simulations with a voltage resp. a current source).

 

[Ryan1234]

* As node C will be your input for loop gain simulation there is nothing to add.
* If you know the load, connect it to node B. Or use a replica of the path which you have disconnected. However, don´t forget that the dc loop has also been broken with the consequence of probable bias problems.
In this case you must use the method described by Middlebrook (combination of two separate simulations with a voltage resp. a current source).[/quote]

Thank for your answer. I still have two questions here.
1. You say "If you know the load, connect it to node B."
@ How could I know the value of the load?
@ Is the load the input impedance looking into node C ?
@ How could I measure it?
2. Is the replica of path another accurate solution ?

Thank you very much!!

 

[LvW]

1. You say "If you know the load, connect it to node B."
@ How could I know the value of the load?

If the feedback path is rather simple, you can see the load by visual inspection, don´t you?

@ Is the load the input impedance looking into node C ?

Yes, of course! That´s the load disconnected by breakink the loop.

@ How could I measure it?

This makes no sense. Either you derive it by simple inspection (I suppose you know your circuit diagram) or it is a complicated/involved function of frequency.

2. Is the replica of path another accurate solution ?

Yes it is accurate - as far as the load itself is concerned. However, watch the bias condition of the closed loop!

To be on the safe side, use the Middlebrook method which is described elsewhere (google/internet, or also in this forum).

 

[Ryan1234]

Depending on my research during these days,
here I sum up the method to break the feedback loop for analysis of loop gain:

1. If the ac source is voltage, then breaks the high impedance node;
__If the ac source is current, then break the low impedance node.
2. Use the low pass filter to break the node. ex: (Large L) or (R, dc=0 ac=1T)
3. It does not need to consider adding the loads on the breaking nodes. Just break the node, and that will be fine.

If I am wrong, please correct me.
If I am right, please feedback to me.
Thank all you guys. I love you~

 

[LvW]

Hi Ryan1234,

you are not "wrong", however what you suggest works only in case you really have a node which sees no (or only a negligible small) load change by opening the loop. This happens not very often.
What do you think, why some more complicated procedures have been elaborated for this purpose?