My upcoming project is a dedicated amp for my full range ESLs. It is well known that the step-up transformer with its high xfm ratio (1:50 -- 1:100) is the weakest link in an ESL. OTOH, the required voltages are too high for direct drive at reasonable SPL, so I am going for the next best thing -- minimal output transformer ratio and fairly high B+ in the output stage.
I will start building the contraption very soon now, starting with the PSU. For now, I would like to invite your feedback on the audio part. A short overview: the output stage is GM70 AB1 fixed bias, B+ of 1200...1500V. To keep it simple it is RC fed from a 6N6P driver; I chose this tube because of its low rp and way higher mu than, for example, the 2A3. The first stage is also an LTP doing the phase inversion and giving some extra gain. The tail CCSes will be solid state, exact topology TBD.
Note: the 4uF smoothing capacitor is just local bypass. The main reservoir cap is 100uF for both channels.
Thanks for your comments! I will be posting the build progress in this thread, too.
Kenneth
I will start building the contraption very soon now, starting with the PSU. For now, I would like to invite your feedback on the audio part. A short overview: the output stage is GM70 AB1 fixed bias, B+ of 1200...1500V. To keep it simple it is RC fed from a 6N6P driver; I chose this tube because of its low rp and way higher mu than, for example, the 2A3. The first stage is also an LTP doing the phase inversion and giving some extra gain. The tail CCSes will be solid state, exact topology TBD.
Note: the 4uF smoothing capacitor is just local bypass. The main reservoir cap is 100uF for both channels.
Thanks for your comments! I will be posting the build progress in this thread, too.
Kenneth
Attachments
Last edited:
It seems to me that the problem with OTs are high Z windings, rather than a high Z ratio. The Hi Z windings having a lot of distributed capacitance. By using a Hi Z/Hi Voltage B+ final stage, this ends up doubling the problems. Maybe for an autoformer type step up this would make sense, but that has HV DC B+ issues on the output. Perhaps an autoformer xfmr with 0 V on the center tap, and -HV on the output tube cathodes instead.
Last edited:
Or a cathode follower or circlotron circuit. In any case you need the OPT first. Custom iron is outside of my budget. I've got some 29KQ6/PL521's for a direct drive ESL amp project but life is geting in the way.
Jim
Thanks smoking-amp,
I would expect that any secondary-winding parasitic capacitance would have less of an impact in my transformer, because of the much lower step-up ratio. In a 1:50 tranny, each pF of secondary capacitance is reflected as 2.5nF at the primary; in my case, only about 16pF. Agreed, my primary winding itself will have much more parasitic capacitance.
Anyway high frequency extension has never been a problem for me in regular ESL transformers. IME it only starts limiting the bandwidth at well over 50kHz. So I'm hoping it won't be a problem here either. I will have to wait and see. I chose the GM70 partly for its low plate resistance, hoping it will easily drive any capacitive reactance.
Kenneth
I would expect that any secondary-winding parasitic capacitance would have less of an impact in my transformer, because of the much lower step-up ratio. In a 1:50 tranny, each pF of secondary capacitance is reflected as 2.5nF at the primary; in my case, only about 16pF. Agreed, my primary winding itself will have much more parasitic capacitance.
Anyway high frequency extension has never been a problem for me in regular ESL transformers. IME it only starts limiting the bandwidth at well over 50kHz. So I'm hoping it won't be a problem here either. I will have to wait and see. I chose the GM70 partly for its low plate resistance, hoping it will easily drive any capacitive reactance.
Kenneth
Jim,
It depends. My OPTs are underway, I didn't pay an arm and a leg for them, considering the requirements. One can always pay more for exotic cores, name tags, or silver wire, but reasonably-priced custom jobs do exist.
