Mike,
That's interesting!
Greg found that with small LC of just 33pF, which we specified on the BOM in fact, that 47R of VAS degen was best.
Your solution is markedly different - fortunately the board can accommodate it, there is provision for phase lead, lag comp, and resistive bypass - but it certainly shows more sonic improvements are possible.
Good news..... I favour a larger output choke, say to 5uH which is a walk in the park for any audio amp, and will be trying this today.
Cheers,
Hugh
That's interesting!
Greg found that with small LC of just 33pF, which we specified on the BOM in fact, that 47R of VAS degen was best.
Your solution is markedly different - fortunately the board can accommodate it, there is provision for phase lead, lag comp, and resistive bypass - but it certainly shows more sonic improvements are possible.
Good news..... I favour a larger output choke, say to 5uH which is a walk in the park for any audio amp, and will be trying this today.
Cheers,
Hugh
That is the understanding of output coils - the stability they offer is a load function and a case of "one size fits all" amplifiers, within the usual meaning here of the term "unconditional stability". Practically, it just means that amplifiers won't then ring noticeably when driving any realistic load. This is nothing to do with amplifier stability.
There are lots of strange comments on forum about other uses for the coil, like HF bandwidth limiting, RF suppression or the notion that it should damp amplifier oscillation but it would seem absolutely the wrong place to be adding such a damper in the hope it will stifle problems upstream and defined within the feedback loop.
It is also said to have some affect on motor feedback interaction with the loudspeaker but, given the the very high frequency and damped roll-off here, one would immediately think of using higher Q measures at low level to control any identifiable problems there.
For unconditional stability, miller compensation is usually not enough. Sometimes you must go sky high with miller compensation before the amplifier is unconditionally stable, and sometimes it never works at all. This is why the output coil is needed most of the time; going without it is a gamble unless you plan to use the amp with a load you know it is stable into.
If the amp is stable normally, and only has problems with reactive loads, then it is appropriate to use an output L/R, because after a point adding miller compensation usually won't help it only, just make it slower and not any less prone to oscillating into the right load.
That is what I think, anyway.
- keantoken
If the amp is stable normally, and only has problems with reactive loads, then it is appropriate to use an output L/R, because after a point adding miller compensation usually won't help it only, just make it slower and not any less prone to oscillating into the right load.
That is what I think, anyway.
- keantoken
My version of this amp with CLG of only 11 ! is already unconditionally stable with just a smallest amount of compensation. 2pF lag + 8pF lead.
With this arrangement and with an i/p filter giving -3dB @ about 140khz, adding 100nF across an 8R load has no visible effect on the scope with 10khz square waves and adding 1uF instead just adds some ringing.
So I do not have to add a coil out of necessity for stability, rather I am experimenting with with various combinations of compensation purely to optimise the subjective sound.
I agree that adding moderate amounts of millar capacitance does not really help reduce ringing with capacitive loads but it does improve the sound of the amp.
I am surprised that you are concerned with oscillation with this design because Greg and I have found that it is extraordinarily stable already.
Last edited:
Actually Kean, I just wanted to add that if we do add a coil to an amp output I do think it is very important for sound quality to damp it sufficiently to stop any tendency for ringing.
For me ringing in the speaker leads is bad news as the ringing frequency will have absolutely no harmonic relation with the signal and thus is just another spurious source of noise with will add audible confusion to the final sound.
For me ringing in the speaker leads is bad news as the ringing frequency will have absolutely no harmonic relation with the signal and thus is just another spurious source of noise with will add audible confusion to the final sound.
Hugh,
I just rechecked in spice the mods I recommended but this time using 47R degen value.
Without the 500R in series with the miller cap and with 47R degen spice showed the 180 degree frequency at around 7.8 meg hz
With the 500R and with no degeneration the 180 degree frequency increased to 21 meg hz
Now, we know that these figures are not realistic because spice is way too optimistic in this regard but never-the-less I think the relative difference is significant even if the absolute values are not
I just rechecked in spice the mods I recommended but this time using 47R degen value.
Without the 500R in series with the miller cap and with 47R degen spice showed the 180 degree frequency at around 7.8 meg hz
With the 500R and with no degeneration the 180 degree frequency increased to 21 meg hz
Now, we know that these figures are not realistic because spice is way too optimistic in this regard but never-the-less I think the relative difference is significant even if the absolute values are not
Mike, how's it look with with 50r of degeneration and no compensation whatsoever (my setup).
Guys,
This is an intrinsically stable amp due to the low loop gain and the mosfet VAS.
Most of the effort here is to tidy up square waves and ringing behaviour after undue capacitive provocation.
Can I suggest that we be guided by the sound quality as we consider the simulations, for the models are imperfect, particularly with fets. It is very difficult to take this course, as it is not exactly PC, but the amp is for listening to music and our consumers are ears, not AP2s...
I'm enjoying the sound as I type, this is a very good amplifier as it stands, 47R VAS degen, 33pF lag comp, 400mA quiescent and 5uH of output inductor with 10R damping resistor.
Cheers,
Hugh
This is an intrinsically stable amp due to the low loop gain and the mosfet VAS.
Most of the effort here is to tidy up square waves and ringing behaviour after undue capacitive provocation.
Can I suggest that we be guided by the sound quality as we consider the simulations, for the models are imperfect, particularly with fets. It is very difficult to take this course, as it is not exactly PC, but the amp is for listening to music and our consumers are ears, not AP2s...
I'm enjoying the sound as I type, this is a very good amplifier as it stands, 47R VAS degen, 33pF lag comp, 400mA quiescent and 5uH of output inductor with 10R damping resistor.
Cheers,
Hugh
Last edited:
Good advice Hugh! If building this amplifier has taught me one thing it's that subjectivism rules!
This entire build has shown me that there are no rules that are correct all of the time and for all amplifiers, and I thank you for encouraging me to trust my ears. The Bjt driven version I have been working on is especially terrible sounding if it is left in the low distortion incarnation suggested by spice. A bit of human meddling got it sounding much better.
Now I challenge you, leave the output choke arrangement in place and remove that miller cap. Tell me what your ears think! You may not like it, but I'd love to hear your thoughts.
If you do this you will have a good idea of what I am currently listening to, though I am using the lower gain zvp3310a, which may cope a bit better with no compensation.
Hopefully you won't get stability issues if you try this. I have tried it with the 2110 in my circuit and had no problems, so you should be ok.
This entire build has shown me that there are no rules that are correct all of the time and for all amplifiers, and I thank you for encouraging me to trust my ears. The Bjt driven version I have been working on is especially terrible sounding if it is left in the low distortion incarnation suggested by spice. A bit of human meddling got it sounding much better.
Now I challenge you, leave the output choke arrangement in place and remove that miller cap. Tell me what your ears think! You may not like it, but I'd love to hear your thoughts.
If you do this you will have a good idea of what I am currently listening to, though I am using the lower gain zvp3310a, which may cope a bit better with no compensation.
Hopefully you won't get stability issues if you try this. I have tried it with the 2110 in my circuit and had no problems, so you should be ok.
Last edited:
Actually Hugh, given that you a using ever so slightly less degen (I am using 51.1R), a higher gain fet and perhaps a slightly more perfect CCS, you may end up needing some lag comp after all, though maybe much less than 33pf.
But I would certainly have a go at removing your lag comp completely before you ultimately make that decision. For interest's sake.
But I would certainly have a go at removing your lag comp completely before you ultimately make that decision. For interest's sake.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.