Little GEM

Status
Not open for further replies.
Jump to Latest
#41
Waly said:
If we accept this definition of "unconditionally stable" (an amp stable at any closed loop gain), then a "conditionally stable amp" like yours will always have some pulse response overshoot (such a "conditionally stable" amp is necessary an at least second degree underdamped system - see the phase response with a dip before ULG, this phase dip is what creates instabilities at e.g. low CLG). This overshoot can be tamed using an input filter, but then it is not realistic to expect, in simulation or practice, perfect step response without one.
Click to expand...

That is what I thing also. Do you accept my previous square wave test, with input filter connected, correct?
dado
 
#42
Waly said:
If we accept this definition of "unconditionally stable" (an amp stable at any closed loop gain), then a "conditionally stable amp" like yours will always have some pulse response overshoot (such a "conditionally stable" amp is necessary an at least second degree underdamped system - see the phase response with a dip before ULG, this phase dip is what creates instabilities at e.g. low CLG). This overshoot can be tamed using an input filter, but then it is not realistic to expect, in simulation or practice, perfect step response without one.
Click to expand...

If you knew the answer , why ask the question.??..🙄

Unconditional stability as noticed by Dadod is not strictly necessary
for amps provided they are set at high enough CLG , yet , it is better
to have the highest possible phase margin as capacitive loading will
not only increase the phase dip abut also , consequently , induce
a local gain peak , hence reducing the gain margin.
 
#43
wahab said:
If you knew the answer , why ask the question.??..🙄

Unconditional stability as noticed by Dadod is not strictly necessary
for amps provided they are set at high enough CLG , yet , it is better
to have the highest possible phase margin as capacitive loading will
not only increase the phase dip abut also , consequently , induce
a local gain peak , hence reducing the gain margin.
Click to expand...

Yes, that is well said. I like to hear different opinions and different suggestion. The pudding is in details.
dad
 
#44
wahab said:
In principle such a computation is made without the input high frequency
attenuation capacitor and without output LR filter, this to check if the amp
is unconditionaly stable.

Indeed , most amps rely on theses filters to be stable as conditionnaly
stables amps allow to reduce THD by the virtue of greater HF loop gain.
Click to expand...

wahab said:
If you knew the answer , why ask the question.??..🙄
Click to expand...

Because, as you claimed above, I disagree that the input filter has anything to do with the amp stability, conditionally or not.
 
#45
dadod said:
That is what I thing also. Do you accept my previous square wave test, with input filter connected, correct?
dado
Click to expand...

The pulse step response, considered as a consequence of the gain-phase characteristic result, is not an independent determination. It is mathematically related to the complex CLG transfer function, e.g. by simply mapping the Fourier transform H(jw) to the Laplace transform H(s) through jw->s and calculating the H(s)/s reponse.

Now, as I said, there could be some large signal implications of the pulse step response (like, as a notable example, slewing effects). Such effects are generally not very well modelled in Spice, as they usually may involve non linear saturation effects in active devices and other non-linear effects that may be poorly (or not at all) included in the device models. The same non-linear effects may induce transient instabilities, but again, these have little to nothing to do with the linear gain-phase responses determined using ac analysis in Spice. To avoid such large signal instabilities, indeed an input filter could be effective, but I still see this as a poor man's stabilizing method.
 
#46
Here is 10kHz square wave simulation without input filter and with and without output filter.
The zobel is changed now, I use 2R7 and 100nF now. Output inductance is 1uH parallel to 2R7 resistor.
dado
 

Attachments

  • square10k-no-input-or-output-filter.jpg
    square10k-no-input-or-output-filter.jpg
    269.8 KB · Views: 399
  • square10k-no-input-or-output-filter-5nF.jpg
    square10k-no-input-or-output-filter-5nF.jpg
    254.4 KB · Views: 392
  • square10k-no-input-or-output-filter-10nF.jpg
    square10k-no-input-or-output-filter-10nF.jpg
    259.4 KB · Views: 363
  • square10k-no-input-filter-10nF.jpg
    square10k-no-input-filter-10nF.jpg
    249.5 KB · Views: 356
  • square10k-no-input-filter-100nF.jpg
    square10k-no-input-filter-100nF.jpg
    259.5 KB · Views: 332
  • square10k-no-input-filter-1uF.jpg
    square10k-no-input-filter-1uF.jpg
    252.5 KB · Views: 161
#48
AndrewT said:
Have you adjusted the slew (risetime) of the test sq?
I don't understand you captions.
Some plots say no output filter, but I see the Thiele Network always connected.
Click to expand...

I use 20 nsec now.
By output filter I mean an inductance parellel with a resistor. Zobel should always be connected(this is my opinion) and I shorted the inductance for those tests.
dado
 
#51
dadod said:
I use 20 nsec now.
By output filter I mean an inductance parellel with a resistor. Zobel should always be connected(this is my opinion) and I shorted the inductance for those tests.
dado
Click to expand...
I can't see the short across the L.
Tap the output before the L or after the L.
I would only investigate by tapping before the L, I cannot recall advice to investigate a particular operation by tapping output after the L.
 
Status
Not open for further replies.
Top Bottom