This topic was already discussed on this thread: http://www.diyaudio.com/forums/solid-state/190040-how-do-you-test-your-amp-stability-oscillation.html
But I am wondering if it is really sufficient to just connect a speaker to the amp and to check for oscillations - what I have already done. In the Elrad magazine where the 550W Mosfet amp was published in 1985, the author tested his amp with many different capacities as I remember.
My objective is to proof that my stripboard design of the honey badger is stable in any case (with all speakers you may think of). Link to this thread: http://www.diyaudio.com/forums/solid-state/294664-diyab-amp-honey-badger-line-grid-board.html
Could somebody advice me which test I should perform to be on the safe side?
But I am wondering if it is really sufficient to just connect a speaker to the amp and to check for oscillations - what I have already done. In the Elrad magazine where the 550W Mosfet amp was published in 1985, the author tested his amp with many different capacities as I remember.
My objective is to proof that my stripboard design of the honey badger is stable in any case (with all speakers you may think of). Link to this thread: http://www.diyaudio.com/forums/solid-state/294664-diyab-amp-honey-badger-line-grid-board.html
Could somebody advice me which test I should perform to be on the safe side?
It certainly wouldn't hurt to have some 1n, 10n, 22n, 47n and maybe 100n caps floating around for a start.
i just read the mentioned article again and the author was using a switchable capacitor box of the E6 series from 2.2nF to 6.8uF and a 1kHz toneburst which is synchronized with the peak of the sine wave to have also square waves in the signal.
I will now build such a capacitor box and think how to get a cheap toneburst generator.
I will now build such a capacitor box and think how to get a cheap toneburst generator.
The old and gone danish hifi magazine "High Fidelity" used to put in 10kHz squarewave and load the amp with 8ohms and 1 uF.
Attached is a picture of my latest creation measured without output coil (which will be added).
With 8 ohm/1uF and a outputcoil, a small resonanse that quickly dies may be visible on some amps (not here)
Besides this, you also need to apply a 1-10kHz sinus and check at the collector/drain of all the transistors in the amp, to make sure that they do not osciallate themselves.
If they do you must add compensation-caps in the circuit.
Attached is a picture of my latest creation measured without output coil (which will be added).
With 8 ohm/1uF and a outputcoil, a small resonanse that quickly dies may be visible on some amps (not here)
Besides this, you also need to apply a 1-10kHz sinus and check at the collector/drain of all the transistors in the amp, to make sure that they do not osciallate themselves.
If they do you must add compensation-caps in the circuit.
Attachments
Last edited:
I think it depends on what the amp is designed to drive. I wouldn't be putting 8R // 1uF across the output of a headphone amp, for example, as this is an extremely long way from what the amp is likely to see in service.
For my 100W amp I did a fair bit of testing into 4R // 470nF. I think halving the expected load resistance is a good start, and adding capacitances in the hundreds of nF for a typical 100W amp. Obviously you wouldn't load a lower power amp this hard.
For my 100W amp I did a fair bit of testing into 4R // 470nF. I think halving the expected load resistance is a good start, and adding capacitances in the hundreds of nF for a typical 100W amp. Obviously you wouldn't load a lower power amp this hard.
Cload @ feedback loop unity gain intercept ~ few MHz is what counts
for global feedback stability, local output Q RF oscillations are possible to 10s of MHz
normally the relatively high Q @ MHz C load on a audio PA is the cable wire pair C of of 12-50 pF/ft
so actually 100s of pF to few nF are probably the "worst case"
and of course most choose to decouple the output with a few uH L||R and provide damping with an addition "Zobel"
ringing on the speaker side of this isolation Z is not related to amp stability at all, despite the prominence given in very old audio rags
unless
if you think uF values are justified by ESL panels driven from step up Xfmer then you should look at the Xfmer series R, leakage inductance - usually these parasitics cause the Z to rise above 50 kHz and in no way looks like uF at typical audio PA MHz negative feedback gain intercept
for global feedback stability, local output Q RF oscillations are possible to 10s of MHz
normally the relatively high Q @ MHz C load on a audio PA is the cable wire pair C of of 12-50 pF/ft
so actually 100s of pF to few nF are probably the "worst case"
and of course most choose to decouple the output with a few uH L||R and provide damping with an addition "Zobel"
ringing on the speaker side of this isolation Z is not related to amp stability at all, despite the prominence given in very old audio rags
unless
include piezo ceramic bender tweeters without any series R (which wouldn't happen in a XO, they will always have series R)
if you think uF values are justified by ESL panels driven from step up Xfmer then you should look at the Xfmer series R, leakage inductance - usually these parasitics cause the Z to rise above 50 kHz and in no way looks like uF at typical audio PA MHz negative feedback gain intercept
This was my baby😉
At that time I had a capacitor box with a 12pos switch that I connected to the output while applying 1kHz square signal.
Just one possibility that was at hand at that time.
Today I would measure phase margin according to the well known standard procedure.
The cheapest toneburst generator is a file created by audacity played on your phone/laptop/mp3player😉
I just relocated this article stretching over ELRAD 2/3/4 1985.
Have a look at the position of compensating cap C9: It is NOT connected to the output - avoiding phase lag introduced by the source follower power-stage, specially when loaded by a cap.
This simple trick helped to achieve a very stable amp under all conditions.
Last edited:
None of your sound sources delivers any significant MHz content - because it is real audio, most of the time delivered by some bandwidth limited audio codec.😉
I remember the afore mentionened 10kHz square test well - stability under these conditions was nice to have - but none the less academic.
Last edited:
the OP's question is Stability Testing, which simply can't be done reliably with frequency limited audio codecs as Test signal source
maybe you're missing some EE assumptions:
a simple "eyeball" stability test for some feedback amplifiers is to look at the square wave/step response overshoot and ringing - with a sqauare wave that has frequency components up to and beyond the amplifier closed loop bandwidth
to apply this to audio PA you may need to bypass input filters, should really be looking at the output stage response "inside" any load isolating series Z
this is not "academic" - if all the conditions for its validity are met the step response does give a practical indication of amplifier stability margin
maybe you're missing some EE assumptions:
a simple "eyeball" stability test for some feedback amplifiers is to look at the square wave/step response overshoot and ringing - with a sqauare wave that has frequency components up to and beyond the amplifier closed loop bandwidth
to apply this to audio PA you may need to bypass input filters, should really be looking at the output stage response "inside" any load isolating series Z
this is not "academic" - if all the conditions for its validity are met the step response does give a practical indication of amplifier stability margin
Last edited:
Maybe I miss some EE assumption, certainly you can provide a useful link to enlighten me.
I can imagine an amp with a local instability of some 10 or even 100MHz - but in case of such a rare occurence I was not sure this will be excited by the input signal, which is normally slowed down not only by input filtering but along the signal path as well.
Certainly the chance to excite such instability is better by applying a sharp transient to the input.
I can imagine an amp with a local instability of some 10 or even 100MHz - but in case of such a rare occurence I was not sure this will be excited by the input signal, which is normally slowed down not only by input filtering but along the signal path as well.
Certainly the chance to excite such instability is better by applying a sharp transient to the input.
- Status
- Not open for further replies.