Megatron Electrostatic Headphone Amplifier

[Shawn]

10 hours ago, JoaMat said:

Nice idea. If you can implement an auto detect function as well – like Elekit has done with TU-8900. 

Power module…

It's a great idea. I’ll likely move forward with the current plan for now since I’ve already ordered the PCB prototypes, but I definitely want to explore how that auto-detect function works.

My guess is that it detects the difference in cold filament resistance between 2A3 and 300B tubes, but I’m not exactly sure how it’s implemented in detail. I’ll need to dig into that more.

Also, from the image you shared, you reminded me that I should probably add a few small ceramic caps for filtering (like C9, C8, C7 on the right side of U1). Thanks for the heads-up.😅

The test PCBs should arrive this weekend. If I’m able to populate and test them successfully, I’ll report back with results.

BTW, do you know what kind of DC-DC chip on the picture? Too blurry for me to read.

[JoaMat]

I don't what chip it is. 

Marking might be???? Very hard to see.
26QH
288
A3HL

Including in curcuit is also dual comparator 2093B and  .05R sense resistor.

Good luck and looking forward a report.

[Shawn]

11 minutes ago, JoaMat said:

26QH
288
A3HL

Find it. I will share the info here anyway if you are curious. The switching freq range is between 900 to 2400(kHz). Lower than the LMQ, but shouldn't have any problems.

https://www.ti.com/product/TPS62913

[JoaMat]

That was fast! Thanks a lot.

[simmconn]

It’s not hard to do a 1:1 clone of the PCB. It’s also a great interview question for a junior electrical engineer to come up with a circuit that does this. However not a lot of them would probably know how a vacuum tube filament/heater behaves.

[Shawn]

Good news: It works!
Measured output is 4.9V for the 5V setting, 6.1V for 6.3V, and spot-on 2.5V for the 2.5V setting.
I haven’t tested the soft-start function yet — if anyone has good suggestions on how to choose a suitable load resistor for testing, I’m all ears.

Bad news: I spent two days figuring out how to hand-solder with a heat pad and heat gun the QFN package. Even when using a stencil, too much paste tends to remain on the pads, which makes them very prone to shorting.
As a result, I ended up ruining two boards and some components. At this point, I really don’t feel like assembling the remaining three boards myself. I think it’s best to leave the rest to a proper assembly house.

Updates:

Soft start works. About 3.5ms rising edge time for 6.3V.

Edited June 9 by Shawn
Updated for soft start testing

[sorenb]

3 hours ago, Shawn said:

if anyone has good suggestions on how to choose a suitable load resistor for testing

I'd use 300b or 2A3 tubes as load
be aware that NOS tubes may have different results from new production tubes

[sorenb]

7 hours ago, simmconn said:

It’s not hard to do a 1:1 clone of the PCB.

Maybe harder than you anticipate
Recently I was sent a not working Blue Hawaii BJT.
Turns out the PSU PCB is a clone of Kevins layout ... just flipped upside down, with components firmly mounted on top of the traces ...a year down the road, it went boom, with lifted traces and some literally burned totally.
@kevin gilmore  has been kind and provided a modified version of his layout to match the mounting studs in the chassis.

 

[JoaMat]

Thanks to @Shawn, who found the device used in Elekit filament supply module, I now have a working DHT switched mode filament regulator ”prototype” (in red circle).

Board size 23mm x 15mm (0.9in x 0.6in)

Soldering looks awful, but remember my eyes are 69 years old and my hands are not that steady.

Edited June 25 by JoaMat

[Shawn]

9 hours ago, JoaMat said:

I now have a working DHT switched mode filament regulator ”prototype” (in red circle).

Great to know that. You could further reduce the PCB size by using a compact bridge rectifier like the CSPB40K-HF. It has a much smaller footprint compared to standard rectifier packages and still handles decent current.

Like the way how you fix and test the 2A3 tube.

 

9 hours ago, JoaMat said:

Soldering looks awful, but remember my eyes are 69 years old and my hands are not that steady.

Recently, I purchased these soldering tweezers from JBC to perform SMD work. Nice and decent.

[JoaMat]

 I used a Mini Hot Plate when soldering the buck converter and inductor. First a thin layer of solder and then up on the plate. 150 degrees Celsius for a minute before up to 250. Rest of components are hand soldered.

