class A w/ varying power levels output kinda defeats the point of class A, doesn't it?
Personally, if you're too lazy to actually turn the amp off, I would use a 'sleep mode' that, after a certain time period of zero source power, would keep the PS on, but disconnect the amp from it. Then, when it detects source power again, it would re-power-up the amp section. The caps stay charged up, everything's working except for the actual amplifier section.
Personally, if you're too lazy to actually turn the amp off, I would use a 'sleep mode' that, after a certain time period of zero source power, would keep the PS on, but disconnect the amp from it. Then, when it detects source power again, it would re-power-up the amp section. The caps stay charged up, everything's working except for the actual amplifier section.
Re: sliding bias control chip
the interesting thing about this chip is that it tries to maintain a 20mv voltage drop on the source resistors. I remember someone mentioned a while ago that the optimal "bias" for a class B amp is one that drops about 20mv on the emitter (bjt) or source (mosfet) resistor.
I guess the LT chip confirms that.
unfornately, I don't know why anyone would use such a complicated device to set-up bias in an otherwise very simple amp.
smoking-amp said:
the interesting thing about this chip is that it tries to maintain a 20mv voltage drop on the source resistors. I remember someone mentioned a while ago that the optimal "bias" for a class B amp is one that drops about 20mv on the emitter (bjt) or source (mosfet) resistor.
I guess the LT chip confirms that.
unfornately, I don't know why anyone would use such a complicated device to set-up bias in an otherwise very simple amp.
Nelson's comments are certainly true of the "before" times.
That is before microprocessor chips and measuring devices
capable of pulling off real time bias adjustment became
available.
You can with some of the new allegromicro hall effect sensors,
some sample and hold circuits and a few a/d converters
do complete real time monitoring of the real bias, and real
output power and then calculate and adjust the bias.
Then you can do peak power detection and sliding bias
all in firmware.
Not trivial to do so. Not so easy to keep the cpu noise out
of the power amp either. Isolation amplifier to drive the
bias recommended.
The krell amps use an inline current transformer and a small
microprocessor to set the bias.
That is before microprocessor chips and measuring devices
capable of pulling off real time bias adjustment became
available.
You can with some of the new allegromicro hall effect sensors,
some sample and hold circuits and a few a/d converters
do complete real time monitoring of the real bias, and real
output power and then calculate and adjust the bias.
Then you can do peak power detection and sliding bias
all in firmware.
Not trivial to do so. Not so easy to keep the cpu noise out
of the power amp either. Isolation amplifier to drive the
bias recommended.
The krell amps use an inline current transformer and a small
microprocessor to set the bias.
There are more subtle problems afoot than the speed with which the bias can be set.
Turn on a class A amplifier and watch the bias settle in. It takes quite some time before the output devices stabilize due to the heating of the devices themselves, the thermal mass of the heatsinks, caps forming, etc. Easily twenty to thirty minutes, and things are still shifting a bit after an hour. All these factors have a non-trivial effect on the behavior of the output devices, hence the sound quality.
And folks expect a dynamic bias circuit to sound like class A? The bias may or may not be able to track, but the devices/heatsinks/etc. scarcely have time to react to the changes.
Now, perhaps if there were some way to keep the devices warmed up, and the heatsinks' thermal mass at equilibrium, and so forth...perhaps then the circuit might stand a chance of sounding like it was class A.
There is.
Dispense with the varying bias and set a nice, constant, fixed high bias. That way you'll be guaranteed that the circuit will be ready when a transient comes along. On paper, bias tracking arrangements look good. Unfortunately, in the real world they don't accomplish what they set out to do. I'm not saying that they don't work--in the sense that music comes out--nor am I aware of any intrinsic reliability problems--but they just don't sound as good as a hot class A amp.
I heard the big Krell (600W?) sliding bias amp a year or two ago out in Santa Fe. Granted it was an unfamiliar system, etc. etc. etc. but I was rather underwhelmed by the sound quality. Not that it was bad, mind you, but for that much money I could easily have bought two or three much better sounding amps and had change left over. Dan D'Agostino would be better served (...I'm sure he's breathlessly waiting for my take on this...) by going back to his earlier design philosophy and building up-to-date KSA-50s or KMA-100s.
