ABX amp switching

I'm now planning to pull my QSC ABX box out of storage to serve as my amplifier switcher in the living room.  AFAIK it has 30A relays for speaker switching, and I have verified that it uses binding posts (and not XLR's, as I had been thinking) to attach speaker connections.  Given that QSC was making professional power amplifiers with many hundreds of watts of power, their ABX box ought to be able to handle fairly high voltage as well as the impressive 30A, and not mess up balanced amplifiers that don't want either output terminal grounded.

I had been planning to use external relays to switch amplifiers, like the Schneider-Magnecraft 199X-12.  I still think this is a cool idea, I can then run minimum length cables to the speakers, with the relays behind each speaker.

Interview with Dan D'Agostino

http://www.monoandstereo.com/2010/10/interview-with-dan-dagostino-of.html

The Momentum is Class AB, not Class A.  Dan says that Class A buys very little additional performance with very much power and heat.  When asked what is special about his Momentum amplifiers, he points to his 35 years of experience designing good sounding amplifiers.

Very Ironic, given that Dan became famous for Class A amplifiers.  That was certainly how I first heard of him, when I first saw the man himself give a demo of his KSA-50 amplifier at an audiophile society meeting in San Diego, CA around 1982.  It has seemed to me and others that Dan made Class A transistor amps acceptable as they had never been before.  Now the leading seller of Class A amplifiers is Pass Labs.  Nelson Pass had previously been associated with the quasi-Class A that some Threshold amplifiers used (but it was not really anything like real Class A) during the 1970's.  But after starting his own company again around 1990, Pass went right to Real Class A with the Aleph amplifiers, and now the XA and First Watt amplifiers, which are all all real Class A.  Dan being Dan the Class A man made it possible for Nelson to be Nelson.

Dan also suggests there are interesting new ideas in his new amplifier he's not prepared to discuss.

That's fine for his business and customers, though what I like is the interesting ideas, so I'm disappointed he's not revealing more.

KSA-150 and KSA-250

It is unclear whether the KSA-150 and KSA-250 used a managed variable bias system.  Reading the manuals carefully, it sounds as if there may be some kind of bias management, but not as sophisticated as Sustained Plateau Bias which officially appeared with the KSA "S" models.

If KSA-250 has multiple bias levels, John Atkinson's famous calculation of the Class A power of a KSA-250 is only true for the lowest bias level, not the higher ones.   (In the first version of this post, I was sure KSA-250 had multiple bias levels.  Now not at all sure.)  I'm guessing the relatively cool running 250 had variable bias whereas the much hotter idling 150 might not have.

The Mark Levinson amplifiers like 331, 332, and 333 definitely use some kind of managed bias system to provide more Class A power on demand.  It says clearly says so in their manuals but only calls it "proprietary" and "developed by Madrigal Labs" and doesn't give it a fancy marketing name.

The KSA-250 and KSA-300S were much larger than the FPB 300 and approaching the FPB 600 in transformer capacity and weight.  The FBP 300 chassis actually seems like the successor to the KSA-200S chassis.  So the successive models appear to be in two groups (though within each group, models do get slightly heavier and certainly more sophisticated over time):

KSA-150

KSA-200S

FPB 300

FPB 400cx

(this group always had something like 3kW transformer)

and

KSA-250

KSA-300S

FPB 600

FPB 700cx

(this group always had something like 6kW transformer)

As time went on, Krell was either getting or at least advertising more power from almost the same amount of iron and aluminum.  The old 150 could double down all the way to 1 ohm and get 1200W.  The FPB 300 doubles down to 2 ohms…where it gets 1200W.  Actually the FPB is said to get more power into 1 ohm, but the 1 ohm power is not rated and it is certainly not doubling down to 2400W into 1 ohm.  I think I've heard that something like 1600W may be available into 1 ohm.  Ultimately, the FBP 300 and FBP 400cx have the same transformer, output transistor assembly, and chassis and the 400cx is rated at 1600W into 2 ohms.

If Krell ever advertised Sustained Plateau Bias for the Evolution models..they don't seem to do so any more.  I thought they did at first, perhaps when Dan was still with the company.  It doesn't seem like they were doing full rail regulation anymore, just a fancy kind of filtering.

I think the sustained plateau bias on my FPB 300 is great, though it could be smarter still and I'd still like a manual indication and control as with the "S" models.  The S models lacked the sophisticated rail voltage regulation, and I think they had slightly higher levels of distortion than FPB, but I like the ability to control bias levels.  If I could preset/hold the lowest bias level, I wouldn't need a second "summer or background music" amplifier.

