Now for something different...

[kevin gilmore]

I am personally not going to spend any time whatsoever doing an A10/B52 board.

Neither am i going to do an opa541 board because i have used my

old power supply boards to test what it sounds like.

The blackhole sounds way better. Especially with the alfet output devices.

I will probably do a modification of the current kgsshv power supply board to do 600/300 volts

for aristaeus/triode amplifiers.

It definitely was worth the effort to reverse engineer ray's piece of crap.

The fact that the stealth,raptor,B52 and now A10 are all exactly the same is

quite entertaining.

Edited March 5, 2012 by kevin gilmore

[spritzer]

That was just NoNoNoNoNoNo trying to muddy the waters. Let's put it this way, the BHSE is a Blue Hawaii but not all Blue Hawaii's are BHSE's. Kerry's version is close to the BHSE and the original design run at full tilt is so as well in terms of spec.

I don't think the OPA541 could ever be used in anything meaningful but then again the Darkhole does just what the Dorkstar was meant to do but properly.

I understand pursuing it for your own knowledge and benefit but it's the production of boards that doesn't make sense to me.

Production of boards as in I will design it for my use and if somebody wants one, they can have it at cost. I only sell high carb items for profit...

[Knuckledragger]

Now the real question is, should I leave the circuit as it is or apply fixes so that it at least sort of works. Isolated heaters are a no brainer but I really want people to appreciate just how bad this amp is.

I'd apply any fixes in the areas of safety, reliability, and simplicity, but try to leave the bad sound as in tact as much as possible. The real question is can you emulate Ray's fugly case graphics?

[n3rdling]

The A10 is the worst amp I've ever heard so this doesn't surprise me.

[HeadphoneAddict]

Sorry for leaving you hanging like that but work called. Thanks Kevin for stepping in and stating rightly that it isn't wise to fuck with the Mafia. We've always known that the A-10 was bad but it is shocking just how bad is is. Ray states a voltage swing 600VP-P (roughly 210V RMS) which is just barely meets while generating way too much distortion so it is clearly clipping at that point. To put this into context the BHSE swings 1500VP-P (530VRMS) and it costs a lot less than the A-10. Hell,the SRM-252S swings 280VRMS and never comes close to clipping... hen we have the issue with no adjustments at all in the amp. Ray states on the top panel that you need matched ECC83's for the diff amp and 6SN7's for the preamp output but this will never work to balance out the amp properly. The diff amp has to be adjusted and ditto the phase splitter. There is also no offset adjustment so that 600VP-P voltage swing is purely theoretical. The amp will never reach that in practice...

To make matters worse these are not enough amplification stages so the gain is driven to the max with bypass caps on the cathodes so this thing is noise central. Massive amount of gain with no voltage swing is never a good idea. Then we have the over all layout of the amp, it's just a mess with stuff run all over the place, it's like auto routing gone mad.

As for the PSU, it is pure shit but Ray was clearly looking at Mikhail for inspiration there. Two bias lines run from the PSU to the headphones when a simple voltage divider in the amp chassis would have done a far better job. Mikhail did the same in the ES-1/2 wasting "precious" umbilical wires on nothing. Then we have the Cincon CMF40/60 switching PSU's used to run all the low voltage stuff from relays to the tube heaters. Grounding the heaters in the 5687 is a shockingly bad idea same as it was in the ES-1/2. Now the real question is, should I leave the circuit as it is or apply fixes so that it at least sort of works. Isolated heaters are a no brainer but I really want people to appreciate just how bad this amp is.

I don't see much of a point in trying to show how bad it is when you could improve upon the design and show how good it could be instead (like the ESX).

[HeadphoneAddict]

I should read the whole thread before posting, sorry...

[kevin gilmore]

So i spent about 5 hours last night trying to find a matched set of tubes so i could have a listen. With about 100+ 5687's

and an unlimited supply of 12ax7's, the best i could do was about 1 db differential and 1.5 db left to right channel.

So matching is just about impossible. And after an hour of running, the numbers changed.

Unlike the ES1/2 for which significant modifications can result in a pretty decent amp, for this mess, the surgery

has to be extreme. Take the output preamp tubes, and turn them into input gain tubes. Take the original input

gain tubes move them to after the volume knob, and apply local feedback from the output tubes. Then seperate

the filaments for the output tubes and lift them to +300 volts. It might actually sound half way decent this way.

[spritzer]

With that level of effort it would be better to just build a ESX mini with ECC81's and 6S4A's as output tubes.

I'd apply any fixes in the areas of safety, reliability, and simplicity, but try to leave the bad sound as in tact as much as possible. The real question is can you emulate Ray's fugly case graphics?

Sounds like a plan but I'm sure we'll draw up and test a fixed version just for kicks. As for the box, It really depends on the transformers used but I'll probably build the first one into a couple of die-cast aluminum boxes which need to the powder coated. I was thinking purple with glimmer or something crazy like that...

[kevin gilmore]

Wow, alex actually does exhibit some learning after the exstata debacle.

Well at least he knows better now where to acquire working schematics.

http://gilmore.chem.northwestern.edu/liquidlightning.pdf

For those that don't recognize it, its none other than a stax srm323

with mosfets in stage 2 and 3, and cavalli style current sources,

and a stacked current source in stage 2.

But for $4k+ ?? AH well the price of progress.

Now alex needs to go back to the liquid fire, remove the tubes, replace

them with jfets, just like the schematic he borrowed from about 1981.

