A DIY electrostatic amp for intermediate DIYers? - Page 4

I could fill a few posts with links to amp designs but many of them aren't suitable for a cheap amp like this. Most are laden with coupling caps which will drive up the price quickly (630v caps aren't cheap if you want something that isn't horrible quality) and some use different voltages which makes the PSU design harder on a budget. We could always just use an unregulated HV supply (like Stax does) but I'd rather have a more simple amp section and a more robust PSU.

The old SRX amp is simple enough but it has to be modified to allow bias for the tubes but it can be made to work with a lot of different tubes. The Tubecad amp is identical to the Egmont and you can't get a simpler circuit to build. There is the DC-coupled Gilmore amp with 6H30pi, 6S4A and 0A2's but it has proved to be tough to build and the 3 separate filament supplies quickly add up.

These amps are built on the cheap so that's not likely to happen. They are more test beds to compare the different PSU and part choices against the BHSE.

He was talking about an all out assault at the high end there so it doesn't apply. 1000V P-P should be more then enough but voltage output is only part of the issue. The Koss E/90 would be the best amp in the world if we only needed voltage swing.

None of you has commented on the circuit I posted on the previous page. I will take this as a most gracious politeness on your parts.

What are the issues besides voltage swings?

Someone posted a few of KG's requirements such as 100uV/s slew rates and 0.01% THD. And 1000Vpp.

It seems to me that trying to meet these requirements cheaply will not be easy. Possible, but not easy.

Any O/P device that is sweeping through 1000Vpp of its transfer curves is going to surface signficant non-linearities. The only tubes that I know about that are super linear over a wide voltage range are vertical deflection tubes and they don't typically operate at these voltages.

But aside from that, to minimize THD from a tube driving the load (even a no-current-required load like an electrostatic headphone) the plate load resistance matters. Which almost always means a CCS which will mean some number of HV sand devices.

Slew rates will require fairly fast devices operating at rich bias points so that they can provide the currents needed to drive the capacitances at the inputs of the next stage. These are not coupling caps, just the normal input capacitance of an active device. For example, if one uses a triode with gain for the output stage its driver has to pump the miller capacitance of that triode. This takes a low Zo and current delivery capacity and speed. If the O/P device is a mosfet the same issue applies. This is one of the reasons that KG uses the grounded grid configuration driven by a fet follower in the BH.

So, if I might enquire again, what are the real requirements for this amp? And what are the engineering tradeoffs you all might be willing to make?

a guy could cram a couple 10m45's in there p2p style if he really wanted.
This is where I think things get weird.
The outputs of the amp are going to be hanging out at like 300V above ground. This is safe for bias as it goes through a big fat resistor, but not for stators which are DC coupled without resistors to the output of the amp.

A cap would be required.

edited/added:
the amp should be able to swing as much as you mentioned, but it will require that the pot be swung all the way up... I dont care, and like it! but others have shown opposition to swinging past about 1/2 spin. Probably not as much an issue in the DIY community, but worth considering when the amps start to get sold second hand. scummy thing to think of, but life.

I find it very elegent in it's simplicity, and love the tube compliment, mainly as we are using the 6s4a on other designs. But what is the cost of iron, and what options are available for these windings? And, I don't see heaters - is that the 25v?

Then there are the other aspects to protect phones - we are talking about some using rather pricey phones with this amp. Usually cheap = risk.

As for tradeoffs, most here (excluding Spritzer) won't know how to answer - they just want cheap that works.

While I do know a switchable bias (have the boards on my desk, and will stuff most of the parts tonight), it simply cannot be done for less than $125, so it will NOT fit the design goals here. It is only mentioned as part of the tradeoffs people are willing to make - what solutions, and, at what cost.

That's my question as well. If it is cheap as chips, simple to build yet can out pace the Stax amps that's not easy to accomplish. The transformer will always be the largest chunk of the budget at up to 100$ and the large PSU caps aren't cheap either. I do like the R-core transformers available on ebay (100w units and the R80-04 unit in particular) for 55$ though they could do with more current handling.

If we were to put a CCS on the output tubes then it will be a massive improvement over the plate resistors but it will be almost impossible to do at this price point. Most of the third party amps sold don't even have a CCS like Rudistor, SP, Woo GES and the RSA A-10. That just goes to show how much of a rip off some of those amps are but for 300$ we would have to settle for resistors. As you rightly point out, it will make the performance into any impedance load suffer so the amp will not be an ideal choice for something like the SR-007 but should do fine with a Lambda or the ESP/950.

My tube of choice for the output would be 6S4A as they are cheap, can handle enough voltage for our needs and easy 6.3v filaments. Now there might be some Russian equivalents or some obscure TV tubes to try out too.

A fixed HV supply like in the Gilmore amps and then switched voltage dividers would do the trick. It's only two resistors so cheap as hell

I could possibly get the PCB's made at university, given the vector/.eps design and simple instructions of what is required. The actual process of getting them from press and peel to etched is fairly straightforward, I have used the method to make elements of scale models before, to great effect but that is where my knowledge ends.

It's also not that expensive to have them made at a board house if they aren't too big. The bottom line is though if we can agree on basic design parameters then we can nail down a schematic and have it drawn up.

