Wednesday, April 5, 2017

Krell is Back !

My Krell FPB 300 amplifier, a very high end amplifier from the 1990's, is back from a full capacitor replacment from the Krell factory in Connecticut.  It sounds fabulous!

And for the first time since I bought this unit from the original consumer owner in 2008, it works perfectly!  It doesn't even clink anymore, the chassis clinking being first thing I noticed (and hated with a passion) when I first got the Krell in 2008).  Because of the clink, I started doing bad things, like keeping the top cover off and relocating to the corner, which never got rid of the clink but created other problems.  But now, along with the temperature not running slowly up to 180F, as it had in the recent past back to the beginning, it doesn't clink either.  The overtemperature cutback and the clinking were clearly related, as I began to figure.  And the clinking was happening more and more as the plateau bias controller was ramping the left channel up to max, then cutting back.  Now the plateaus simply float on top of the highest peaks, without any limits to full operating capability ( cutback).  Just the way the amplifier was designed and advertised, but I had never experienced until now, the second time my Krell has been back from a factory repair, and I might add, the first time since Rondi D'Agostino took over operating the company again, as she had in the beginning.

(Less clinking (none so far!) than even my fixed bias Aragon 8008 BB, which always clinks about 10-30 seconds after shutdown, and sometimes clinks during warm up.  The Krell, so far, has not made a single clink (!!!), even when I pushed to to max volume for 20 minutes follwed by a shutdown to standby, I waited a few minutes, and no clink!)

I had been becoming worried I might never see this incredible amplifier working correctly.  It was too much to ask.  It was an engineering tour de force when it was introduced.  Or more precisely, the engineering tour de force was the statement reference amplifier from which the FPB amplifiers were derived, the Krell Audio Standard 2 amplifiers from 1993 were the ones to introduce the key technologies: sustained plateau bias with full power regulation and balanced operation.  The full power and balanced operation (source of the initials FPB) are not mere bonus bragging rights, they are a key part of making the revolutionary sustained plateau bias system work.  It is a very complicated amplifier, though what it does, in principle, is quite simple.  I've always been impressed by the ideas behind the amplifier, I thought I thought of some of them myself beforehand, and always thought that was the way it should be done.  I even vaguely remember mentioning my preferences, and some of these ideas, when talking to D'Agostino when I met him at an audio meeting around 1982 demonstrating his very first commercial 50W Class A amplifier, that  everyone at the time was saying was the best thing ever.

"I want a 400W Class A amplifier, which doubles down Class A power into lower loads like an ML2,  down to 1 ohm, without a fan!  I hate fans!" I said.

"It would weigh as much as a car and cost a mint." D'Agostino said.  "It's just physics."

"Well you could do 3 things. You could use transistors and circuit designs intended for higher temperatures.  You could use massive, truly massive external heatsinks that got more than just too hot to handle.  That's fine with me, hot is no problem.  You could raise the class A bias up and down, as need be, so there's less power not delivered into the load.  Not sliding bias, that sucks.  But just hold different levels for awhile, so it's not noticeable that you are changing them.", I said.  Or I think I said some of those things anyway.

Actually, the FPB does more, which I could have mentioned because I knew at the time, the regulated rail voltage supplies, and the full balanced operation, and those are part of the picture too because they make the sliding bias system sonically transparent, any effects are completely canceled out.  But I hadn't thought of that connection.

I've always thought it was a brilliant design.  Perhaps too brilliant for a mere audio company, even if it is Krell, to keep running over the long haul though the Service Department.

But I am pleased!  Krell did it for me this time!  My Krell works perfectly, sounds great, super, superior, just making everything listenable, with the hottest input and output cables, no dampening tweaks needed, just great, correct, perfect.  The way it should be.


As audio fads have come and gone, and the D'Agostino's had been driven out of Krell, I wondered if anyone cared enough anymore to get this amazing amplifier completely working again.

Fortunately, just about the time last year I decided to send my Krell back to the factory again, Rondi D'Agostino, who it should be remembered was the co-founder of Krell Industries and the original CEO, was back at the helm as the CEO, the dispute between the D'Agostino's and their venture capitalist investors from the 2000's finally having been resolved.

So now that Rondi is back, Krell is apparently doing things right again, and unlike my first Krell repair experience in 2010, when the company was in the hands of the vultures, this time (2017) it got done correctly.

It was a perfect experience all around.  The repair cost estimate was produced a week after the amplifier arrived at the factory, and less than two weeks after that the amplifier was delivered by a Freight company, without even a scratch on the original factory box.