[spritzer]

Amazing similarities but let's see how Stax did it in 1982:

and the 323A:

[n3rdling]

Do you guys also have the LL PSU schematic? If so, what do you think of it?

[kevin gilmore]

I'll do up the power supply later. I don't think much of the power supply.

The amp part has 30db less THD than the exstata. Way better.

about -80db second harmonic distortion across the full frequency

range, and a rising third harmonic that is -70db at 20khz.

Peak to Peak voltage really is close to 1350.

Some people are going to like the mosfet sound better than

the bipolar of the kgsshv. Nothing stops you from building

the kgsshv with mosfets however, other than the fact that

the parts are not isolated, they are a direct drop in for the 2sc4686a

Edited March 9, 2012 by kevin gilmore

[spritzer]

Yeah, Kevin has been working on a KGSSHV setup sourced exclusively from Mouser and that uses fet's instead of the BJT's.

[kevin gilmore]

The parts situation is a royal pain in the ass for sure, unless you had the foresight to buy lots

of stock of stuff a few years ago. Which some of us did.

The idea is to have all the parts sourceable from mouser.

On my kgsshv power supply, the ksc5026 replaces the 2sc3840 directly

The ksc1156 can replace the 2sa1486 with an added zener on the drive transistor, but

a better solution is the pin compatible (you have to flip it) Fqp3p50

The lt1021 is a digikey only part. For the amplifiers, when the 2sc4686a parts

dry up, the 2sc4686 parts will still be available, or you can use the same ixys ixta3n120

alex uses which is a special order part. Or the fqpf8n90c at +/-450

the lsk389 or the singles are available at another distributor and are standard production parts.

The 600 volt pnp is available as a 2sa1807 which is surface mount only, and

the adapter boards will be needed. Or the 2sa1413 which may be discontinued.

Alex's design is all mouser parts except for the 2sj74 which is discontinued

but still largely available. At some point these will not be available anymore

and then linear systems will really HAVE TO GET OFF THEIR FUCKING ASSES

and deliver the p-channel parts.

The bipolar version of my kgsshv is twice as fast (250khz -3db) than the fet

version (130khz -3db) due to significant gate capacitance of the output fets.

The n-channel input part is 30% faster than the equvalent p-channel input part.

Although you can go bipolar here too.

It all sucks i tell ya.

Unless you are Nelson Pass and can pay stupid money for custom

Static Induction Transistors which are n-channel only.

Edited March 9, 2012 by kevin gilmore

[Clarkmc2]

I find it encouraging that SemiSouth is actually interested in the "high-end audio" market for their devices. I had previously thought that the volume was much too low to put any effort into it.

Perhaps Nelson might be doing us a favor in the end after all. SemiSouth might make more devices available on their own at some point.

http://semisouth.com/archives/610

And yeah, those SITs were frighteningly expensive. I am no engineer, and I am still trying to wrap my head around how and why an SIT works. I read that the original Japanese manufacturers could not understand them either; they just made them to spec.

Edited March 10, 2012 by Clarkmc2

[spritzer]

Things are indeed getting bad but we can only hope that there will still be some demand for high voltage BJT's.

The IXYS part is available at Mouser, you just have to search for the TO220 version (P) and not the SMD (A). Naturally not available outside of the US...

[guzziguy]

Certainly that doesn't stop you. You have many friends here who would be happy to buy them for you and ship them to Iceland. I'm sure a certain amp designer does it all the time.

[spritzer]

Yup, I get regular care packages from Kevin.

[kevin gilmore]

And i get bottles of the black death in return.

And goodies from birgir's bakery, whose version of fruit cake

is just amazing. (not what we think of as fruit cake in the usa

which is generally useful for pounding nails and not much else)

Birgir calls the stuff "stolen" but stolen here is not anything

like that.

Its a great deal.

Testing indicates that to get the same distortion levels as the kgsshv does at 18 watts per channel,

you have to go to 38 watts per channel on alex's fet design. Or lower the power and suffer significant

THD at higher frequencies.

I completely expected this.

All parts now ordered for the preamp and blackhole.

Mouser must love me.

Edited March 10, 2012 by kevin gilmore

[DouglasQuaid]

And goodies from birgir's bakery, whose version of fruit cake

is just amazing. (not what we think of as fruit cake in the usa

which is generally useful for pounding nails and not much else)

Birgir calls the stuff "stolen" but stolen here is not anything

like that.

Edible fruit cake? Now I've heard it all. As for the amp project... It's like trying to build a Yugo from the ground up. Sure sounds fun, but when you're done who the hell would want to use it?

[spritzer]

Let's just put it this way, there is a reason for the pile of amps, dacs etc. in a spare room.

[wink]

He's gonna have a rummage sale....

[deepak]

Let's just put it this way, there is a reason for the pile of amps, dacs etc. in a spare room.

Build the Liquid Lightning please

[kevin gilmore]

So do we do a circuit board that is identical to the LL production version, or do

we do a modified version thats better? Off board or on board heatsinks version.

Really, taking the kgsshv board layout, the last 2 stages are identical, so

changing the front end it pretty trivial. And the kgsshv power supply works

great at +/-400 volts with only 2 resistor changes.

And its not a rummage sale... its a rubbish sale. Especially the rudistor clones.

Edited March 10, 2012 by kevin gilmore

[jdkJake]

Ah, stollen. I miss stollen. My grandfather (also a baker by trade, a Dutchman though) used to make us stollen. A far cry from fruit cake indeed. At least the way he made it.

Sounds like a more than equitable arrangement.