How would you know the phase margin with an oscilloscope? Or the gain margin?
On the other hand, as you wish 🙂
On the other hand, as you wish 🙂
Oh, I suppose I wouldn't then, I was going for the " At the very least you would need to use a scope and investigate the stability margins"
Is it likely that an amplifier with 30dB gain would be unstable?
Is it likely that an amplifier with 30dB gain would be unstable?
Last edited:
It could, or it couldn't, just knowing the CLG is not enough, you need to know the frequency where the CLG becomes 1, and what is the phase margin at that point.
Increasing or decreasing the CLG, changes that point of "1" closed loop gain, thus changing the phase and gain margin.
Increasing or decreasing the CLG, changes that point of "1" closed loop gain, thus changing the phase and gain margin.
Thanks, I need to do more reading. What other equipment would I need other than a function generator and oscilloscope to determine those quantities?
For a chip (that we don't have details for) then it would be quite acceptable imo to determine this empirically i.e. keep increasing the feedback factor until instability sets in and then reduce that level back down using judgement as to what is a 'safe' level. A scope and squarewave testing into various loads (some reactive) would give a final assurance that all was well.
Thanks Mooly, so from what I understand so far phase and gain margin can be detrimental to the high frequency stability due to phase shift effectively turning negative feedback positive and causing oscillation and high frequency distortion which would be visible on a square wave?
I could notice the amp has some stability problems even without testing or simulation because "the designer himself said so":
notice some details:
1) one way to get instability is to have too much internal phase shift (which may be unavoidable) coupled with too much NFB, which will turn positive feedback instead of negative at some frequency causing oscillation, or near enough to cause instability
2) the designer added a very small 15pF cap C809 in parallel with the main NFB resistor, 68k R811, with a crossover frequency of 150kHz , meaning NFB will start to increase only above that frequency
3) he is *starting* with low NFB, he´s increasing it only at a very high frequency and yet he had to "stop" it somewhat by adding an otherwise useless 1k resistor R809?
Meaning he couldn´t apply that 15 pF cap end to end?
Those three details together make me *guess* the designer has a small instability problem and he had to correct it.
No big deal because I suppose the amp works properly but I wouldn´t do what he clearly avoided,meaning increasing NFB by lowering R811 or increasing R813 .
Let me say this maybe politically incorrect statement: IF I had this amp on my bench and had to reduce gain for a paying customer for any reason, I woud test a few NFB combinations and re-compensate so gain at 150kHz does not change and I do not get in trouble caused by myself, but this is "a Forum answer" , so the *practical* answer is: "attenuate the input and do not mess with NFB" 🙂
By the way, I was not the only one suggesting a similar one.
By the way, I missed .001uF C805 in parallel with input 100k R837 (thanks Mooly), so now I suggest a new attenuator: add a series 22k resistor instead of one of the input black arrows 🙂 , replace 100k R837 by 10k and C805 by 100pF , which will start acting at 30 kHz, safe enough.
Attenuation will set new sensitivity around 400/450mV which shoud be acceptable.
If not enough, rise the added resistor from 22k to 47k, but it might be overkill.
notice some details:
1) one way to get instability is to have too much internal phase shift (which may be unavoidable) coupled with too much NFB, which will turn positive feedback instead of negative at some frequency causing oscillation, or near enough to cause instability
2) the designer added a very small 15pF cap C809 in parallel with the main NFB resistor, 68k R811, with a crossover frequency of 150kHz , meaning NFB will start to increase only above that frequency
3) he is *starting* with low NFB, he´s increasing it only at a very high frequency and yet he had to "stop" it somewhat by adding an otherwise useless 1k resistor R809?
Meaning he couldn´t apply that 15 pF cap end to end?
Those three details together make me *guess* the designer has a small instability problem and he had to correct it.
No big deal because I suppose the amp works properly but I wouldn´t do what he clearly avoided,meaning increasing NFB by lowering R811 or increasing R813 .
Let me say this maybe politically incorrect statement: IF I had this amp on my bench and had to reduce gain for a paying customer for any reason, I woud test a few NFB combinations and re-compensate so gain at 150kHz does not change and I do not get in trouble caused by myself, but this is "a Forum answer" , so the *practical* answer is: "attenuate the input and do not mess with NFB" 🙂
By the way, I was not the only one suggesting a similar one.
By the way, I missed .001uF C805 in parallel with input 100k R837 (thanks Mooly), so now I suggest a new attenuator: add a series 22k resistor instead of one of the input black arrows 🙂 , replace 100k R837 by 10k and C805 by 100pF , which will start acting at 30 kHz, safe enough.
Attenuation will set new sensitivity around 400/450mV which shoud be acceptable.
If not enough, rise the added resistor from 22k to 47k, but it might be overkill.
^ Thanks for the full and informative answer.
Yes, I understand it's best to advise to attenuate..........Like I said, if it was me I'd try and get rid of some of that gain 😉🙂
Thanks again, very interesting
Yes, I understand it's best to advise to attenuate..........Like I said, if it was me I'd try and get rid of some of that gain 😉🙂
Thanks again, very interesting
Thus reducind the noninv imput impedance to less than 10k ( about 8,7 at dc ), meanwhile the inv imput impedance is 68k ( 69 in fact ). Even though the original already had some imput impedance inbalance, i think increasing it this much is not a smart ideea, you increase the chance of a larger dc offset, and it simply looks bad ( at least to me ).
Also i am not sure that reducind C805 to 1 tenth of it's original vallue, could be safe, i am not sure why the original is that large, but it might play some role in the stability issue, so i think it should stay the same.
I do not know if anyone has noticed but i already gave some suggestion of what can be done, that way you do not affect in any way the amplifier, and you can still choose whatever divider values you want. It does involves some work, a DIP8 chip, a small power source, but i think it's worth it. If needed i can draw a schematic, but it's so simple that i shouldn't need to 🙂
All the best.
Yes, you could certainly be sure the final result was stable in the real world with such testing.
Also (and no one ever seems to mention this) when designing amplifiers, the feedback ratio is almost always determined by looking at the maximum output voltage required and the input voltage available... and why wouldn't you do that ?
Problem with that approach is that it doesn't guarantee optimal sound quality in my experience. All amplifiers are different, there are no hard and fast rules.
If you want to measure the phase response and stability margins accurately then you need a good scope and signal generator and the ability to alter the already constructed circuit to derive its open and closed loop gains together with a plot of phase shift of the signal passing through the amp. Only then can you theoretically determine whether its going to be stable or whether you need to alter things (which with an STK you may not be able to).
So the quick and easy way is to do it empirically.
- Status
- Not open for further replies.