I'd like to advise you not to pursue that direct-drive amp, unless you have non-full range ESLs with very narrow D/S spacing. Otherwise there's just not enough headroom to get good dynamics. On top of that, the DC feedback loop is likely to be a pain to get right, and there's the safety concern, too. Finally tube life might be short, those deflection tubes were not designed for continuous high plate voltages.
I thought a long time about direct drive but in the end I just abandoned the whole idea (at least for my situation, i.e. full range).
Kenneth
I have yet to year a pair of ESL's that I really love. I know very little about them or the requirements to drive them properly (expect that everyone says they are really hard to drive). I'm sure part of the reason I've never really loved them is they probably weren't being driven properly.
Would you mind offering a "Dummies Guide" to what you want in a good amp for ESL's? Someone that's going to the trouble to build their own specialized amp obviously knows what they need.
Would you mind offering a "Dummies Guide" to what you want in a good amp for ESL's? Someone that's going to the trouble to build their own specialized amp obviously knows what they need.
Sure, there's 4 major things really:
1. The amp must tolerate having its outputs shorted for a considerable amount of milliseconds. This is important as the step-up transformer can go into core saturation when there is too much bass power in the material. If that happens, inadequate SS amps will trip their protection circuit which is annoying. Inadequate VT amps with gNFB will distort grossly which is, well, ... annoying 🙂 Not sure what zero-feedback VT amps would do.
2. At high frequencies the ESL is predominantly capacitive. The panel capacitance (~1nF for full range) is reflected to the amp as a 2.5uF capacitance (1:50 step-up transformer) or even 10uF (1:100 transformer).
Now the amp needs to be able to supply a lot of reactive current in order to impress the signal voltage on that capacitor. Since voltage and current are in quadrature here, the output stage will not dissipate a lot of heat but it must be able to source and sink a lot of current.
3. In conjunction with (2) above, the amp needs to be absolutely stable and not oscillate with a large phase shift at the output (capacitive load). Fortunately most well-designed amps are.
4. Outside of the above fault conditions, the amp must have enough power because sensitivity is low on ESLs, especially full range. About 40-50W would be a minimum. Less will work but we will lose a lot of dynamics.
For completeness and not getting flamed I should add that
a) there exist compensation networks that can alleviate some of the above, often at the expense of an even lower sensitivity
b) non-full range ESLs have considerably less problems on all 4 counts
My design will mainly avoid problem (2) because the step-up ratio is so much lower. Problem (1) can be solved by getting a bigger core or by putting a resistor in series with the primary (ouch efficiency!). The transformer in my project might have better coupling between pri and sec, hence lower leakage inductance, hence better LF behavior. Will have to wait and see.
Hope this helps
Kenneth
Last edited:
All but the last point are part of the challenge (if I could only figure out how to make it work class AB without a transformer in the audio chain).
Is not tube life cathode emission life? If I operate a tube at 1/20 of its rated cathode current I would not expect it to have a short lifetime.
But this thread is about your amp. Can you give some details of the OPT's?
Jim
Hi, Jim, thanks for your interest. The OPT's are potted 200W EI cores, nothing of the fancy core stuff like C-cores or nickel. They have better-than-usual insulation between both windings to withstand the voltage. They also have multiple taps so I can experiment with the transformation ratio a bit. Finally there's some feedback windings (not drawn), in case I want to make some local cathode feedback.
What I'm aiming for is a "usable prototype" of sorts, to show the approach works and then use as my day-to-day amp. I won't have the time to build a prototype and a "real" unit.
Re the lifetime of tubes in DD amps, it's not something I experienced first-hand but many times I heard or read this from people who actually built one. Not sure what the precise failure mode is, though.
Kenneth
What I'm aiming for is a "usable prototype" of sorts, to show the approach works and then use as my day-to-day amp. I won't have the time to build a prototype and a "real" unit.
Re the lifetime of tubes in DD amps, it's not something I experienced first-hand but many times I heard or read this from people who actually built one. Not sure what the precise failure mode is, though.
