The DIY Electrostatic amp thread

[FrankCooter]

I think R7 is out of place, too.

All R7 does is give some adjustment to R8. Maybe I'm missing something, but I don't see that it matters if R7 is either above or below R8. Together R7/R8 form the cathode bias resistance for the push-pull 7591 outputs. Unlike the usual arrangement of this type, R7/R8 do not set the plate currents in the push-pull pair, in this case the plate current is set by the 10m90 cathode resistor. Varying R7 moves the plate voltage of the 7591 pair up or down in relation to ground. Ideal plate voltage to ground (the offset) is 0V.

I originally tried a LTP as the first stage. It wasn't stable and would blow the regulator pass transistor, the associated zener string, and the filter capacitors of the positive rail. The regulator had no problem with a dummy load. Getting rid of the LTP cured the problem, but at this point I went with a simple CLC filter to avoid any further issues.

As drawn, your design won't work because it will put the DC plate voltage directly on the headphones. If you added coupling capacitors to the output it would be OK. The whole point of the bipolar HV supplies is to allow for direct coupled output by setting the plates of the output tubes at 0V in relation to ground. K.G. and Spritzer would be the sources for a better explanation of this type of circuit.

[chinsettawong]

Hi Frank,

Thanks for sharing your design. I might try to build it, if you don't mind.

Wachara C.

[spritzer]

Yup R7 and R8 form the adjustable cathode resistor which is used to zero out the DC offset. Same setup as is used in the Stax SRM-T1 and SRM-007t amps amongst others.

DC offset is the great enemy here as while the drivers are fine with it in theory, it kills the voltage swing. Single Power ES-1/2's had an offset of 100+V and the difference is quite startling once you get rid of it.

[dsavitsk]

All R7 does is give some adjustment to R8. Maybe I'm missing something, but I don't see that it matters if R7 is either above or below R8.

I was thinking that you were zeroing offset between the two plates, in which case R7 would need to be in a different place. But, yes, as you have it, you can zero it w/r/t/ ground, though you can't do it for both sides.

But, why does that matter? The headphones appear to be differential, don't connect to ground, and thus have no zero reference? Instead, they just need to sit some voltage below the bias.

As drawn, your design won't work because it will put the DC plate voltage directly on the headphones.

So, you'd either simply increase the bias to some voltage above the plates, or you could tie the B+ to ground and consider it a negative rail only. Something like the LL1668 as a plate choke will drop ~7V in this setup, which seems pretty incidental, and will keep the offset from bias pretty close to what you want.

[kevin gilmore]

The bias has to be +580 (for example)

referenced to L+ minus L- (or R+ minus R-)

So if you make -400 volts zero volts, make +400 volts 800 volts

then if the outputs are prefectly matched at 400 volts, then the

bias voltage needs to be 980 volts.

This has scare people in the past.

For 5 pin headphones, you really need L+ and L- and R+ and R- to be

within about 20 volts of each other.

Edited April 20, 2011 by kevin gilmore

[spritzer]

Nothing wrong with setting 0V as something else until somebody plugs in an Orpheus or a normal bias Stax. Bye, bye 5000$ drivers...

[chinsettawong]

Why will the drivers go bye-bye?

Wachara C.

[guzziguy]

Nothing wrong with setting 0V as something else until somebody plugs in an Orpheus or a normal bias Stax. Bye, bye 5000$ drivers...

Then I guess that we should call this amp "The Destroyer".

[FrankCooter]

Probably been discussed before, but any amp design with direct-coupled HV outputs makes me a bit nervous, even with a zero offset. Everything works fine in normal operation, but what happens in any given failure mode of the outputs? Is there a practical voltage limit (besides the bias) to the HV rails?

This project has encouraged me to try something a little more ambitious. I'm considering something similar but with direct- coupled 801A DHT outputs. This would require 600V rails. Any additional safety issues involved?

[FrankCooter]

Then I guess that we should call this amp "The Destroyer".

"Now I have become Death, the destroyer of worlds." - J. Robert Oppenhemer quoting the Hindu god Vishnu as he watched the explosion of the first atomic bomb.

I've built my share of "Vishnu" amps and would prefer not to build any more.

Edited April 21, 2011 by FrankCooter

[spritzer]

Why will the drivers go bye-bye?

Wachara C.

The voltage will overcome the natural insulation of the air inside the driver and burn up the diaphragm. I've seen this way too often on the normal bias sets which were used with transformer boxes. The later SRD boxes used 450V zeners to clamp the output plus the protection in the earcups but it just didn't always work. The phones will continue to work but it is a vicious cycle since the charred remains of the diaphragm shorten the D/S gap so the driver is even more sensitive to arcing so it happens again and again.

For the HE90, Sennheiser clearly states in the specs that they should never see anything in excess of 1000V P-P. That would be double the bias voltage (so 460V P-P for the normal bias Stax sets) but I think the SR-007 can take a bit more than that. They aren't really bothered by 900V bias and the driver construction sure points to that.

On this subject, it is amazing just how messed up ESP drivers can be yet still produce sound. The first Micro-Seiki MS-1 I received produced sound but it was very faint. When I opened them up there wasn't any diaphragm left, just small crumbles scatted all over the driver yet it made sound...