Grey
Turn on a class A amplifier and watch the bias settle in. It takes quite some time before the output devices stabilize due to the heating of the devices themselves, the thermal mass of the heatsinks, caps forming, etc. Easily twenty to thirty minutes, and things are still shifting a bit after an hour. All these factors have a non-trivial effect on the behavior of the output devices, hence the sound quality.
And folks expect a dynamic bias circuit to sound like class A? The bias may or may not be able to track, but the devices/heatsinks/etc. scarcely have time to react to the changes.
Now, perhaps if there were some way to keep the devices warmed up, and the heatsinks' thermal mass at equilibrium, and so forth...perhaps then the circuit might stand a chance of sounding like it was class A.
There is.
Dispense with the varying bias and set a nice, constant, fixed high bias. That way you'll be guaranteed that the circuit will be ready when a transient comes along. On paper, bias tracking arrangements look good. Unfortunately, in the real world they don't accomplish what they set out to do. I'm not saying that they don't work--in the sense that music comes out--nor am I aware of any intrinsic reliability problems--but they just don't sound as good as a hot class A amp.
I heard the big Krell (600W?) sliding bias amp a year or two ago out in Santa Fe. Granted it was an unfamiliar system, etc. etc. etc. but I was rather underwhelmed by the sound quality. Not that it was bad, mind you, but for that much money I could easily have bought two or three much better sounding amps and had change left over. Dan D'Agostino would be better served (...I'm sure he's breathlessly waiting for my take on this...) by going back to his earlier design philosophy and building up-to-date KSA-50s or KMA-100s.
Grey
Unfortunately no matter how smart your quiescent current controller is you have to remember that when you change the quiescent current you also change output stage transconductance, something that is not a problem in a fixed bias arrangement.
A class-a/b amplifier sacrifices linearity for efficiency but is simple.
A sliding-bias class-a amplifier also sacrifices linearity for efficiency only with considerably more complexity.
An amplifier that can be manually switched between fixed bias class a/b and full class-a offers the best of both modes without the compromise.
A class-a/b amplifier sacrifices linearity for efficiency but is simple.
A sliding-bias class-a amplifier also sacrifices linearity for efficiency only with considerably more complexity.
An amplifier that can be manually switched between fixed bias class a/b and full class-a offers the best of both modes without the compromise.
High efficiency class A
So it seems that a sliding bias class A (you could even consider a non-switching class AB amplifier an extreme case of this) is not the ultimate solution.
But of course you can have a floating very low voltage power supply that delivers a big current for the class A amplifier output stage and have a class AB power stage driven by the same signal that drives the floating power supply. This could even be better than a normal class A since the voltage across the output transistors remains virtually constant, which is good for linearity. Still you get the benefit of a higher efficiency of the class AB amplifier. You only get some additional dissipation from the class A stage but that is relatively small because of the low supply voltage.
This principle has been used by (I think) Technics (New Class A) in the past. A nice example can be found here (from Shinichi Kamijo):
http://www.ne.jp/asahi/evo/amp/J554K2955/index.htm
Steven
So it seems that a sliding bias class A (you could even consider a non-switching class AB amplifier an extreme case of this) is not the ultimate solution.
But of course you can have a floating very low voltage power supply that delivers a big current for the class A amplifier output stage and have a class AB power stage driven by the same signal that drives the floating power supply. This could even be better than a normal class A since the voltage across the output transistors remains virtually constant, which is good for linearity. Still you get the benefit of a higher efficiency of the class AB amplifier. You only get some additional dissipation from the class A stage but that is relatively small because of the low supply voltage.
This principle has been used by (I think) Technics (New Class A) in the past. A nice example can be found here (from Shinichi Kamijo):
http://www.ne.jp/asahi/evo/amp/J554K2955/index.htm
Steven
Personally, I think the QUAD current dumping principle is a more elegant way of combining a low-power class A amplifier and a high power amplifier. At least it doesn't require floating power supplies or anything like that.
