Re: Problem with Spice - it won't allow me to change bounds of graph



"Michael Noone" <mnoone.uiuc.edu@xxxxxxxxx> wrote in message
news:Xns962A9A7DFA2BAmnooneuiucedu127001@xxxxxxxxxxxxxxxxx
> "Larry Brasfield" <donotspam_larry_brasfield@xxxxxxxxxxx> wrote in
> news:VAZ2e.37$oI2.968@xxxxxxxxxxxxxxx:

[step vs. ramp cut]

>> I observed similar behavior. The circuit appears to
>> have DC convergence problems in the transient
>> simulation and AC simulation, so I've attributed
>> that strange starting behavior to that issue without
>> studying it closely. I trust the simulator's result
>> once the transient simulation settles at a believable
>> operating point. I take that as occuring after the
>> strange, unrequested initial flailing ceases.
>
> I already have all the components needed (well, different FETs, but
> similar enough that I think they will work) sitting around the lab
> anyways, so I'm going to see how it performs in the real world. If I get
> similar results, is there any way to modify this circuit so as to
> improve response time?

Here are the changes I would make:
1. Eliminate or greatly reduce R2.
2. Degenerate the M1,R1 gain stage to get a more
stable and predictable gain versus operating point.
A source resistor would be a good addition for this.
3. Set the zero established by C1,R5 to be near the
lowest frequency pole created in the M1,R1,M2 stage.
4. Adjust the M1,R1 gain for best loop response.
(Note this likely interacts with step 3, so it will could
be a little tedious.)

> I looked back at my original post from two weeks
> ago and couldn't find anything along those lines, though it's possible
> that I missed something due to the very large size of that topic (over
> 150 posts! eek)

Perhaps lost in that maelstrom was a simple circuit I posted
with about 8 uS response time. You should hope to get 1 mS,
perhaps with revision of your current configuration once the
above 4 changes are properly applied.

--
--Larry Brasfield
email: donotspam_larry_brasfield@xxxxxxxxxxx
Above views may belong only to me.


.