Krell's "Sustained Plateau" Design - some thoughts
I downloaded the patent yesterday and ran through it quickly... I was quite interested to discover what Krell has been doing. It's Patent # 5,331,291
In brief they use a current sensing transformer on the ground leg of the speaker output line to trigger what looks like some gates sitting on the output of a comparator to "adjust" the bias to keep the amp in Class A on peaks.
A few thoughts came into my head about this idea. Mostly about the speed of things and what is going to lead or lag and how the amp's feedback is going to react to these abrupt level shifts, will the transitions be smooth enough and what about DC balance?
Assuming the gates doing the switching are faster than the amp (not a bad assumption) and assuming the amp has been "slowed down a bit" for best results (anyone got the freq resp specs handy?) it seems to me that if the sense transformer uses any core material it may lag a bit, the comparator will have some lag, and what about the rise time of the output stage in response to the increase in bias (step). What about the balance between the two rails in terms of tracking the bias shift?
I wonder if anyone looked at the distortion produced *at* the bias switching point when the amp is playing?
And, given that the amp starts out in more or less low bias AB, and then is plateau'd up as current is required, doesn't the effective headroom *drop* with output? So, for a passage with a series of high dB peaks, there could be compression or clipping on sucessive peaks - one which passes before the bias can shift up, and one after or as the bias has shifted up...
Actually, it kinda seems like a backwards idea, unless I have it wrong... that is start with the amp in pure class A and *reduce* the bias on loud passages, to obtain more headroom! You want the highest bias on the lowest passages, where the xover distortion will be most noticable.
Oddly, unless I have gone a bit daffy, when running into a really low Z load, as the Krells are legendary for doing, this scheme works out to be a kind of SOA protection? Since to maintain Class A into low Z the bias will need to be pumped up quite a lot... which will limit the swing, and limit the max current as compared to the same amp in AB... or did I run the train off the tracks??
Ok, so you get to claim a "Class A" amplifier, and you get to use smaller heatsinks that otherwise would be required (not that they are so small), but from what I understand of the design there is no loop feedback around the output stage, meaning that the distortion specs are not particularly low (and they're not), so while ur "eliminating" xover notch distortion, is this important at all in this case?? Doug Self has shown that xover notch distortion in Class B *might be* less of a problem than AB xover notch distortion... which leads me to scratch my head a little bit.
What do you folks think?
Well, the abrupt transition on a peak would be a problem
in my book, but there it is.
At one point there was talk of an "anticipator" circuit that would
somehow devine that a big peak was coming along and make
the transition in advance, but I haven't seen any evidence that
such a thing existed, and of course it still doesn't answer the
question about large abrupt transitions up or down in bias and
what they might do to the sound.
Also I was unaware of any claim that Krell operated without
On the other hand, you can go listen to a Krell, and decide
whether Dan's approach was valid or not.
I listened to the Krell Standard amps some time ago that had the S.P.B arraignment and i can tell you i was quite impressed!
My eyes were opened to many products and CD's that day. so i remember it well.
I heard nothing that alerted me to the fact that anything was going on.
Now for comparison, listen to a Carver amp. at low volumes they have a clean top end, but raise the volume just a little bit to the point where the power supply starts switching rails and you can hear it. it gets grainy and fatiguing.
I had these in my studio (they were there before me!) and i could mark spots on the master fader where the sound was grainy, if the fader was between any of those spots (with all else being the same) everything was fine, but get close to any of those marks and my ears would get tired fast!
I swapped out the carvers for my Bryston 4B and it went away and i gained a whole bunch more accuracy on the low end! sheesh!
I would love to see some discussion on what makes the 4B so good for low end?
Here is a question for the pros, what is the delay time from input to output in a power amplifier? surely there has to some small amount of delay right?
Couldn’t an anticipator circuit be built that would look at the incoming signal before the first stage and adjust bias accordingly. it seems to me looking at the output would be the wrong end to kiss so to speak.
I think bias is moved up to the next level well before the current being drawn would cause it to leave Class A operation at the existing bias level.
It maybe possible to get it to ramp from level to level in a controlled manner but it would clearly need to be a rather rapid process.
This means it would only leave Class A operation if the audio waveform was slewing very quickly.
I think that if well implemented it is a pretty good idea.
Roughly speaking maybe if the slew rate at which the bias changes is faster than the slew rate of the amplifier then it maybe possible to gaurantee that the amp never leaves Class A operation.
On the other hand if this were the case then feedback may not be able to correct any glitches at the output while the transistion from level to level was taking place.
I don't know, but in theory it could keep the bias always set for a value that was enough to sustain the momentary output current plus some amps. If it could do that faster that the amp could be expected to raise the output current those extra amps, I guess it would work. The max output current rise and speed is defined by, say, max output level at 20kHz in 4 ohms so it can be calculated without surprises.
PS rejection ratio?
This may be a silly question/statement but...
Isn't it a feature of most 'complex' SS amps that they are quite immune to 'noise' and amplitude changes on the power supply lines?
If you have a PSU rejection ratio thats high enough, can you just change the voltage, leaving the current more or less alone, at it's presumably high level, and not 'hear' anything on the output?
"A few thoughts came into my head about this idea."
Me too....... My audible impressions are the FPB 200 doesn't sound as good as my old KSA-80 does. I had the opportunity to compare the later amps to my old KSA-80 and all felt the KSA-80 sounded better in several aspects. The downside of the old KSA series is that they need to be left on all the time and are fussy in other ways. Of course today I use Pass Labs stuff I've built myself but it would be fun to build a KSA-80 and soup it up as I have with my Aleph 2's......
Does anyone have a KSA 80 schematic they can send me or post here????
That could work but you'd need to use a switch mode power supply which is pretty complex to design and may not be able to change its output voltage very quickly.
A multi-tapped transformer would possibly be doable but you'd need multiple sets of rectifiers and filter caps to create the various supply rails.
Still it is a pretty good concept.
These being krells...
I assume price is not really that much of an impediment to either of the PSU solutions you mention. A big honkin' transformer with multiple taps, caps etc sounds more Krell of old, I have no idea whether they've used smps for the newer designs.
|All times are GMT. The time now is 10:51 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio