Now for something different...

[kevin gilmore]

spice models on tubes are notorious for incorrect output voltage swings.

With a B+ of 300v, a plate resistor of 2.5k, the maximum voltage

the tube can possibly do is 225VPP. And that maximum is theoretical

because the tube is going to be hard saturated at the bottom, and

the linearity is going to be crap.

with a plate resistor of 1.5k its even worse, under 200VPP.

Voltage gain this way will be under 2.

Its all about the Rp of the tube, in this case spec is 800 ohms, but

will quickly rise to 1k+ as soon as you start putting signal thru it.

[spritzer]

It's a good approach to discard any recommended setup when dealing with electrostatic output stages. The engineers were not thinking about these requirements as we only care about low distortion voltage swing.

[Remolon]

Kevin, you are right. Better go back to the past and work with the plate characteristics curves. If I do so:

With B+ = 300VDC and Rp = 2K5, for a good response the bias should be -20V, and the input consequently limited. Then Vpp would be 130v (99dB for the SR-007?)

[morphsci]

[nikongod]

130vp-p isnt that little. You could do a little better, sure, but if you think there is an interesting reason to try it, try it. There is more to life than 15Kvp-p voltage swing....

You may like the results from using the 12V filament supply for the input tubes as a -12Vsupply and running a simple SS CCS to it. It saves a tube and a few caps...

Why 12ax7?

You have one negative feedback point indicated, but nowhere for it to go.

If its going to run open loop, and you will only get ~130Vp-p out of it at most, why not just use 12au7 for the inputs? In a long tail pair you can get freakishly wide bandwidth out of 12au7 and it should drive the 2a3 nicely.

Maybe one of the "middle ground" tubes - 12az7 has always intrigued me for high gain and low Rp, although I have never used one.

Wire the cathodes of the output tubes in parallel directly on the schematic. Skip those resistors between the 2 of them. It wont happen without about 5 heater supplies anyways...

[spritzer]

You are talking about max voltage swing here which is far, far from being linear. If you are aiming to design then worst electrostatic headphone amplifier ever then it's the right track but otherwise I'd start from the beginning.

[Remolon]

The discussion has been very helpful. Thanks.

I will make the initial trial with the only change of the bias, it will be -20V

[Remolon]

R45 is useless but Rk4 and Rk5 facilitate the use of not matched 2A3 tubes, adjusting the bias individually.

I changed all the input tubes. I will start with 6DJ8, and the following values for the other components

Rload, Rf 10K

R1 27K4

Rk1, Rp1 825R

Rg1, Rg2, Rg3, Rb4, Rb5 221K

Rk2, Rk3 100R

Rp2, Rp3 47K5

Rg4, Rg5 1K

Rk4, Rk5 10R

Rp4, Rp5 2K5 (5W)

C2, C3, Cp2, Cp3 1uF

B+ 300VDC and Bias -20VDC

[Remolon]

I will make the first trial not with the 2A3 but with the 5687. I have not found a reliable Spice model for 5687, so I am working on a graphical solution.

With a voltage supply of 400VDC, a plate resistor of 10K and a grid bias of -8 volts, I have calculated a idle current of 20mA and a maximun output of 220Vpp. I will test a fixed bias from the 12.6 volts filaments winding and a cathode resistor of 402R.

I have ordered the custom-made power transformer, and have finished the PSU, the input differencial amplifier and the stepped attenuator. Here is a picture (the differencial amplifier PCB without sockets and valves).

Depending on the results, I will test a different tube looking for more volts, more current or do nothing.

[Grahame]

This post on el reg appealed to me for some reason today ,,,

Re: I will laugh so much...

ReBirth is sounded pretty dated now, though. There's far more convincing virtual analog synths around; although ReBirth ran nicely on a 60MHz processor and was amazing for its day. Which was 16 years ago. I would expect some future TB-303 clone to be running a SPICE simulation of the actual circuitboard*

And that user interface? What were they thinking? Yes to emulating the TB-303 sound, no to emulating the TB-303 pattern editor

* OK, admission time, I tried this. You have to run the simulation at 1MHz in SPICE to capture the oscillator triggers, and there are enough non-linear components that you have to use something like conjugate gradient to solve it, and that puts it out of the 1MHz sample rate even on a 3GHz machine. You could possibly run the filters at a lower sample rate and analytically create the input waveforms though - the SPICE waveforms, even with non-perfect components, are pretty much mathematically perfect. But I digress.

[Knuckledragger]

I remember futzing with an alpha version of ReBirth in late '96. As much as I dislike softsynths, I'd love to see a modernized version of it. Propellerheads EOL's RB338 quite some time ago and folded it into Reason. There's also an iPad version. Neither is really satisfying. It'd be nice to have a real 303 emulation option with a waiting time slightly less than that of an x0xb0x.

[spritzer]

Any of those should work. No BOM as we never do one and we just need schematics for P-P amps.

Speaking of something different, here is something I've been working on as an attempt to burn off some of my surplus parts:

It's a reworking of the Stax "A" circuit with a diff input. Really needs feedback but not enough gain...

