Hello fellow DIY-ers. It's been a while since i posted here but i need some advice. I'm trying to build the amplifier below but i would very much like to know if i messed anything up before i actually test it and it blows in my face... I toook the idea from the Local Hero amplifier and altered it a little bit mainly because i do not have 12AU7 low mu tubes for the driver stages, and the gain of the first stage was HUGE. Now if i were to say i came up with this and then found the Local Hero i bet few would believe me. But anyway here's the schematic. Any ideas would be appreciated. The power supplies aren't shown but they will each have their regulator.
An externally hosted image should be here but it was not working when we last tested it.
The 100k input resistor will generate noise, which is likely to be noticeable at low volume settings.
A 12AX7 makes a poor cathode follower, and hence a poor phase splitter in this circuit.
I am unclear why you have a large Miller capacitor on the second voltage amplifier. Note that any difference in value between the two phases will lead to HF unbalance and hence a rise in HF even order distortion.
I can't quite work out what you are trying to do with the EL34 cathode circuit, but there appears to be far too much resistance.
My overall impression is that the concept/architecture is wrong.
A 12AX7 makes a poor cathode follower, and hence a poor phase splitter in this circuit.
I am unclear why you have a large Miller capacitor on the second voltage amplifier. Note that any difference in value between the two phases will lead to HF unbalance and hence a rise in HF even order distortion.
I can't quite work out what you are trying to do with the EL34 cathode circuit, but there appears to be far too much resistance.
My overall impression is that the concept/architecture is wrong.
Thanks for the reply.
I could replace the 12AX7 gain/ pi with another 6DJ8 and be content with it that way, and not have to force the gain to a specific value, just leave the first cathode unbypassed. The miller capacitors are there to limit the bandwidth of the part enclosed in the local feedback loop because that's what they did with the Local Hero amplifier or else it would certainly oscillate. I could alternately clamp the high frequencies returning from the transformer secondary by inserting a cap between the driver tubes cathodes... but then my amplifier might also amplify any RF that it may pick up from it's input.
Ok what i am trying to do is use only local feedback. The signal from the transformer output is symetrically fed to the EL34 cathodes AND to the driver tube cathodes, just like mr. Koren did with his design. What do you mean by too much resistance? I can tell you that i intend for the EL34's to have a "mixed" bias scheme as you may have noticed , for the sake of avoiding runaway. But i am open to any suggestion.
I could replace the 12AX7 gain/ pi with another 6DJ8 and be content with it that way, and not have to force the gain to a specific value, just leave the first cathode unbypassed. The miller capacitors are there to limit the bandwidth of the part enclosed in the local feedback loop because that's what they did with the Local Hero amplifier or else it would certainly oscillate. I could alternately clamp the high frequencies returning from the transformer secondary by inserting a cap between the driver tubes cathodes... but then my amplifier might also amplify any RF that it may pick up from it's input.
Ok what i am trying to do is use only local feedback. The signal from the transformer output is symetrically fed to the EL34 cathodes AND to the driver tube cathodes, just like mr. Koren did with his design. What do you mean by too much resistance? I can tell you that i intend for the EL34's to have a "mixed" bias scheme as you may have noticed , for the sake of avoiding runaway. But i am open to any suggestion.
I suppose most people think of global feedback as returning the output signal to the very first stage in the amplifier... But i agree with you. Then i should say i'm trying to build an amplifier that uses as local a feedback as possible (because we know we can't have NO feedback at all and expect the speakers to be properly controlled).
I hadn't spotted the mixed bias. Thermal runaway is a symptom of a fault, so you should not need to design as though it is an expected behaviour. You are just throwing away voltage.
Have you calculated what load the PS will see at 10 or 20kHz?
Have you calculated what load the PS will see at 10 or 20kHz?
CFB is already in the schematic. UL not possible because i've already finnished the OPTs 🙁 I would like to use Schade feedback, but one little question: why is it not so widespread? Does it sound awkward by any chance? Also, since it's plate feedback, would it not be easier (no coupling cap) and more efficient (higher gain within the loop) to return the signal from each EL34 plate to it's driver's cathode? Would that get along well with CFB? Some have done it and like it but i\m trying to get my head around how it actually works. Presume i do that. The following happens:
- the driver tubes (which i hope are biased correctly for a linear enough operation) can sense what their power tube counterpart does and keep it under control.
- the power tube gets feedback through it's cathode from the OPT output, however the OPT is never linear (or else it wouldn't need to be compensated)
- the power tube tries to respond nonlinearly for the output to be as it "thinks" it should be
- the driver tube senses the nonlinearity in the power tube's frequency response and fights against it
At least that's what i think...
But it's quite likely to happen with old worn tubes. If the amplifier works ok and i like it i will of course buy new tubes and make it fully fixed bias, also improving stability by not having the cathode RC groups.
No i haven't 🙁 I'm at work right now...
Sorry, but I could not read the schematic (the image is too small)... I would just stick with CFB if you already have it, no need to complicate matters with Schade. There are quite a few designs using Schade plate-to-grid feedback and it works well when you don't have an UL or CFB OPT, a search on this forum should turn up quite a few hits...