200 degrees C

The schematic for the Krell 200S indicates MJ15024 and MJ15025 power transistors, although I have read that for best results you need the custom versions Krell actually ordered from Motorola which are only similar to MJ15024/25 but better.

The specs say the 15024 has a maximum Normal Operating Temperature of 200 degrees C.  (That is like 2.4 times higher than any temperature I've actually measured on the heatsinks.)

Device dissipation is 250W per device, and 16 amps.  The derating for power is 1.43 watts per degrees C, so at 80 degrees C the max dissipation would be 171 watts.

I suspect the transistors used in my FPB 300 are at least this good, and I believe there are at least 10 transistors per channel (and maybe 20 or more).  Therefore, even at 80 degrees C, one channel could be dissipating at least 1700W (and that's not counting the power consumed in regulated power supplies, drivers, etc).  and there are 20 outputs devices in each channel.  Therefore, even at 80 degrees C, each channel is capable of dissipating more power than the power supply could possibly deliver, 3400 watts.  And that's not counting the power supply and driver transistors, etc.

Bottom line is that this amplifier is safe to operate at the full maximum temperature of 80 degrees C, which is about 180 degrees F, and still deliver full power.

In my most recent testing, it does not seem like the left channel is rising to the next plateau.  Instead, it looks like it is simply drifting high.  The plateau bias starts it at the right level, but then it slowly drifts up, more like 50-100W than 150W.  Of course if it ultimately drifts up to 80 degrees C, plateau bias computer puts it back down again.  After an hour of idling with inputs connected to the Audio GD Dac, I was seeing about 500W consumption, with one channel being about 145F and the cool one being 115F.  Actually, the hotter channel is probably in a better temperature range for good sound.  After a few hours of playing the amp was happily consuming about 600W, which is perfectly fine, with  155F/135F temperatures, also fine, once again, the drifty hot channel actually being more linear.  Transistors tend to get more linear when they are hotter and/or run with greater current.

If the amp is kept playing music, it might cycle through the 180F max but then settle down to 155/135F with about 600-700W continuous consumption, which is about optimal.

If I could set the bias level to level 2 (600W) that would be nice, but I'm effectively getting about the same thing, just with an occasional higher reach that warms the amp up quicker.

That Class A Feeling

I've been getting that feeling.  That same feeling I got when, as a kid, I wandered into Mel Shillings' fabled Music and Sound, which had relocated…into my neighborhood in Southern California!  They had the huge amps, the planar speakers that created extraordinary stereo images (about the first decent stereo I had ever heard), and something else I forgot about until just now…

Air Conditioning!  Running my Krell with typical power consumption in the 600-900W range is generating a lot of heat.  And, since my listening room is the living room, which happens to have the thermostat, almost as soon as I turn the amp on the Air Conditioning comes and and pretty much stays on as long as the Krell is running.

Well I'm sure the Krell loves being the center of attention again, right in the center of the room, getting the main blast of cooled air from the ceiling vent on the other side of the room.  But while it's bracing to sit down into the stream of cold air, and brings back memories, it's also expensive and it even seems to make me a little sick.  And when I'm done listening the rest of the house has been chilled about 3 degrees below normal.

The problem here that I hadn't quite solved when I tried moving the Krell into the corner--shielded from cold air by the couch and then thought it was overheating--is that the thermostat is right on the opposite side of the same room, and the fully air conditioned air hits the Krell and practically bounces right back to the thermostat.

Back then I did get a new thermostat installed…but it never quite satisfied me wrt programming (it was basically impossible to program on the wall, that's another story for my Computer Critical blog).  What I really wanted then, and now, was a portable "personal" thermostat I could locate to another room.  The perfect other room would be my bedroom, which often doesn't get warm enough in winter, but isn't much bothered by heat from the Krell.

Well they do make such a thermostat system now, the Prestige IAQ 2.0 System, which has a portable thermostat called the Portable Comfort Control.

OK, I could just turn the thermostat up, but I hate it when I forget to do that.  And what if I'm running the Krell all day???  I think thermostat in the master bedroom is the perfect place…I could just leave it there always except during parties I'll put in back in the living room.  Or if a friend is doing lots of baking in the kitchen.