Probably been discussed before, but any amp design with direct-coupled HV outputs makes me a bit nervous, even with a zero offset. Everything works fine in normal operation, but what happens in any given failure mode of the outputs? Is there a practical voltage limit (besides the bias) to the HV rails?

This project has encouraged me to try something a little more ambitious. I'm considering something similar but with direct- coupled 801A DHT outputs. This would require 600V rails. Any additional safety issues involved?

Two things you can do, add a servo with the output running through relays which shuts off if the offset goes too high. The Stax SRM-717/727 and the T2 use this setup with the threshold at around 20V.

Another Stax trick, feed the output through 5k1 resistors to protect the headphones and the listener.

In reality this isn't much of an issue since the distances inside the headphones are so great that insulation breakdown should never be a factor. Stax did ground the SR-Omega chassis just to be sure but I doubt it was ever really needed.

[chinsettawong]

Hi Spritzer,

Thanks for your explanation. Do you think that the chance of arcing will be less if the stators are coated with insulation paint?

I always put a layer of insulation on my stators, and I have never experience any burnt driver yet. My highest bias voltage was 1.2KV.

Wachara C.

[spritzer]

It certainly won't hurt.

[15lyl]

It takes me almost 8 years to complete the KGSS, because due to the lack of 2sa1968, now I use a 2sa1413 instead and reduce the supply voltage to 345v.

When I test the amp with a CRO, there is an about 100 KHz sawtooth wave oscillation, I connect two 560pf caps. between B E of 2sc3675 in the third stage, the sawtooth gone. But when I listen with my Stax SRX mark3, there is still small squall oscillation; even I increase the caps.to 2nf. If I connect the caps. between 2sc2705 base and 2sc3675 collector, all high frequency gone.

Will anyone help?

[kevin gilmore]

well you should use the 2sa1486 instead of the 2sa1413 if you are talking about

the power supply. There are much better power supplies now.

The 560pf caps are definitely the wrong compensation in the wrong

place. Look at the latest kgsshv amplifier schematic here.

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

what you want are the 5pf caps between the (in your version) collector of the 2sc3675

and the base of the 2sc2705.

Edited April 28, 2011 by kevin gilmore

[15lyl]

Thank you for your quick response Kevin. 2sa1486 is not available in Australia too.

The noise was from the bias voltage. When my probe touched the bias, the noise disappeared, then I put a 5uf cap. after the 4.7M ohm, it has been fixed. Now I need to adjust and listening.

Where can I read your KGSSHV the whole article? It is not in the HeadWIze site now?

[kevin gilmore]

I think you are now in the right place for a few of my new projects

A 5uf cap after the 4.7M resistor is a BAD idea for O2's and most of the

latest headphones. I'm sure you can find a distributor that will sell

you parts and ship them to australia.

[spritzer]

Yeah, that cap is a very bad idea.

[15lyl]

ALL my problems are from my very old EI transformer, which produced a big voltage transient surges when switched on and killed 4 of 2sa 1968, now I need to put 2 metal-oxide-type varistors in the secondary.

I think I need a toroid transformer to rewire the power supply.

[pidesd]

hi frank cooter,

you said your amp had a pair of resistor loaded 6sn7 s...would it perform well that way? i have a few 6sn7gt(vt-231) lying around and thought i could put them to use.

could you provide a schematic?

-py

[pidesd]

birgir,

are the values the same as frank posted in last page in your 6as4 edgmont? is it just a tube substitution?

[spritzer]

My 6S4A is very different from what Frank did, it's more like the original with a 47K plate resistors for the 6S4A's and a +/-350V PSU.

[FrankCooter]

hi frank cooter,

you said your amp had a pair of resistor loaded 6sn7 s...would it perform well that way? i have a few 6sn7gt(vt-231) lying around and thought i could put them to use.

could you provide a schematic?

-py

My amp used CCS loaded 7591's, not resistor loaded 6SN7. If you want to use 6SN7, it's probably best to stick with the original Egmont schematic that Spritzer posted.

[spritzer]

I'm with you there Frank, I see little sense in CCS loading the 6SN7 especially if the goal is to build something cheap. I am itching to build a simple 7F7/7N7 version some day though and cram it in a tiny chassis.

[FrankCooter]

For something simple, cheap, compact, with a mu of about 20, and the ability to handle 14w at 400V, a triode connected 6P14P-EV (Russian 7189/EL-84 eq.) looks very promising. Guitarists love this tube. It's built like a little tank. Out of a dozen I bought off Ebay, I put together 2 quads that biased withen a couple of mA. I'm working on an amp right now that's a further simplified version of the 7591 amp. Total cost under $200. No CCS, but 400V rails with 20mA on the finals, and a gain on paper of about 700. Solid state power supply based on an Antek toroid, a couple of Radio Daze $12.00 chokes, and some most likely counterfit Panasonic electrolytics. Chassis as tight and compact as possible. Sort of the AK-47 stripped to basics approach to functionality. "Staxes for the masses!" Sometimes it's as much or more fun to try to build something out of next to nothing as an "over-the-top" fantasy. And if it sucks, it's no great loss.