Thanks for the bridge tip. I have some small Schottky bridge rectifiers (CDBHD240-G) I intend to try first.
 

[JoaMat]

With help of above Mini Hot Plate I managed to build another three small regulators.

Here they are on a “motherboard” with rectifiers and electrolytics to achieve 12VCD. The trimmers are for adjusting output voltage, 2.5V for 2A3 and 5V for 300B.

Now I want a regulator for 6,3V and 7.5V. The buck converter above is limited to 5.5V, so might try LMQ66430. @Shawn What inductor did you use with LMQ66430?

[Shawn]

11 hours ago, JoaMat said:

What inductor did you use with LMQ66430?

I used an inductor from Coilcraft, MPN: XGL6020-222MEC. The manufacturer claims it has the lowest DCR in the industry for its class, and it’s also very compact at around 6.71 x 6.71 mm.

For the output voltage selection, I set it up with a few jumpers to switch between different resistor values for 2.5V, 5V, and 6.3V outputs.

If you need help calculating those resistor values, just let me know.

Also, I’m curious. When you switch from 300B to 2A3, did you modify any surrounding circuit parameters, or can you drop them in directly (aside from changing the filament voltage)?

Edited July 7 by Shawn

[JoaMat]

Thanks. XGL6020 inductors are probably excellent. With TPS62913 I use Wurth 2.2uH 4.4 x 4.1 mm, which works all right.

I don’t change anything when switching between 300B and 2A3 (except from filament voltage). Both tubes need 3.6K cathode resistors @20mA. EL34 needs 1.6K and EML 20B needs 0.6K.

I use small daughter boards with resistors and jumpers to change cathode resistance.

[Shawn]

20 hours ago, JoaMat said:

I don’t change anything when switching between 300B and 2A3 (except from filament voltage). Both tubes need 3.6K cathode resistors @20mA. EL34 needs 1.6K and EML 20B needs 0.6K.

I’m planning to use ±400 V for the supply with an anode current around 23 mA. I noticed that on Kevin’s PCB layout for the 300B, the cathode resistors are shown as two 3300 Ω parts. Do you think I should replace those with 3600 Ω instead?

Also, for filament voltage switching, I’m using the pin header setup shown below(MPN: 0010897120). Each row corresponds to 2.5V, 5V, 6.3V, and 12.6V, and I also left one extra row reserved in case I need it in the future.

[JoaMat]

I’m using a single cathode resistor. When you are using two 3300R resistors, as on your board, I guess that is equivalent to one 1650R. If you follow silk screen you will get -100V or so on output. You probably need to double resistor values. Maybe someone out there can tell more about this.

[Shawn]

On 7/9/2025 at 1:34 AM, JoaMat said:

If you follow silk screen you will get -100V or so on output.

Last night I did the first power-up. I used some very inexpensive 300Bs just in case anything went wrong and to avoid risking the expensive tubes.

Balance was actually very normal. About 5 V on one channel and 3 V on the other. But the offset was around –100 V on both channels. I’m guessing this is the “–100 V output” issue you were referring to? Let me know if that matches what you’ve seen.

I haven’t had time to swap the cathode resistors yet. Measured Vgk is about –42 V, which is clearly not deep enough. I’ve ordered some 7.5 kΩ resistors, which after calculation should bring Vgk closer to –90 V.

Also, for anyone interested, here are the LMQ66430 output voltage setting resistor positions I used on my board:

M20-8760542 is a header pin jumper.

1–2: reserved positions

3–4: 6.3 V

5–6: 5 V

7–8: 2.5 V

9–10: 1.5 V

Hope this helps anyone planning similar setups. 

[JoaMat]

 

-

Edited July 10 by JoaMat

[JoaMat]

1 hour ago, Shawn said:

I’m guessing this is the “–100 V output” issue you were referring to? Let me know if that matches what you’ve seen.

Yep, that's what I've seen.

[Shawn]

Today the 7.5 kΩ resistors I ordered arrived. To swap all eight cathode resistors I pretty much had to remove every screw on the chassis.

Long story short, the offset is much improved, going from –100 V to about +10 V.

Big thanks to @JoaMat for the help!