How about a small push-pull tube amp with the ouput taken off the primary of the output transformer? John Broskie illustrates this approach in the Nov. '99 issue of "Tube CAD". Basically, he uses the first 3 sections of a Williamson amp. The input is a the first section of a 6sl7, the phase-splitter is the second half of the 6sl7, and the outputs are 6sn7 connected to the primary of the output transformer. Output to headphones is via coupling caps off the transformer primary. The weakness he cites in this approach is the impedance mismatch between the 6sn7s and the output transformer. How about sustituting the 6sn7s with 7189a? The 7189a can take 400v in triode mode and has a gain of about 20. Appropriate Hammond transformers run about $45.00 apiece. Coupling capacitors could be the exellent Russian 630v k-40 pio type. These are available on Ebay for a couple of bucks apiece if bought in bulk. Use a tube rectified power supply with a CLC filter. Total investment could be under $400.00, less if you've got a good junk box. Would this be a potential solution?

We could always just admit that it's not realistic to make a high performance stat amp for $300 and open up the budget a little... or at least include some options on the board for those willing to spend a bit more.

Gents, I'm figuring that if I keep posting concepts you will tell me what's right and wrong with them. Then perhaps we can find a cheapo design.

So here is Poor Man 2. Explanation below:



Simple and cheap is the name of the game. We know that the most expensive item will be the PS so the amp has to be dirt cheap. Some compromises might have to be made.

In Gilmore's BH he uses a dual jfet for the input diff amp. For two reasons: 1) to get good matching and 2) to maintain as equiavlent thermal drift as possible. But, these are n-channel devices whose outputs are at a positive rail. This requires him to use the current tunnel to translate the signals to the bottom rail. Very nice circuit, but lots of extra components.

If we use individual discretes, however, we can use PNP devices for the input pair. And if we select HV BJTs we can run these all the way down to the negative rail. In this case the ZTX560s are 500V devices. The only thing we need to be careful of is their power dissipation.

The are biased at 1mA each. With a 51k load their collectors are at -450V where they can swing the grid voltage needed for the EL34s.

The ZTX458 NPN BJTs are there as followers to provide a high impedance buffer for the diff amp so that we can get halfway decent frequency response. The 458s are only 400V devices so we actually have to find higher voltage devices for this slot or connect their collectors to a -300V supply which we can probably get from the -500V supply. Only a few mA are needed.

The BJT followers drive the tubes. Don't gag on the cathode bypass caps. We need them to get enough gain from the O/P tubes.

The amp has no NFB and its gain is over 1000. It can push 900Vpp. It has no NFB at the moment but has decent theoretical THD, less than 1% but more than 0.1%. It will be hard to do better than this without more sophisticated circuitry, but we might be able to tweak here and there. Frequency response into 150pf will not be as good as the BH, but it won't be terrible. I have not looked at slew rate. It might suck, but it might not be too bad either.

In addition, the CCS on the tail of the diff amp can be driven from an opamp servo so that the stator DC offset is 0VDC. Or we can just put an offset adjust pot on the CCS and hope that it won't drift too much. Also the 820R resistors can be split with a trimpot to equalize the output offsets. They will also drift, however, from thermal flucutations.

If this were actually built it would be good to thermally glue the ZTX560s together. Best we can do with two packages.

So, are we getting into the ballpark?

The first cavalli amp as shown outputs 300 vdc continuously. The roughly 560
volt bias has to be on top of the 300 volts. So 860 volts. Many people
get very nervous putting what is esentially an unlimited current at 300
volts on top of your head.

I did an amp almost exactly like this 10 years ago.

Otherwise you need output caps.

The second cavalli amp is similar to a BH except that it is grid drive.
For large voltage swings on the output grounded grid is much more linear.
But this will definitely work. The KGST which has not been published yet
is very similar to this.

Agreed on the higher linearity of the GG configuration. This mod could probably be added to the amp above.

Alex, that is quite good stuff... at least for me who has 4 quads of el34 laying around. Would this also work with kt77 (since I also have a quad of them)?

Except for the jfets, I have all that laying around, so I'm at about $5 additional cost so far... l

Seriously, the cheapest el34 are the JJ / Mullard / Electro Harmonix, all about $80 per quad shipped. Any way to keep this with the 6s4a, about $20 shipped? If not, it is what it is.

Would this require close matching of the tubes, or is there a biasing sequance, theoretically speaking?

As shown, the output stage is idling at only 8ma*2 Why not more? an EL34 can easily take it. At least 20ma 80ma @ 1000V (bridge rectified) is within the limits of inexpensive transformers.

Would using a long-tail pair for the output stage dump too much gain? It would save 2 caps & 1 resistor.

Perhaps the loss of gain in the output stage with a long tail pair&lower impedance plate resistors could be compensated for by reconfiguring the "driver" BJTs to provide a little gain.

Pabbi1, answers below. But, before that, here's a grounded grid version. It is slower than the original but can be sped up. Also. we should look into mosfets for the active devices.



In this amp we have relieved the need to replace the ZTX458s because they are no longer there and the 757s see much lower voltage. In fact, probably 150V transistors could be used in their spots which would lower the junction capacitances a little.

I don't see why KT77s could not be used. But haven't checked to be sure.

I started with 6S4 but in this configuration they create a higher Zo for the amp thereby reducing its speed. Which is why I moved to the EL34. But I'll look at them again.

Tube matching won't hurt. Tubes that are really far apart may be hard to balance. But with a balance control on the diff amp we can probably compensate for anything reasonable. Would you like to see the opamp servo?