kevin gilmore

my opinion is that it is very close to a 717. definitely sounds different from a 727.

goldenreference.PDF

new division. *schiit

maybe there are 2 other caps hiding in there. pretty much have to be. max gain for a 6sn7 is 20 and max gain for a 6sl7 is 70, so 1400 would be max voltage gain at 6% distortion. cut it down to a gain of 500 and you are still at 2% distortion. with no feedback frequency response is going to be much worse at 20khz, like 6db. the resistor current sources make it worse. stupidly poor design that seems typical of schiit. kind of wonder that if they ever did make something really nice, people would notice.

there is only one coupling cap per channel, so the circuit may be slightly different than shown. plate resistors instead of cascade current source is pretty lame these days. Completely open loop with no feedback is the result of not enough overall gain, and this is going to make the thing highly unstable. at least they used Justin's stax jack. The only piece of quality in that mess

my favorite power supply is the GRLV, it outperforms everything I have seen by a large margin.

you probably want the version with the on board servo then. my guess is that the input impedance is something like 10k now that the lsj and lsk dual parts are available, a 2 channel version would be possible with a much higher input impedance

so all of that (the last pic) is that just the headphone amp, or is it also used as the output buffer for the xlr/rca way goofy if its just for the headphone. and I really don't like digital volume control.

I have not listened to the 1541, waiting for a USA distributor. the lower end unit is probably similar to the denafrips unit and the same price. Both are likely very good, Birgir has the denafrips and the original soekris boards

output buffer may be a diamond follower, hard to tell. rest of it after the dac certainly looks like stacked current mirrors. certainly a complicated design similar to what you see inside integrated circuits. obviously assembled with high resolution pick and place machines. much better built than a lot of stuff I have been seeing these days. Now a year or 2 to make the software perfect and the thing will be gold. no sanded off parts either like the denafrips, so they are proud of their work. Only one board error with that hand soldered diode. nice stuff and an indication of whats coming.

Lucky that birgir does not have a basement, otherwise the collection would be much bigger and heavier. I happen to know someone with this problem. however he does cheat, no way he can put his car in the garage. last I knew there were several hundred pounds of transformers where the car should be.

one comment about the tubes, have tried many tubes on testbench, 12au7,12ax7,12at7,7308 etc you need a mu of at least 33 and a transconductance of at least 12500

the card edge sockets were .156 spacing on everything I have seen. The connector for the aux input slot is the same thing.

More difference in the tube between 100 and 150v

The VAS and output stage are from the 717. Resistor pullup on the vas stage. The 727 had a current source that replaced the resistors and really was better once you fix the local feedback. I'm sure that the stax mafia can make retrofit current source boards birgir's thoughts on the sound are exactly the same as mine, the high end is rough and screechy also all the extra connections on each of the plugin boards including the servo are going to get noisy over the years. very cheap connectors, not the augat that I use. also kind of wonder if they really are rated for voltage

per channel the -15v current is around 8ma tubes, 6922 and 7308 come to mind. there are also some subminiatures the change in sound with higher voltage is VERY subtle...

raising the plate voltage to 150 puts the tube in a more linear transconducance region. use stn9360 or 2sa1486 and there is plenty of clearance. in other news, going to 24V instead of 15V for the jfet input based versions of everything also makes the front end amplifier slightly more linear

The idea was a 7 pin header with one pin missing, or a single pin header might rethink that originally I was thinking of soldered in right angle pins

actually chinsettawong's hand built things have better looking drivers I just shipped birgir a latest set of kingsound garbage, why he buys this crap, no idea

the opamps are strictly the servo's, not sure there is any reason to change that part of it. the diamond buffers are strictly the output buffers, unity gain, the servo's are on the main board

Put a heatsink on the back of the buffers

the resistors do waste about 15 watts total of heat. so if you really want it completely right, +400 (or 450) -400 (or -450) +100, +6.3 (x 2) and -15 and then if you want the TS version, floating ac filaments for the output tubes. Done right, its a lot bigger than the original.

I can do that so on the t8000, there are places to install jumpers, evidently because stax was not sure of what tubes they would be able to get, and so they setup the jumpers to do things like 12ax7. Problem is that (waiting on final confirmation from birgir) for the 6922, the 2 x 6.3v filaments are wired in series, so not only do each section of the tubes have to be super well matched, but filament current between the 2 tubes also has to be matched, and the common connection (its a voltage gain of 1000 remember) increases cross talk. so 2 x LDO regulators, set at 6.3V each

standalone grhv for 100v and regulated filaments

12 inch and 10 inch drivers? similar to what Wilson does?