In contrast, when I had sent the Amplifer back to Krell in 2010, it seemed to be taking a long time for the repair, and I didn't want to rush the factory.  I was hoping they would take the extra time to fix the clink.  Finally after two months, I called, and they said it had been misplaced but they would send it back to me immediately, which they did--deducting the shipping cost because of their delay.  When it arrived the original problem that had caused me to send the amplifier in--noticeable distortion in the right channel--had been fixed, but two other obvious problems had not been fixed.

A constant since I purchased the amplifier in 2008 was that the left side seemed to be running too hot.  At any given time it would be running WAY hotter than the right side.  For quite a long time, I didn't know how to understand this.  I knew that Class A amplifiers ran hot.  Was this too hot?  Was the problem actually that the right side was running too cool?  The fact that the right channel in 2010 developed easily noticeable distortion suggested that the right channel might have been running too cool, maybe it was the much hotter left channel that was correct.  Until around 2012 I did not have any way of measuring the heatsink temperatures.  So I hadn't complained about the imbalance, only about the clinking (and about which nothing was done).


I know that many people don't consider sustained plateau bias amplifiers "true Class A," and some might even not understand that it really can and does run very high bias currents when it can and when it needs to in order to maintain effective Class A operation.  At times I had my doubts too.  But I have investigated the operation of this amplifier in considerable detail since acquiring it in 2008, so I know what it does.

Prior to the mind blowing Krell Reference Standard amplifiers, Krell had backed away from the true Class A operation of the first 50 watt model in order to get much higher powers.  So for example it was well known that the KSA-250 didn't run nearly the bias it would need for 250 watts of Class A power.  John Atkinson did the calculations, it was really "just" a high bias Class AB amplifier.

Back in 2008 I had decided to get much higher into the High End than ever before.  I had just had a dental near death experience...months of intense pain and a jaw so sore I could barely slip a straw into my mouth to drink liquid foods.  After weeks of getting no good answer from my dentist, I went to my medical doctor, who immediately prescribed effective antibiotics and directed me to find another dentist at the dental school.  Finally, months later, it had all worked out, and in celebration I bought myself the speakers I had long dreamed of owning, full range electrostats nearly 8 feet tall in a line source configuration: the Acoustat 1+1's.  I had remembered reading in the International Audio Review about the 1+1's, Peter Montcrief liked the idea in principle of having a narrow planar electrostat, though critical as ever he wanted it half as wide still and twice as tall: the 1/2+1/2+1/2+1/2.

I had also heard other Acoustat models, and generally liked them, actually somewhat sorry I hadn't gone that route instead of endlessly modifying my LS3/5A's.

But now I had the Acoustats, I really didn't have a suitable amplifier for them.  About the highest end amplifier on hand was a Parasound HCA-1000A, with neither enough power, nor enough magic.  I ran with the Parasound for awhile, applying vast amounts (more than I realized) of EQ in order to get ruler flat response at the center of the room (where, as it turned out, there are huge suckouts at the room modal frequencies).  And playing pretty loud, with lots of EQ, the little HCA-1000A actually shut down.  I had never seen that happen before.  (I could have gotten by with less EQ, and in fact I did for awhile, but this experience convinced me I *needed* a higher power amplifier, and I justified it on this basis.)

So what kind of amplifier should I have for my infamously power and current hungry speakers?  Acoustats were so famously incompatible with (unstable) amplifier that Acoustat had introduced their own special "trans-nova" amplifier with a very unusual circuit.  But somehow that seemed too pedestrian to my grandiosifying mind.  Besides, there weren't any of those on eBay at the time, and there were lots and lots of Krells.

I had long admired the pairing of Class A amplifiers and Electrostatic speakers, going back to the pairing of Mark Levinson ML2's and stacked Quad ESL-57's in the well known HQD system.  (It was a very special playing of an HQD or similar stacked Quad system in 1976 at a rock musician's home that had literally blown my mind, turning me into the high end nut that dropped out of college for awhile and worked in a high end salon.  But then discovering the affordable LS3/5A's and then-affordible refurbished tube amplifiers, I got back to having a normal life in the software field, and minor audio hobby on the side, probably for better.)  And also with Stax Electrostatic Headphones, where they made a big deal about Class A amplification in the matching amplifiers.

As I understood it then (and basically now also) the Class A amplification has a distinct advantage for electrostats.  For one thing, they had inherently lower distortion and the electrostats are very revealing of any distortion because they are so transparent.

Some Class A amplifiers also had other advantages, for example the aformentioned ML2's have double the amount of power into each halving of the impedance, so 25 watts into 8 ohms, 50 watts into 4 ohms, 100 watts into 2 ohms, and 200 watts into 1 ohm.  Amazingly the ML2's had all that power into low ohms, but were pretty much absolutely limited to 25 into 8 ohms, which made it very suitable indeed for the Quad ESL-57's which would arc above that voltage, but needed lots and lots of current to maintain equal power into high frequencies, because of the capacitive load, almost just like a capacitor.

Now in this context, when I say "Class A amplifier" I am thinking about a solid state Class A amplifier, preferably with massively huge heatsinks, and no fans.  I went to work at an audio salon mostly lusting after tubes, but it wasn't long that for some reason I began to lust more for the big solid state amps.  I distinctly remember lusting for the beautiful Threshold 800 for example.  But I did want something more like "real class A."  I thought about that a lot, beginning around 1978.

Tube amps may not be so great with transformer coupled electrostatic speakers.  A custom tube direct drive amplifier would be the *perfect* thing for any electrostatic speaker iff (and a very big iff) designed correctly.  This has been done, but never done correctly commercially as far as I know.

(A friend of mine has such a design, built for a few Quad ESL 63's, using a particularly linear very high voltage tube, of which I have a stockpile...but building the amplifier would be months of dedicated work for me, so I can't imagine doing so now.)

Notably there was an Acoustat direct drive amplifer, which used tubes way beyond their ratings and so was noted for reliability problems.  Likewise, sad to say, the legendary Beveridge 2A, though that might have been a tad better than the Acoustat, it too could too easily become unreliable.

So transformer coupling is far more reliable, in the real world of commercial stuff anyway, and not the far reaches of our imaginations.  And with transformer coupling of an electrostatic speaker, you need the high frequency current that pretty much only high bandwidth transistor amplifiers can provide.

Anyway, knowing (or thinking) I needed at least 400W (at 4 ohms), I didn't originally think of a class A amplifiers, just big brutes like the Aragon 8008 BB (which I also have now, but got somewhat later, when the Krell was sent in the first time for repair).  The Aragon actually has over 600W into 4 ohms, and even more into 2 ohms.  But I missed the first sale of an 8008 BB I saw.  Then I also thought I could get a Bryston, with a 20 year guarantee if I bought a brand new one at the real dealer price.  But that would be so boring, I figured.  I want to leap for the best, not be stuck with something forever because it's guaranteed.

When I started my search, mostly online and in my collection of Stereophile, Absolute Sound, and Audio magazines, I didn't really know much about Krell amplifiers, other than that the first was true class A but fan cooled, and later ones got higher power but not so much class A.

It was during this search for a new amplifier that I discovered that the kind of amplifier I had hoped and dreamed about HAD been made, about a decade earlier, in the FPB amplifiers.  And I found one that looked minty but was priced $1500 less than a new Bryston.

The FPB had the sustained plateau bias, and also the doubling down in power down to (at least) 2 ohms (they don't advertise 1 ohm but it does have even more power into 1 ohm just not double) that is useful for electrostats, and also regulated power supplies...another feature of the ML2's that I had long hoped for, electrostats of course would benefit from even higher transparency and reveal the otherwise unstable power...also as it turns out regulated power is extremely important for the plateau bias to work well.

So this was the amplifier of my dreams, but the very first time it became a nightmare because of the clinking chassis.  It was not just the clinking, but it was very unpredictable.  It wouldn't just clink some time after turn on or turn off, it would clink every so often, perhaps 30 minutes or so, but unpredictably so.

So here I was, in my mind gambling on the very idea that I could (really) tell the difference between amplifiers in their electronic performance, and one amplifier was making a very real difference, a very bad one.

Much much later I just came to appreciate the Krell so much the clinking seemed more like just a minor flaw that could be lived with.  But at first, it drove me mad.

Actually, my very first thought was "oh, how stupid, it uses some clinking relays to chose the different bias levels."  It took some time, and even opening the top cover, to determine conclusively that wasn't true.  Now I know the bias levels are indeed controlled entirely electronically, though of course there are protection relays on the power only (not on the outputs).

But if this amplifier was going to clink, I decided I was going to minimize it.  I tried having a top cover and having no top cover.  Even removing the chassis pieces which hold the top cover.  I tried tightening screws (they take a Torx head) and loosening them (which made more sense to me) and even retorquing all the screws on one side to a specific amount.  It was in fact after that last operation (and after the first repair) and having moved the amplifier to a corner by the couch that I noticed it was running so hot that I could not believe it to be normal, that probably my retorquing had broken the amplifier, that I put it away in storage for my dream of full disassembly, reassembly, and then shipping back to Krell or someone else again).  During one power cycling it actually shut down (but it had gotten very hot already).  About 3 years later I took the amplifier out of storage and found, surprise, I hadn't broken it at all, it ran OK in the center position of the room, though getting pretty hot, but damn it I was going to get my money's worth out of the first repair, I'd just run it until it sounded bad and completely quit working.  So though it was more than 7 years from the first repair to the second repair, there was an intervening period about about 3 years when the amplifier was in storage and I thought it wasn't working.