Sorry about that. The original was large though i don't know what happened.
Yes the last time i tried Schade feedback i didn't like it's sound that much...although it wasn't in a power tube.
So if i am to keep the feedback limited to 2 stages perhaps i could compensate each side with a cap over the resistor like they were 2 separate SE amplifiers? Or maybe that would make things a lot worse...
Edit: or... maybe i should just be content to use CFB + a bit of GNFB... triodes don't need much GNFB anyway...
Yes the last time i tried Schade feedback i didn't like it's sound that much...although it wasn't in a power tube.
So if i am to keep the feedback limited to 2 stages perhaps i could compensate each side with a cap over the resistor like they were 2 separate SE amplifiers? Or maybe that would make things a lot worse...
Edit: or... maybe i should just be content to use CFB + a bit of GNFB... triodes don't need much GNFB anyway...
Last edited:
My sentiment exactly, make the GNFB variable or selectable, so you can adjust to taste...🙂
Thanks for the encouragement. Yes that was my first idea but it would have been a poor performer because i hadn't modified the transformers for CFB yet... As it is CFB also helps improve the audio quality beyond what you'd expect from the transformers, from what i've read about it. Still, they're power transformers, very heavy, U core, modified by me to have some minimal interleaving. Paper thin core laminations, designed to be efficient, although may not be that stellar for audio. I guess i'll have to find out when i test the amp...
The reflected load should now be 5K p-p, which is the datasheet spec.
The reflected load should now be 5K p-p, which is the datasheet spec.
If you're not going to have feedback, why not make the amp as linear as possible open loop?
I'd start with a CCS tail on the output stage.
I'd start with a CCS tail on the output stage.
The amp I described in this thread measured quite well, sounded great, and used only local feedback in the output stage and CCSs.
See this thread for background on what the local feedback does to the effective curves of the output tube circuit. (makes them better than a triode connected KT88 or even a 300B)
Just another idea to consider.
Edit: This could easily be adapted for EL34s.
See this thread for background on what the local feedback does to the effective curves of the output tube circuit. (makes them better than a triode connected KT88 or even a 300B)
Just another idea to consider.
Edit: This could easily be adapted for EL34s.
Re: MrCurwen
"If you're not going to have feedback, why not make the amp as linear as possible open loop? I'd start with a CCS tail on the output stage."
---
Using a CCS tail under a class A P-P power stage is 3rd harmonic compressive, not particularly linear. What Alan W did, was use a compensating 3rd H expansive driver stage (balanced diffl. cascode, 6H30pi) in front to compensate for it.
One can easily see this problem with a CCS tail:
(Using the E55L datasheet here since it has gm versus Vg1 on pages 5 and 7, and gm versus Ip on page 9. The CCS tail constrains one to complementary currents, so page 9 is the correct graph to use here.)
Looking at page 9 of the E55L, one has gm (S) plotted versus plate current. For P-P class A one gets the S and flipped around S curves to overlap and sum for total output stage gm. They obviously sum to a big hump in the middle. Giving compressive 3rd harmonic distortion.
If one uses grounded cathode outputs (P-P again), then the gm (S) graph on page 5 or 7 versus Vg1 is the correct graph to use. This classical approach sums much better to a constant gm.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
300B has a very similar (to E55L) gm curve versus Vg1 (page 5):
http://frank.pocnet.net/sheets/084/3/300B.pdf
"If you're not going to have feedback, why not make the amp as linear as possible open loop? I'd start with a CCS tail on the output stage."
---
Using a CCS tail under a class A P-P power stage is 3rd harmonic compressive, not particularly linear. What Alan W did, was use a compensating 3rd H expansive driver stage (balanced diffl. cascode, 6H30pi) in front to compensate for it.
One can easily see this problem with a CCS tail:
(Using the E55L datasheet here since it has gm versus Vg1 on pages 5 and 7, and gm versus Ip on page 9. The CCS tail constrains one to complementary currents, so page 9 is the correct graph to use here.)
Looking at page 9 of the E55L, one has gm (S) plotted versus plate current. For P-P class A one gets the S and flipped around S curves to overlap and sum for total output stage gm. They obviously sum to a big hump in the middle. Giving compressive 3rd harmonic distortion.
If one uses grounded cathode outputs (P-P again), then the gm (S) graph on page 5 or 7 versus Vg1 is the correct graph to use. This classical approach sums much better to a constant gm.
http://frank.pocnet.net/sheets/009/e/E55L.pdf
300B has a very similar (to E55L) gm curve versus Vg1 (page 5):
http://frank.pocnet.net/sheets/084/3/300B.pdf
Last edited:
Re: SpreadSpectrum
That link with a local feedback graph comparison is quite illustrative. One can exceed the 300B with just some resistors.
That link with a local feedback graph comparison is quite illustrative. One can exceed the 300B with just some resistors.
Yes, you can set the circuit for a mu of 5 (higher than the 300B's 3.8) and still have better linearity and more power output without class *2 operation.
Of course, you pay for it by having to have more power supply voltages. Not a big drawback in my opinion but should be noted.