When the AC isn't running, the Krell could still use some air movement, so I just bought the Econo-Cooler from Cool Components (these aren't sold by many online stores btw).  I've generally hate fans in equipment.  But I'm OK with fans I can dial down to the lowest speed myself…and if it's noisy I'll try something else.  I only wanted to start with one Econo-Cooler (CP-EC) but it didn't seem like I could order just one and get the full ps-300 power supply which has many possible voltage levels.  They also have control modules with digital temp readout, thermostatic control, and triggers, but I'm worried about the difficulty of getting the thermostatic control to work properly.  I don't want to put it directly on the Krell--that might fry it (the thermostat).

Another thing is to have connections for my "backup" amplifier, the Aragon 8008 BB.  It doesn't generate quite as much heat in operation (about 1/2 the amount of idle heat) and it has now been moved into the corner where it probably won't affect the thermostat as much.  I'm thinking that in the long run I want to have it hooked up through my QSC ABX box (but not directly) so I can actually do ABX tests.  But I don't trust the speaker relays in the ABX box (nor the connectors, nor the box itself) so what I'd do is use the box to switch a DC current which would then control a pair of 300V 40A DBDT relays right behind each speaker.  For decent ABX tests, I might eventually have to have my own computer control that "works" the relays a bunch of times so you can't figure out A/B just on the relay sound (even the QSC ABX box itself has this problem a bit…with some practice you can hear the difference in the relay sound).

Well I looked at lots of things, starting with pins that could be directly inserted into the cheap 5-way binding posts at the back of the Acoustats (still one of the best options).  Finally I decided to use a familiar old solution--banana plugs.  I'll "stack" the plugs (plug one plug into the holes on the other) to connect the speaker to the A amplifier or the B amplifier.  The Acoustats will remain connected with bare wire that goes to one plug, and so will the Krell.

(BTW, no use in plugging bananas directly into the 5 way posts on the Acoustats.  Heavy banana plugs don't take long to fall out of those cheap posts, and are often hanging part way out.)

The only local Radio Shack had hardly anything anymore (sad!!!) so I ordered these from Parts Express.

OK Class, how much Class A Power ?

I've mentioned before in this blog how manufacturers routinely inflate Class A power claims by various technically correct but misleading means.  I've also discussed the correct calculation, which is 2RII where I is actual idle bias and R is the load.

The Krell FPB 300 has different idle bias depending on situation.  Temporarily I have seen it consume about 1200W continuously for 30 minutes or so after turn on.  But over a whole day, it can't sustain much more than 700W, which seems to rest on an second bias plateau around 600W.  The bias circuit kicks everything down when the heatsinks start getting above 180F.

For 600W idle, that's 300W per channel, and about 2 amps across the 170V maximum peak-to-peak voltage.  So 2RII gives us 32W into 8 ohms and 16W into 4 ohms.

An amplifier with 1/2 the voltage swing but dissipating the same power would be running 4 amps in consuming that 300W, and therefore have 4 times as much Class A power.  The price for having more maximum power on tap given the same dissipating capacity is that you cannot have as much Class A power.  People would like to ignore this but it is basic math.  If you want more peak watts from the same quiescent consumption, you have to give up the Class A Watts.  Also, Class A Watts do not easily double down as they are limited by current in most amplifiers.  So Class A Watts will halve into each halving of impedance.  The famous Mark Levinson ML2 did not do this because the amplifier was voltage limited at each lower impedance down to 2 ohms--so it did indeed "double down."  But it was not practical for many people, and especially me, because the maximum power into 4 ohms was 50 watts (25 watts at 8 ohms).  And you get that 50W from a single channel mono block about as big and heavy as my Krell.

Here are the Pass Labs .8 specifications.  My Krell can continuously dissipate about the same as Pair of Pass Labs XA60.8's, which have a combined ship weight of 214 pounds.

Cook that Krell

I was all worried that one channel of my FPB 300 was running hotter than the other.  I shouldn't have been, my Krell, which was sitting in the corner of the living room for 4 years (during my recent building boom) waiting for me to get around to prepare for getting fixed, was operating normally.

The FPB *has* to heat up to the normal max temperature to trigger something like recalibration.  It then sets itself back down to lower, usually in the 600W bias plateau, but if the amp has been driven hard enough it will get set back to the 300W minimum bias plateau.  (My wattages are the actual amplifier power consumption…what is available as Class A power is different and will require a long explanation later.  What has revolutionized my understanding of the FPB amplifier is watching it's power consumption from moment to moment over hours of listening.  Now I think I understand it.  It makes very good use of the ultimate limited resource--the heatsink area of the amplifier.  You could say it "optimizes" the resource for the requirements of the situation.  I am happy to be even more a fan of Sustained Plateau Bias.  It's a way to approximate a much larger non-adjusting Class A amplifier, compare my FPB 300 with the metal of a pair of Pass Labs XS300's, which have similar power into 2 ohms, but use about 6 times as much metal.)