As it turned out, in the end it didn't entirely quit working, I had just tracked the problem so well by then I could tell not only was it not normal, but it was getting worse and it was idling to maximum temperature in 30 minutes which was clearly unacceptible.  So I described the overheating to Krell exactly and they gave me an RMI in October and I finally shipped (by Freight) the amplifier to them in March (they say you could take your time in getting the amplifier to them, and they didn't give me any problems at all).

Meanwhile, during the second wait time (mostly me not getting around to doing the full repack, which is a bit challenging and required me to buy new foam pieces) I had fun with the Aragon again, this time discovering it had large distortion from underbias which I fixed.

So, I'm always having fun one way or another.

But it's wonder to have the Krell back, and now it doesn't overheat or clink at all.

At idle, where the Krell seems to run at 300W (I think it's been reprogrammed to the 400CX bias levels, which is 50W lower for the idle, but then has a 400W instead of 300W maximum...the 400cx being basically the same amplifier but with the CAST front end connection AND the reprogrammed bias, everything else is the same) it heats up to around 108-116F, I measured that after 4 hours and also overnight.

During very heavy continuous use for several hours the heatsinks will eventually start reaching above 150F, but I've never seen them at 160F so far.  That's consistent with typically using the next plateau level in both channels, that's 600W, and just briefly a level or two above that.

Back before the repair I generally measured the 600W operation over the long haul.  But when the amplifier started up, it would be at 600W, then 750W, then 900, then perhaps up to 1300W, then it would kick back down to 600W, and some seconds after that there would be a clink.  That was because apparently the bias controller in one channel was ramping up to something like maximum, 180F was being reached (and I measured that also), and then it was engaging a limit to the second pleateau or 600W.

If this thermal limit is mentioned at all in the FPB manual it is very parenthetical, but in the earlier KSA-300S it is spelled out quite clearly in the manual and you can see the bias lights show it: if the temperature gets "too hot" (they don't say 180F but that's what I have measured as being the limit) you are limited to the first two plateaus.

What was wrong was not the limiting action--that is normal operation--but that one channel should just keep ramping up to maximum.  The other channel didn't do that, and it stayed cooler, pretty much as both channels are right now, 140F in pretty high power  use, and 160F in sustained very high power use, and below 120F at idle.

In the end, it wasn't just that the left side was routinely around 180F and there were clinks every 30 minutes or so.  No it was ramping up to max bias so much it was clinking several times a minute, and going to 180F at idle in 30 minutes.  In all that, it never shut down, because it never goes over 180F thanks to the thermal limit on plateaus.  But it was clearly wrong, maybe on the last edge of doom, and annoying because of the continuous clinking.


Unlike the Aragon, which needs high cap input cables and high inductance zip cord wiring to sound nice (mellowed down), the Krell does actually sound pretty mellow if you use those things.  You should not, you should use the most transparent cables, then you get the most incredibly transparent sound.

In my case, I'm using 18 inches of 18pF per foot and heavily shielded Blue Jeans LC-1, and Canare 4S11 speaker cable.  If I put those on the Aragon it sounds harsh, but it sounds just right with the Krell, perfectly balanced.

At all times the Krell has this incredible depth I have called multilayered.  There's just more depth there, and things are fully separated from one another.

It's just magic what it does for my system.  That's why I put up with so much before, until this point.


I am sadder but wiser in some ways still.  I used to believe the FPB had it's own peculiar magic with a zero feedback output.  The output stage was "on it's own" to deal with the very powerful and complex  back EMF from the electrostatic speaker.  As such it simply reacted by the simple impedance of the and immediate, without taking the long route through the compensation networks and the input stage and all all over again.

It turns out the Krell does not have loop feedback to the input stage, the standard of nearly all solid state amplifiers, it has feedback to the driver stage immediately behind.  So the output is not non-feedback as in my imagination.  The feedback in both input and output parts of the amplifier allow super high bandwidth, so I recall the FPB 300 was measured to 400 Khz (full power) while the spec was only 300 Khz.  IIRC the Aragon is around 60kHz and that's typical.  One recommendation is that the amplifier like every other stage should have 10 times the nominal bandwidth required so that the total doesn't become audible.  So 400 kHz is not unreasonable, it's fantastic actually, but it's not the output devices own 50Mhz or so either.