[Remolon]

Does this graph represent the operation of the schematic?

Is the operation point Va=400V Vg=-18V and 4.5mA?

What is the maximun output Vpp, and with how much distortion?

[spritzer]

This is still work in progress and I have yet to optimize the circuit for the 6SN7 (supposed to use 12AU7's) so the values reflect that. I will never be able to run the tubes to the max give the constraints of the transformer I'm using. Running the tubes at 6mA is tempting (what I did with the B-10, max rating for the tubes) but the transformer can't sustain that so I'll probably run at 3-4mA to keep it nice and cool.

I'm also not sure about the cathode bypass on the second stage as it's probably more trouble than it is worth. The amp is currently hovering around 50dB worth of gain which is enough for me with 7.2VRMS available from the source.

It will be north of 1000V P-P but the distortion will be significant.

[Remolon]

Sorry for the prophet, but I consider the Stax Model "A" as a more qualified candidate to be “the worst electrostatic headphone amplifier ever...” displacing the one I had proposed:

This is a generic circuit which accepts different valve combination.

The choose of the right combination between 6dj8, 12at7, 12au7, 12ax7 and others for the input differential amplifier will provide an adequate gain. In my case, I am using 6dj8 for V1, V2 and V3 to have a gain of 25dB.

As I said, I am making the first attempt with the 5687WA for the output. If it works as expected, I will considered other candidates for the output tubes depending of what is needed, more amplitude or more current.

If the election is the 6SN7-GT and a power supply (B++) of 600V, I would not use the values on theModel "A" schematic. I think it is better to work with other load resistor and bias; as a starting point, this one:

Rp4 = Rp5 = 30K ( referred to my schematic) and Vg = -10V, fixed or with Rk4 = Rk5 = 1K0. Theoretically, Vpp = 320 (450-130), with less distortion and a little more that 100dB for the SR.007 (off course, the 6SN7 admits much higher plate voltage but I would prefer not to exceed 450V).

My choice would be for fixed bias, separately adjusted for each valve, and Rk4 = Rk5 = 10R for an easier control.

[spritzer]

Lower plate resistors are a bad idea for electrostatics. 50-70K is the golden range.

[kevin gilmore]

Lower value plate resistors saturate the output tubes in a hurry.

Try it and you will see. Constant current sources way better

And lots more parts

[Remolon]

I would appreciate numerical values specific for this case. The datasheet specifies a cathode average current of 20mA and a maximum of 70mA, and the calculated values with a plate resistor of 30K and a voltage supply of 600V are, obviously, even lower than the average of 20mA at any point of operation.

Probably there are other reasons I am not aware of them. Any indication is welcome.

[kevin gilmore]

The real way to look at this is to calculate what the Rp of the tube does when the

tube is hard on. For example for a 6ca7 in this kind of circuit, a 30k plate resistor

on 600 volts ends up with about 70 volts at saturation across the tube. You can

calculate what it is for different tubes. But the last 20 volts before saturation is

highly non-linear, and with a 600v supply, and 30k i would be surprised if the

peak to peak output voltage would be more than 300 volts with <1% thd.

I'll calculate more when i can get to a real computer later.

For mikhails es1 disaster with the output tubes already biased to -150v, and 50k plate

loads, the maximum peak to peak voltage (stator to ground) was also less than 300v

on a 800 volt power supply.

[Remolon]

It is what I had calculated, 320Vpp, with the tubes biases to -10v and a plate load of 30K, as you can see in my previous post. That is 113Vrms, 101dB for the SR-007, with much lower THD that the proposed Stax Model “A”.

The bias to -150v is a big mistake, I think that the 6SN7-GT should be biases between -10v and -12v, -8v if not much output voltage is needed, but in no case beyond -14v.

[Remolon]

Eppur si muove

 

I have already finished and tested the amplifier. Built with two 5687WA, for the output push-pull, three ECC88, for the differential amplifier, and fixed bias, around -8V. It works very well.

 

I am using it with a pair of SR-407, and with the potentiometer at 11 the SPL is enough for my listening habits.

[deepak]

I'm interested why you guys have decided to do a 300B electrostat amp after all. From reading old posts on Headfi I got the impression that it would be a poor idea.

[spritzer]

Ohhh it is a poor idea. Building the B-10 was a poor idea but I did it anyway. The 300B is linear but it is limited in terms of voltage so you need output transformers. Costs are starting to mount to say the least at this point.

I'm doing it though as I like to build crazy things and I have some 300B's here. This is my version of spring cleaning...

While on the subject, an even crazier amp design crept into my head this morning which will be even more expensive and ludicrous.

[wink]

"While on the subject, an even crazier amp design crept into my head this morning which will be even more expensive and ludicrous. "

Let me guess...

A solid state version of the DIYT2 using using series germanium sand with oil-filled paper & teflon caps.

[spritzer]

I don't think there ever was any Germanium sand that could handle 1500V... My plan is for a spud electrostatic amp with just two triodes and DHT's at that. I know ETF.12 had a competition along these lines but I'm going to be far stricter to myself. No push-pull, current sources, parallel tubes etc. Just transformer, triode and transformer...