Kenneth
Looks like you did think about it! You may end up using those feedback windings because a triodes has a positive feedback path though the plate to grid capacitance. You have some local feedback from the unbypassed cathode resistors but since the transformer and ESL are inductance and capacitance they must have a resonant frequency. Will the tube current act as a parallel resistance to load it down? I wanted to build something and what you are doing would have been my second choice.
Jim
Jim,
I'm not sure I follow you on the plate-to-grid capacitance; wouldn't that give negative feedback, seeing as how plate voltage is out-of-phase with grid voltage?
Yeah the cathode resistors are a necessary evil. The 330R are for creating a path for cathode current without requiring a center tapped filament supply. The 100R are sense resistors to measure the cathode current -- I might sub these for 10R.
You are right that the reflected panel capacitance and the transformers leakage inductance make a parallel LC circuit. This will give an impedance peak (not dip) somewhere in the low end of the audio band for typical full-range panel-transformer combos. As you say, the tube's rp will damp the peak. All this has little effect on the amplitude curve of the speaker, since the tubes can impress the correct voltage on the C easily enough at that resonant frequency.
Will be starting cabinet building this weekend!
Kenneth
I'm not sure I follow you on the plate-to-grid capacitance; wouldn't that give negative feedback, seeing as how plate voltage is out-of-phase with grid voltage?
Yeah the cathode resistors are a necessary evil. The 330R are for creating a path for cathode current without requiring a center tapped filament supply. The 100R are sense resistors to measure the cathode current -- I might sub these for 10R.
You are right that the reflected panel capacitance and the transformers leakage inductance make a parallel LC circuit. This will give an impedance peak (not dip) somewhere in the low end of the audio band for typical full-range panel-transformer combos. As you say, the tube's rp will damp the peak. All this has little effect on the amplitude curve of the speaker, since the tubes can impress the correct voltage on the C easily enough at that resonant frequency.
Will be starting cabinet building this weekend!
Kenneth
Attachments
I think I was wrong there. I must have been thinking about class C RF amps.
That's some nice stuff in the photo! Are those two rectifier tubes next to the four power tubes?
Jim
Thanks, most of that stuff is very old. Yes they are half-wave (single diode) rectifiers type U19 / CV187. They have top caps and a nasty British B4 socket but I got them cheap, and a pair will do up to 500mA.
Kenneth
Kenneth
Some build photos.
Got some work done since my last post in this thread.
The amp will consist of several decks in a rack. In the photos you can see the rack I built for the amp. The fixed shelf is the filament supply deck. The two other shelves in the photo are the plate power supply deck (left) and the final stage deck (right). There will be one more deck for the driver stage (not yet made).
Kenneth
Got some work done since my last post in this thread.
The amp will consist of several decks in a rack. In the photos you can see the rack I built for the amp. The fixed shelf is the filament supply deck. The two other shelves in the photo are the plate power supply deck (left) and the final stage deck (right). There will be one more deck for the driver stage (not yet made).
Kenneth
Attachments
I've put some comments strongly favourable to direct drive on another thread. Some thoughts:
1. Although you lose the "free" bias-boosting B+ when you have a transformer, you do gain the immense value of less chance of death by electrocution.
2. Although ESLs that use matching transformers are to my ears clearly worse than DD (but still fabulous), your transformer is inside the feedback loop and that makes a lot of difference.
3. I'd do everything possible to create a higher-voltage solid state amp and avoid the many practical issues with "glass audio." I surely do not think tubes are better sounding.
1. Although you lose the "free" bias-boosting B+ when you have a transformer, you do gain the immense value of less chance of death by electrocution.
2. Although ESLs that use matching transformers are to my ears clearly worse than DD (but still fabulous), your transformer is inside the feedback loop and that makes a lot of difference.
3. I'd do everything possible to create a higher-voltage solid state amp and avoid the many practical issues with "glass audio." I surely do not think tubes are better sounding.
- Status
- Not open for further replies.