Pioneer did research around the time of the release of their A400 Amp - many moons ago - and concluded that they would not include a headphone socket that switched the speakers off, in the then usual way, because they found that in inclusion of a switch in the o/p of the amp audibly deteriorated the sound.
More recently George Stancheff has developed the very nice sounding "lightspeed" attenuator which uses light dependent resistors ( LDRs ) and thus can avoid any non soldered contacts in the input circuit. George believes that this is one of the most significant contributing factors to the excellent sound of this device.
So although it may be a useful research tool I would avoid have a switched contact in the amp's o/p - unless it was perhaps a mercury switch ( and this is just an untested idea of mine )
I'm really interested to hear how the choke sounds.
Compesation
I just reduced my lead compensation from 100R + 47pF to 100R + 22pF. Lag compensation remains unchanged at < 5pF.
This had a profound positive influence on the sound. More "foot tapping", fine detail, "air" and feeling openness and freedom. All is all the sound is now much more engaging and "live" than before.
There are three reasons why personally I have no concerns about some ringing with i/p filter removed, square waves applied, and load of 8R + 0.1uF.
1) The i/p filter is normally not removed.
2) The input signal will never have a leading edge as fast as signal generator square waves.
3) The capacitive element of my usual load is way lower than 0.1uF
For me, this test is a useful tool to make sure the amp is not going to burst into oscillation under certain conditions ( as mine did with the lead compensation removed entirely ) but as long as this criteria is met I see no reason not to use one's ears as the judge of what degree of compensation is needed.
scope images coming soon
mike
Last edited:
Mike,
You have followed Linsley Hood's idea that phase lead is best; lag compensation should be avoided if at all possible.
I've found in practice that a combination works best. Seems you have too?
I agree about ringing. It's a nice to have, but not essential if a LP filter is fitted to the front end and you are not driving Quads.
Cheers,
Hugh
You have followed Linsley Hood's idea that phase lead is best; lag compensation should be avoided if at all possible.
I've found in practice that a combination works best. Seems you have too?
I agree about ringing. It's a nice to have, but not essential if a LP filter is fitted to the front end and you are not driving Quads.
Cheers,
Hugh
Just wanted to clarify one thing:
I do not want to give the impression that I don't mind if my amp o/p "rings". I would definetly prefer that it does not ring.
Ringing is a spurious HF waveform that has no harmonic relation to the signal and as such clearly come under the category of "noise" which for me is the No: 1 enemy of good sound.
However, I do not really mind if there is some ringing under specific test conditions that are very different from the normal operating conditions.
For me it will be much more interesting to leave the i/p filter and normal leads & speaker load in place and then apply square waves ( in very small doses so as to not fry the tweeters ) and see what real life conditions look like.
I do not want to give the impression that I don't mind if my amp o/p "rings". I would definetly prefer that it does not ring.
Ringing is a spurious HF waveform that has no harmonic relation to the signal and as such clearly come under the category of "noise" which for me is the No: 1 enemy of good sound.
However, I do not really mind if there is some ringing under specific test conditions that are very different from the normal operating conditions.
For me it will be much more interesting to leave the i/p filter and normal leads & speaker load in place and then apply square waves ( in very small doses so as to not fry the tweeters ) and see what real life conditions look like.
Last edited:
Yes Hugh,
The combination of both seems to enable one to use less compensation overall when compared with just using either one on it's own.
My default position on this at present is that if the amp is fundamentally unstable into pure R - as mine, is it's current configuration is - then some lag compensation is needed.
I have the impression that lag compensation is most effective against fundamental instability into pure R and lead compensation is most effective in dealing with difficult loads.
In practice of course the two combine to deal with both elements.
Hope this make some kind of sense !
cheers
mike
Mike,
It seems you have arrived at the same conclusion I have been stomping my feet about - lower compensation provides more enjoyment and it makes a very noticeable difference.
I believe removing it altogether is the ultimate.
As you say, you can't do this due to either the extra gain of your VAS fet or the use of more feedback, or both.
Now, this is one place where my little ZVP3310A with its relatively low gain might be a winner.
Can I suggest trying more phase lag compensation and no phase lead. Despite most saying that this will sound worse, I have actually found that phase lag (miller) compensation is actually the better sounding of the compensation methods for this amplifier. This is may be because it swamps the highly variable input capacitance of the mosfet??
I am itching to get home and try the choke output, believe me!
It seems you have arrived at the same conclusion I have been stomping my feet about - lower compensation provides more enjoyment and it makes a very noticeable difference.
I believe removing it altogether is the ultimate.
As you say, you can't do this due to either the extra gain of your VAS fet or the use of more feedback, or both.
Now, this is one place where my little ZVP3310A with its relatively low gain might be a winner.
Can I suggest trying more phase lag compensation and no phase lead. Despite most saying that this will sound worse, I have actually found that phase lag (miller) compensation is actually the better sounding of the compensation methods for this amplifier. This is may be because it swamps the highly variable input capacitance of the mosfet??
I am itching to get home and try the choke output, believe me!
I'm not sure yet if zero compensation is the ultimate goal or if there is just an ideal amount that hits a sweet spot for each design.
I'll try the combinations you suggest soon but right now my goal is to build an amp that can be used in the living room - so more fine tweaking will come after that.
I'll try the combinations you suggest soon but right now my goal is to build an amp that can be used in the living room - so more fine tweaking will come after that.
If you eschew compensation, then for proper Bode Nyquist stability you rely entirely on the internal parasitics in the semiconductors, principally the VAS of course, and any stray parasitics floating around in the layout.
In a bipolar, these parasitics are highly variable and of very low Q (ratio of impedance to resistance, a function of the charges swimming around in the depletion layer) and rather fuzzy. I've also found good quality (= silver mica) caps are an advantage in this role as well, probably because they are highly stable and offer very high Q.
The parasitics between gate and channel on a mosfet are rather better than between base and collector - because of the galvanic isolation I would think - and this could be a reason the mosfet VAS is superior sonically. But no matter what you do, the best sound seems to come as you approach minimum values, which is a bind.....
There is a lot written about this, but very little highlights the connection between good sound and compensation. It's absolutely critical, particularly wrt sibilance, which is often the best sign of undercompensation.
Cheers,
Hugh
In a bipolar, these parasitics are highly variable and of very low Q (ratio of impedance to resistance, a function of the charges swimming around in the depletion layer) and rather fuzzy. I've also found good quality (= silver mica) caps are an advantage in this role as well, probably because they are highly stable and offer very high Q.
The parasitics between gate and channel on a mosfet are rather better than between base and collector - because of the galvanic isolation I would think - and this could be a reason the mosfet VAS is superior sonically. But no matter what you do, the best sound seems to come as you approach minimum values, which is a bind.....
There is a lot written about this, but very little highlights the connection between good sound and compensation. It's absolutely critical, particularly wrt sibilance, which is often the best sign of undercompensation.
Cheers,
Hugh
Choke test
All,
I've spent the last two hours listening to the amp with and without an output choke.
The beauty of this mod is I can clip the choke in and out real time while the music is playing. I can do a pretty good a/b test with no time wasted inbetween.
The verdict? Absolutely awesome. The choke, to my ears, has no negative influence. Instead, it adds a little refinement and really sort of pulls the picture together. The mids especially feel like they're solid blocks and the separation is just amazing. Bass, mid and treble seem in better proportion. Treble is nice and clean.
It is perhaps is a little more laid back now, but in a very very good way, much like a valve amplifier.
The only tiny downside is the bass is maybe ever so slightly less punchy, but not enough for me to be worried.
I like it a lot and will be leaving it in place!
All,
I've spent the last two hours listening to the amp with and without an output choke.
The beauty of this mod is I can clip the choke in and out real time while the music is playing. I can do a pretty good a/b test with no time wasted inbetween.
The verdict? Absolutely awesome. The choke, to my ears, has no negative influence. Instead, it adds a little refinement and really sort of pulls the picture together. The mids especially feel like they're solid blocks and the separation is just amazing. Bass, mid and treble seem in better proportion. Treble is nice and clean.