About the fastest amplifiers I know are the Spectral ones and they go at least to 1 Mhz and they are highly praised, and out of my imagination in cost, especially new, and even 20 years old far pricier than my mere FPB.

So here's a conundrum.  In the one case you may get "simpler" designs, with fewer total stages for example, and in other you may get lower distortion, much higher bandwidth, higher class A power, and so on.  Which way to go?  Well I don't see the Krell as being all that much more complicated than, say, Pass Labs, but with somewhat more practicality.

I think I'd truly like a pair of Pass Labs XA200.5's, or better, as a "true" class A amplifer.  That would give me half as much possible Class A power per channel (see numbers below for 400cx) but of course truly continuous Class A--"true Class A"--operation rather than a complex bias control system.  And, it would weigh about three times as much, and consume more than twice as much continuous power than the Krell does during full tilt operation--and the Krell consumes far less at it's idle.

So--while I continue to lust at the likes of the XA200.5 (and I think my new Carrier Air Conditioning could handle full tilt operation here in sunny San old one couldn't handle the Krell) I do notice that the Krell has some fairly practical advantages, not even counting the cost.  One doesn't think of the Class A Krell as an enviornmentally advantageous amplifier (btw, all my electricity comes from wind power through a regional program...which has helped make Texas the #1 state in wind power) but compared to what would need to do to achieve some similar things makes the Krell relatively environmentally advantages--and by huge measure actually compared to the amp with 1/2 the rated Class A power, far higher distortion, and lower bandwidth.)  Of course, the ultimate replacement would be a pair of XS 300's.  At that point, there may be additional disadvantages than the mere cost of electricity for operation and cooling, though that would be considerable, it would be worth it of course, I guess.

But Class A operation, without the cost of a mint, isn't precisely achievable in lower impedances either, I do not believe.  Krell's FPB give you the advertised 300W "true" class A power (or whatever, and up to some limit of time) into 8 ohms, it doesn't double down.  That 400W (now that mine has been updated to the 400cx bias levels, as per factory in 2010) Class A power does not at all double down into lower is the maximum level.  Therefore, there is "less" class A power into 4 ohms, 2 ohms, and so on, in full class a (but again, not at all determining the ultimate limit).

Those number in the 400W case are 400, 200, 100...a mere 100W class A power into 2 ohms.

That's still quite a lot, however, lots of primitive stuff didn't have ANY power into 2 ohms, and the Krell has a total of 1600W into 2 ohms, just only 100W of it being class A.

Anyway, the story is more complicated, but chances are 100W of class A power into 2 ohms, or whatever it is, is good enough.  That is the sort of thing I've learned.  Though it's great to have the bandwidth and power capability into 2 ohms, in real life there better not be more there there, in the upper octave of the speaker, for very long.

So thanks Dan for giving me what was possible, anyway, it's a truly miraculous product, and exactly what I wanted.

But it obeys the laws of physics too.  (That is how I think of it, though perhaps I know better, as a force of nature.  Without output stage feedback, as I had imagined it, it surely would be, and perhaps not always to the better.)


Another example of the high end Class A amplifier is the Mark Levinson ML 20.6.  I sometimes forget this is the successor to the ML2 of a decade earlier, and it has 100 watts into 8 ohms, doubling down to at least 400 watts into 2 ohms (and 800 watts into 1 ohm???).  It's a true fixed bias Class A design, however I strongly suspect the bias doesn't give you the full Class A power into impedances lower than 8 ohms.  Like the Krell, however, it has regulated rail voltage supplies.  This was Stereophile Editor John Atkinson's reference amplifier for a decade or so, I think he replaced it with a 334 or perhaps 33H.

I'm perhaps too strongly attracted by the idea of Class A.  People tell me that you can get very linear class AB amplifiers, and that Class A is no magic guarantee of the best linearity.  However, it seems to me that no device is linear beyond or even close to cutoff, so the 'linear output' of a Class AB amplifier is always a composite way better than it's parts, and that bothers me.  There is also the issue of transistor on/off switching, which you'd rather they not do, but that can be avoided by various tricks like the ones Nelson Pass used in the original Threshold 800 and 400 amplifiers, and I think has become less of an issue with modern devices or tricks like Nelson's are always used--I'm not sure which.

At the end of the day, I want as much Class A as I can get, and all the way and beyond is fine.

No comments:

Post a Comment