I have no personal experience with sliding bias class A, but I guess that until someone invents an electronic version of the crystal ball, it will be very difficult to prevent any sliding bias class A system from temporarily leaving the class A regime when a large and sudden signal peak with a low rise time occurs.
The LT1166 is basically a class AB control loop (non-linear common-mode loop used for class AB control). With class AB control loops, you can make very nice non-switching class AB amplifiers that do not suffer from thermal tracking problems immediately after a change in volume (this is always a problem to some extent in conventional class AB amplifiers). The only thing I don't understand is why LT chose to use a product rule. A harmonic mean rule is much nicer, as it keeps the minimum output device current at half the quiescent current, rather than tending to zero when the current through the other device tends to infinity.
I have no personal experience with sliding bias class A, but I guess that until someone invents an electronic version of the crystal ball, it will be very difficult to prevent any sliding bias class A system from temporarily leaving the class A regime when a large and sudden signal peak with a low rise time occurs.
The LT1166 is basically a class AB control loop (non-linear common-mode loop used for class AB control). With class AB control loops, you can make very nice non-switching class AB amplifiers that do not suffer from thermal tracking problems immediately after a change in volume (this is always a problem to some extent in conventional class AB amplifiers). The only thing I don't understand is why LT chose to use a product rule. A harmonic mean rule is much nicer, as it keeps the minimum output device current at half the quiescent current, rather than tending to zero when the current through the other device tends to infinity.
Nelson, have you got a patent on sliding bias class A or on non-linear common-mode loops used for instantaneous class AB control (or on both)? In the second (or third) case, can you give me a reference to the patent? The oldest article I've ever seen about class AB control loops is the Huijsing and Tol paper published in the Journal of Solid-State Circuits in 1976. I'd be very interested to read older material on this subject, if any.
I've never used the LT1166, but I have used class AB control loops made with a transistor array and some discrete transistors in an audio power amplifier. That works very nicely.
I've never used the LT1166, but I have used class AB control loops made with a transistor array and some discrete transistors in an audio power amplifier. That works very nicely.
My patent 3,995,228 was submitted September 1975, and as
far as I can see describes a circuit which would do either, but
addresses Class A specifically. The basic circuit is similar to,
but predates the work of Cordell and separately, Hawksford.
They both obviously intended Class AB operation.
As an aside, I developed an extremely interesting circuit in
this area just as I was walking out the door at Threshold.
It remains a secret.
far as I can see describes a circuit which would do either, but
addresses Class A specifically. The basic circuit is similar to,
but predates the work of Cordell and separately, Hawksford.
They both obviously intended Class AB operation.
As an aside, I developed an extremely interesting circuit in
this area just as I was walking out the door at Threshold.
It remains a secret.
Nelson, thank you for giving me your patent number. The circuit on the first page of patent 3 995 228 looks very interesting indeed, but it doesn't seem to be a class AB control loop as described by Huijsing and Tol. If I understand it correctly, it eliminates cross-over distortion, but it doesn't eliminate thermal tracking problems due to varying output transistor temperatures.
"but I was rather underwhelmed by the sound quality."
____________________________________________________
I would agree with Grey on this one. Although very sophisticated devices are available to control bias on demand as happens in the Krell's one would really need to control the bias a bit before it is needed to be sure the circuitry is stabilized before that level is required. Thats pretty impossible to do......
I believe the best solution is a combination of constant high bias class A, and water cooling. That way one can enjoy the best of both worlds.
I've owned a Krell FPB 300 and a KSA-80B albiet at different times and from my recollection the KSA 80B if left on all the time sounded much better overall than the FPB 300 but had less slam. The 300 was a very short time owned amp for me...I then traded it for a BAT tube amp. That is a whole other story though.........
On another note(pun intended):To me, the Pass designs still sound much more like real live music which I am very fortunate to be able to hear live quite frequently here in SLC. That has become far more important to me than the heat produced, or the slam capabilities of any particuluar amp
Mark
____________________________________________________
I would agree with Grey on this one. Although very sophisticated devices are available to control bias on demand as happens in the Krell's one would really need to control the bias a bit before it is needed to be sure the circuitry is stabilized before that level is required. Thats pretty impossible to do......