It is perhaps is a little more laid back now, but in a very very good way, much like a valve amplifier.
The only tiny downside is the bass is maybe ever so slightly less punchy, but not enough for me to be worried.
I like it a lot and will be leaving it in place!
Put the full Thiele Network on the output, not just the Zobel half, nor just the inductor half.
Adjust the various component values for best performance from the amp. Then the Network should be inaudible. But the amp may sound better because it is now driving a predetermined HF load for which it is optimised.
All the Nchannel complements have higher transconductance.
One of the reasons I adopted the inverted version NFetZilla for my early experiments.
Adjust the various component values for best performance from the amp. Then the Network should be inaudible. But the amp may sound better because it is now driving a predetermined HF load for which it is optimised.
All the Nchannel complements have higher transconductance.
One of the reasons I adopted the inverted version NFetZilla for my early experiments.
Hugh,
I've no idea how many uH this choke is, but I suspect more than that. Did you see my post detailing the choke? It has 3-4 feet of wire on it!
Yes, I will get some nice magnet wire and wrap it around a suitable bobbin. Good thing you left room on the pcb for one huh? Wish I had on mine!
I'm starting to get really suspicious that perhaps these great changes in sound may be due to oscillation (not all, but some). Unfortunately it takes some sort of RF analyzer to really find out. I think I need to find an RF signal and phase meter, if there is such a thing. Hopefully it'll be at least as good a a spectrum analyzer? Better yet, a spectrum analyzer with phase measurement!
The main failure of my generator was because of a crowbar SCR across the +5V supply that was firing too soon. 5V rails are 5.02-5.04V, so it must be an error. It could not have been part of an overcurrent protect. However the current draw is 1.5A, and I hope that's normal for this circuit.
For now the generator seems to work, but unfortunately isn't very useful. I will put it back together and try to get usable output by clipping the output with diodes.
- keantoken
The main failure of my generator was because of a crowbar SCR across the +5V supply that was firing too soon. 5V rails are 5.02-5.04V, so it must be an error. It could not have been part of an overcurrent protect. However the current draw is 1.5A, and I hope that's normal for this circuit.
For now the generator seems to work, but unfortunately isn't very useful. I will put it back together and try to get usable output by clipping the output with diodes.
- keantoken
Kean,
Nice troubleshooting! Hope you get a working arrangement soon.
I'm reasonably sure I am not getting oscillation. I have a 100meg scope and any serious oscillation should show some sign. I have scoped it extensively.
The difference between the choked versus non choked output was not night and day, and perhaps was too optimistic in my description. But considering I was expecting it to be detrimental to the sound, having a sound that was as good or maybe even better was something to be very excited about. I also now have very good stability!
Nice troubleshooting! Hope you get a working arrangement soon.
I'm reasonably sure I am not getting oscillation. I have a 100meg scope and any serious oscillation should show some sign. I have scoped it extensively.
The difference between the choked versus non choked output was not night and day, and perhaps was too optimistic in my description. But considering I was expecting it to be detrimental to the sound, having a sound that was as good or maybe even better was something to be very excited about. I also now have very good stability!
Hmm,
I was hoping to be able to use a lower inductance on the choke and just dropped the wire length from over 3 feet to 50cm (20"). It had absolutely no effect on the overshoot at all! This thing needs far more than the usual amount. Fortunately the choking seems to have no negative affects sonically.
I was hoping to be able to use a lower inductance on the choke and just dropped the wire length from over 3 feet to 50cm (20"). It had absolutely no effect on the overshoot at all! This thing needs far more than the usual amount. Fortunately the choking seems to have no negative affects sonically.
Hi Andrew,
Yes, please tell us how to approach this or post some links to references.
Do you mean a choke & a snubber = Thiele network ?
Is there a method of calculating the correct values
thanks
mike
Thanks for these tips Hugh,
Do you think that silver mica caps are better than polystyrene ?
I have understood till now that polystyrene is the best available.
mike
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.