As you are driving the amp hard, it will hit the top normal temperature, which appears to be about 165 degrees F, then get set back, and this may happen over and over.  Ultimately, when driven hard enough, the amplifier becomes a high bias Class AB, with 300W minimum bias, and there is no more setback, consumption just falls immediately after every peak.  Many people may not drive the amp that hard--I have the least sensitive loudspeakers ever sold, and with the most difficult to drive load.

Once both channels have hit max temperature once or twice and been set back, they begin to have fairly identically, so I am wondering if hitting max temperature is a necessary thing to keep the amp working right.  Moving the amplifier to a different location is going to cause differential air flows…which has a profound effect on passively cooled heatsinks, and a new recalibration--hitting max temperature--will be necessary before things are balanced again.

As I was using my amplifier for awhile, the left heatsink had it's bias setback, and before long it was the cool heat sink compared to the right, which took it's time to get to 165F, but after some harder playing it did too, and then the right channel got the bias setback.  The power consumption fell to no less than 600W, but during ever more brief pauses in the music it would ramp down to 300W.

I am just guessing that a recalibration occurs when the bias is set back.  It would make sense for the amp to do its recalibration on thermal setback, as then it is changing the bias in a big jump anyway.

165F is a fairly high temperature for audio amplifiers, but it is nothing in the world of power inverters, for example, which require similar parts.  The Krell was specifically designed to run this hot as part of it's normal behavior, and as a way to get more class A power from a limited amount of metal, even if it's not usually considered a limited amount of metal.

Update: I'm seeing 175F after a day of hard playing.  I think the ultimate setback point is 180F as it was for the KSA-x00s amplifiers.  It the sensors see that, the amp kicks down the bias.  When the heatsinks have been cooler, they don't measure the temperature as quickly as the sensors do, so the rise in setback point actually demonstrates the sensors read quicker than my measurement of heatsink temps, as well as the actual temperature being higher than I thought.

It's only to be expected that if there is a setback point, eventually things will hover around it.  So 180F is a normal Krell temperature.  If the amp shuts down, or if I see 200F, I should call Krell.

Power is in the 700's with more than a second silence dropping to 600W.  Still takes about 20 seconds to get pushed down to 300W, so the nominal bias level is 600W consumption now.  So this is a long term sustainable bias level.  Actual wattmeter readings are typically in the 700's.

That bias level is comparable to A pair of Pass Labs mono XA60.8's, each of which weighs about the same as the Krell, and continuously dissipates 375W per mono block chassis.  So Krell is made to cook at max temperature, Nelson Pass designs his amps to have a more moderate temperature--at the cost of twice as much metal for similar performance.  And of course the XA60.8 has no where near the voltage or current capability of the Krell either, even as a heavier pair of amps.

The first XA amp which beats the FPB voltage spec of 60V is the XA200.8, which has 62 volts.  Only the XA160.8 and XA200.8 (and XS300) have 50A current rating.  Krell doesn't give a current spec, but Martin Colloms measured the FPB 600 as having 117A and described it as arc welding.  The XA200.8 weighs 177 pounds (per mono block) and dissipates 750 watts per channel.  I pushed my FPB to 1400+ watts draw, but only for a dozen seconds as it was already warm.

Krell FPB amps are the poor man's Pass Labs…   The Eco-Freak Pass Labs…

People will say Pass Labs is fundamentally different, all about simplicity, etc.  The FPB amplifiers have a no-feedback output which is simply an array of emitter followers in parallel and differential push-pull.  That's simple and elegant. Stuff ahead of that is a bit complicated, of course, including but not limited to the bias system.  But nevertheless the FPB specs bandwidth at 240kHz and 0.1Hz.  The distortion is spec'd at 0.02% at 1kHz and 0.15% at 20kHz and that's for full power.  Pass typically specifies 100kHz and 1% distortion.  (Of course that means he's too indifferent to specs to bother to measure distortion just before it begins to rise near clipping.  But it won't out-do the Krell either.)

I keep forgetting this, but Krell specs the full output (300W) input level as 2.35V RMS, but I continue to believe that more input is possible to reach the full peak output implied by 60V, 450 watts into 8 ohms.