I have been thinking that it would be fun to make a KT88 SE amp this way and see how it compares to a 300B amp as far as distortion goes.
Re: Bloodmist
You will need a pretty good OT to put that much loop gain around the CFB winding. You say this is a re-worked PWR Xfmr... CFB just under the output stage may work.
Is this amp running in class A? (I see bypassed cathode resistors under the output cathodes)
Taking FDBKs back to the driver cathodes certainly can work, but the high gain available there would more likely only be stable from some primary points, like plates or UL taps. (low % CFB has too much % signal going
-through- the OT, however some McIntosh designs get away with it because of the higher 50% CFB) FDBKs to the driver cathodes can work so well that I have heard complaints that an amp doesn't sound like a tube amp anymore. The RCA handbook has an example 50 Watt amp. The driver cathodes make better sense than going back to the driver grids or plates, which cause low impedance problems to drive.
If one does not have output tube CFBs, and the amp is running in class aB, then another interesting option is to put current sense resistors in the output cathodes with crossed R feedbacks to the driver cathodes. Also put R feedbacks from the output plates back to the driver cathodes as well, and proportion them so that the V to I feedback ratio mimics the expected V to I ratio for the load. Then the "local" driver loop will enforce constant gm sum for the output stage with little effect from load Z. (and so minimizing crossover distortion for the class aB output)
(something of a simulated CFB, but using all local feedbacks)
You will need a pretty good OT to put that much loop gain around the CFB winding. You say this is a re-worked PWR Xfmr... CFB just under the output stage may work.
Is this amp running in class A? (I see bypassed cathode resistors under the output cathodes)
Taking FDBKs back to the driver cathodes certainly can work, but the high gain available there would more likely only be stable from some primary points, like plates or UL taps. (low % CFB has too much % signal going
-through- the OT, however some McIntosh designs get away with it because of the higher 50% CFB) FDBKs to the driver cathodes can work so well that I have heard complaints that an amp doesn't sound like a tube amp anymore. The RCA handbook has an example 50 Watt amp. The driver cathodes make better sense than going back to the driver grids or plates, which cause low impedance problems to drive.
If one does not have output tube CFBs, and the amp is running in class aB, then another interesting option is to put current sense resistors in the output cathodes with crossed R feedbacks to the driver cathodes. Also put R feedbacks from the output plates back to the driver cathodes as well, and proportion them so that the V to I feedback ratio mimics the expected V to I ratio for the load. Then the "local" driver loop will enforce constant gm sum for the output stage with little effect from load Z. (and so minimizing crossover distortion for the class aB output)
(something of a simulated CFB, but using all local feedbacks)
Last edited:
I'm not sure where to start... i didn't imagine you folks are more active at night 🙂
The amplifier will (ideally) be a class AB1, because efficiency of the "triodes" is low as it is. A CCS would unfortunately limit me to class A1 operation.
Yes that's the very least it can do but ideally it should also return to the driver cathodes.
Certainly didn't work for an EL84 amp i built in the past, it turned into an RF jammer. But that was just my stupidity, i was trying to get it to work at unity gain (which is nearly impossible). I would prefer to also use feedback to the driver cathodes for this amp, however if i can't it will still make a great amp with less than 10dB GNFB + CFB.
The real question is how would i go about implementing CFB to the drivers in a stable fashion? Meaning how do i limit the bandwidth of the returned signal. Do i just add a cap between the driver cathodes to snub out-of-phase high frequencies? I see the way they added to the Miller capacitance of the drivers in the Local Hero but some complain that it kills the amp's slew rate...
The amplifier will (ideally) be a class AB1, because efficiency of the "triodes" is low as it is. A CCS would unfortunately limit me to class A1 operation.
Yes that's the very least it can do but ideally it should also return to the driver cathodes.
Certainly didn't work for an EL84 amp i built in the past, it turned into an RF jammer. But that was just my stupidity, i was trying to get it to work at unity gain (which is nearly impossible). I would prefer to also use feedback to the driver cathodes for this amp, however if i can't it will still make a great amp with less than 10dB GNFB + CFB.
The real question is how would i go about implementing CFB to the drivers in a stable fashion? Meaning how do i limit the bandwidth of the returned signal. Do i just add a cap between the driver cathodes to snub out-of-phase high frequencies? I see the way they added to the Miller capacitance of the drivers in the Local Hero but some complain that it kills the amp's slew rate...
One thing i forgot to add i decided to ditch the 12AX7 and the distortion figures went through the floor immediately (at least in a simulation) and even using the reviled GNFB (only about 7dB) along with my desired CFB...
Instead i used only 6DJ8 tubes which give the amp a better slew rate. In reality i'm using 6N1P and the predicted figures for the 6DJ8 match the real life 6N1P every time with only minor differences (which i can account to tube aging).
Instead i used only 6DJ8 tubes which give the amp a better slew rate. In reality i'm using 6N1P and the predicted figures for the 6DJ8 match the real life 6N1P every time with only minor differences (which i can account to tube aging).
- Status
- Not open for further replies.