I believe the best solution is a combination of constant high bias class A, and water cooling. That way one can enjoy the best of both worlds.
I've owned a Krell FPB 300 and a KSA-80B albiet at different times and from my recollection the KSA 80B if left on all the time sounded much better overall than the FPB 300 but had less slam. The 300 was a very short time owned amp for me...I then traded it for a BAT tube amp. That is a whole other story though.........
On another note(pun intended):To me, the Pass designs still sound much more like real live music which I am very fortunate to be able to hear live quite frequently here in SLC. That has become far more important to me than the heat produced, or the slam capabilities of any particuluar amp
Mark
OK, I'll give you a hint. The trick is to track and hold (for a while
anyway) the minimum value of the sum of the currents through
the positive and negative halves of the output stage. Regardless
of the dynamics of output current, this value is the bias, and
can be used to control an AB bias with great precision. There is
a very cute circuit which uses a couple op amps tapped outside
of the signal path and not requiring an isolated supply which then
drives an opto-isolated transistor in the bias circuit.
The decay time of the track and hold is set at a value lower than
the thermal time constant of the heat sinks and slower than the
audio signal.
anyway) the minimum value of the sum of the currents through
the positive and negative halves of the output stage. Regardless
of the dynamics of output current, this value is the bias, and
can be used to control an AB bias with great precision. There is
a very cute circuit which uses a couple op amps tapped outside
of the signal path and not requiring an isolated supply which then
drives an opto-isolated transistor in the bias circuit.
The decay time of the track and hold is set at a value lower than
the thermal time constant of the heat sinks and slower than the
audio signal.
Hmmm...as in a peak hold circuit with a resistor to bleed the cap down over a given time period? At least conceptually that doesn't sound too difficult.
It still seems that you're going to be faced with a lag problem. Take something like the Gladiator sound track that starts off on the soft side of normal and just keeps getting louder and louder and louder. Unless you purposefully build in a little overshoot on the assumption that louder is to come, you're still going to be playing follow-the-leader when the music exceeds your previous set point.
If you really want to get serious about this kind of thing, you're going to need to read the music onto a hard drive (we're assuming a digital signal here), scan the piece, calculate and record the necessary bias, then play. Only a single scan would be needed as long as the bias numbers were recorded permanently. All kinds of cute tricks could be brought to bear on the problem once you're working ahead of the music instead of behind it.
Didn't I read that Linn has a music server dingus based on hard drives? Might be a good platform for this sort of thing. Of course, if you give the PC people who live, eat, and breathe MP3 files a little nudge, they might play with it. On the other hand, given the sound quality of those files, it fits into the why bother? category.
Grey
It still seems that you're going to be faced with a lag problem. Take something like the Gladiator sound track that starts off on the soft side of normal and just keeps getting louder and louder and louder. Unless you purposefully build in a little overshoot on the assumption that louder is to come, you're still going to be playing follow-the-leader when the music exceeds your previous set point.
If you really want to get serious about this kind of thing, you're going to need to read the music onto a hard drive (we're assuming a digital signal here), scan the piece, calculate and record the necessary bias, then play. Only a single scan would be needed as long as the bias numbers were recorded permanently. All kinds of cute tricks could be brought to bear on the problem once you're working ahead of the music instead of behind it.
Didn't I read that Linn has a music server dingus based on hard drives? Might be a good platform for this sort of thing. Of course, if you give the PC people who live, eat, and breathe MP3 files a little nudge, they might play with it. On the other hand, given the sound quality of those files, it fits into the why bother? category.
Grey
These guys are have been discussing a PC CD drive based player.
Some have decided to buffer the output to reclock the data.
Would work fine with Grey's proposal. doesn't seem like any records are being set on making actual players...
http://www.diyaudio.com/forums/showthread.php?s=&threadid=17910&perpage=15&pagenumber=21
Some have decided to buffer the output to reclock the data.
Would work fine with Grey's proposal. doesn't seem like any records are being set on making actual players...
http://www.diyaudio.com/forums/showthread.php?s=&threadid=17910&perpage=15&pagenumber=21
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.