For me, this voltage is acceptable, but if someone wants to get even closer to 0 V offset, I think 7.0 kΩ might be the sweet spot.

Also worth mentioning: the G-grid adjustable pot on the board offers around 2–4 V of Vgk trimming range.

My 300Bs are running at ±400 V, since I feel 450 V is pushing things a bit too much.

I think the main reason(Correct me if I'm wrong) for the original –100 V offset was that the cathode resistance was too low, allowing too much current. This made the anode current too high, and because the EL34 CCS isn’t perfect, it compensates by lowering its voltage output. With the original 3300 Ω resistors, I saw the B+ drop from 400 V to 370 V under load.

One more note: after increasing the cathode resistor, Vgk sits around –80 V. For safety, I’d recommend replacing the original 220 uF/100V cathode bypass caps with 160V rated ones.

More pictures will be posted later.

[JoaMat]

How much offset can you trim away with the trimmers? I guess one need about 15K to get rid of +10V offset (with 2 x 500K resistors in the voltage divider).

[kevin gilmore]

so more than a few people are getting clobbered with 300b that don't even come close to the original published western electric document.

of the 7 different varieties i have, only the original 1960's manufactured western electric comes even close to the spec. and the new western electric tubes come in low vgk and high vgk versions. (The B series tubes)

just means you have to mess with the resistors. and yep its a major pain in the ass.

the el34 original version pretty much worked +/- a few volts no matter what tube you stuffed in there as long as they all from the same manufacturer.

or build a megatronxxl  now known as optimus-prime.  more money, but less of a balance problem.

 

[Shawn]

6 hours ago, JoaMat said:

How much offset can you trim away with the trimmers? I guess one need about 15K to get rid of +10V offset (with 2 x 500K resistors in the voltage divider).

The grid’s pull-up resistor is R5 (680 K), and the pull-down side is two 500 K resistors. The 10 K trimmer (pot) has a very small effect, just about 1–2 V of offset adjustment range.

With the current 7500 Ω cathode resistors, Vgk is roughly –76 V. To reach the WE datasheet’s recommended value, the cathode resistance would need to increase further, but that would push the offset even farther out of range.

So the grid trimmer really needs to have more authority. The 10 K value is clearly too small here; something like 50 K or even 100 K would likely be a better choice.

I’ve noticed that at this: 75 V operating point, there’s slight low-frequency distortion at high volume. I’m not sure if that’s because Vgk is too negative (too “deep”) or not negative enough.

2 hours ago, kevin gilmore said:

just means you have to mess with the resistors. and yep its a major pain in the ass.

Yep, it’s a real hassle. At the very least, having to remove and reinstall the PCB every time is incredibly annoying.

[kevin gilmore]

probably easier to change the cathode resistors on the current sources.

[Shawn]

10 hours ago, Shawn said:

I’ve noticed that at this: 75 V operating point, there’s slight low-frequency distortion at high volume. I’m not sure if that’s because Vgk is too negative (too “deep”) or not negative enough.

A quick update on this build.

Regarding the low-frequency distortion at high volume that I mentioned earlier, I did some troubleshooting.

First, I checked the GRHV output for the +300 V rail. It's extremely stable and consistently holds at 300 V, so it doesn’t seem like any current limiting is kicking in.

By chance, I noticed that the D3 LED (next to the 2SA1968 and I'm using STN9360 in this spot) flashes during loud low-frequency passages. That seems to suggest the current mirror may be going into over-saturation.

This might be related to the two 150 kΩ resistors labeled R8A101 and R8B101 in the schematic. In the original Megatron, there is actually just a single 150 kΩ pull-down resistor instead of two. That difference might be causing the issue.

Also measured the anode voltage of the 12AX7 tied to the current mirror. Under loud, bass-heavy signals, it shows a clear voltage drop from about 355 V down to 345 V momentarily.

I’m planning to try shorting one of the 150 kΩ R8 resistors to see if that improves things.

Updated:

Turns out the current mirror itself was fine. The culprit was actually one of the PSVANE 12AX7 tubes.

I suspect there may be a reason these were on sale at TubeDepot. At least the one I got had issues under load.

Anyway, swapped it out and the problem is now completely resolved.

 

Edited July 13 by Shawn
Issue fixed