Hello everybody!
I'm working on a EL84PP amplifier, in pentode mode, much similar to the Baby Huey. A link to the original thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=72536&highlight=baby+huey
The problem is: I get much better results, in terms of output power and distortion, without the shunt feedback, by connecting the 220k ECC83 anode resistors directly to a well filtered B+ point (the G2 regulator, 300V).
Instead, when implementing the same circuit as Baby Huey, with 47k resistors from EL84 plates (340V B+, cathode bias) and whatever shunt feedback resistor (I got the best THD results from 33k), the distortion increases dramatically, and the output power is limited at about 3-4W by early clipping. Removing NFB (low, about 3dB) does show that the clipping exist even at the ECC83 plates, that's not an EL84 fault.
I have also to say that the OPT I'm using are cheap, but even without shunt feedback I got a -0.5dB point at 30kHz at 5W output. Maximum 5% THD power is 10W. With shunt feedback I couldn't go further than 4-5Vrms on 7ohm load, that means a couple of watts or so.
Frequency response at 1W, at 30Hz and 30kHz (compared to 1kHz):
- with shunt feedback +0.13dB and -0.95dB;
- without -0.1dB and +0.1dB.
So even frequency response is marginally better without shunt fb.
What should I look for? Even modelling the circuit with SPICE (even other circuits, with EL34 or 6L6) I cannot make the shunt feedback work, I mean the distortion ALWAYS increases with shunt feedback.
Thank you all in advance!
I'm working on a EL84PP amplifier, in pentode mode, much similar to the Baby Huey. A link to the original thread here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=72536&highlight=baby+huey
The problem is: I get much better results, in terms of output power and distortion, without the shunt feedback, by connecting the 220k ECC83 anode resistors directly to a well filtered B+ point (the G2 regulator, 300V).
Instead, when implementing the same circuit as Baby Huey, with 47k resistors from EL84 plates (340V B+, cathode bias) and whatever shunt feedback resistor (I got the best THD results from 33k), the distortion increases dramatically, and the output power is limited at about 3-4W by early clipping. Removing NFB (low, about 3dB) does show that the clipping exist even at the ECC83 plates, that's not an EL84 fault.
I have also to say that the OPT I'm using are cheap, but even without shunt feedback I got a -0.5dB point at 30kHz at 5W output. Maximum 5% THD power is 10W. With shunt feedback I couldn't go further than 4-5Vrms on 7ohm load, that means a couple of watts or so.
Frequency response at 1W, at 30Hz and 30kHz (compared to 1kHz):
- with shunt feedback +0.13dB and -0.95dB;
- without -0.1dB and +0.1dB.
So even frequency response is marginally better without shunt fb.
What should I look for? Even modelling the circuit with SPICE (even other circuits, with EL34 or 6L6) I cannot make the shunt feedback work, I mean the distortion ALWAYS increases with shunt feedback.
Thank you all in advance!
Further experiments with Spice confirmed that shunt feedback does work: with simplified OPT model (only primary inductance) I estimated the output power keeping the 2% THD fixed.
Without gNFB, I had 4.61W with shunt feedback and 4.44W without. With gNFB (about 4.7dB) I have 15.35W with shunt feedback and 6.06w without.
I'm guessing there's some strange OPT behaviour that suck out the output power. Probably the circuit is unstable in some way: could it be?
But didn't the "Baby Huey" circuit have the possibility to adapt low performance OPTs?
Without gNFB, I had 4.61W with shunt feedback and 4.44W without. With gNFB (about 4.7dB) I have 15.35W with shunt feedback and 6.06w without.
I'm guessing there's some strange OPT behaviour that suck out the output power. Probably the circuit is unstable in some way: could it be?
But didn't the "Baby Huey" circuit have the possibility to adapt low performance OPTs?
Giame,
The low power you see is a big tip off that the input stage is clipping and failing to drive the EL84s to full power. I ended up using 22K for the shunt feedback set resistor and got 7 watts (Ultralinear Mode) with a 285 volt rail. I never did distortion measurements at all, just listened to it.
You said:
"Without gNFB, I had 4.61W with shunt feedback and 4.44W without. With gNFB (about 4.7dB) I have 15.35W with shunt feedback and 6.06w without".
This doesn't make any sense to me. If I saw the last sentence by itself I would simply conclude that the Zout of the amp was way too high without feedback and the tubes delivered power was being dissipated in the Zout rather than in the load and adding globval feedback fixed that by reducing Zout. If that was the case, however, I would have expected to see similar difference between the with and without shunt feedback readings ???.
You talked about a screen regulator so I assume that you are using Pentode Mode rather than Ultralinear Mode. That will cetainly give you a very large Zout without some form of feedback and some global feedback is going to be required regardless of if you also have shunt feedback or not. If your OPT has screen taps I strongly recommend Ultralinear Mode.
Cheers,
Ian
The low power you see is a big tip off that the input stage is clipping and failing to drive the EL84s to full power. I ended up using 22K for the shunt feedback set resistor and got 7 watts (Ultralinear Mode) with a 285 volt rail. I never did distortion measurements at all, just listened to it.
You said:
"Without gNFB, I had 4.61W with shunt feedback and 4.44W without. With gNFB (about 4.7dB) I have 15.35W with shunt feedback and 6.06w without".
This doesn't make any sense to me. If I saw the last sentence by itself I would simply conclude that the Zout of the amp was way too high without feedback and the tubes delivered power was being dissipated in the Zout rather than in the load and adding globval feedback fixed that by reducing Zout. If that was the case, however, I would have expected to see similar difference between the with and without shunt feedback readings ???.
You talked about a screen regulator so I assume that you are using Pentode Mode rather than Ultralinear Mode. That will cetainly give you a very large Zout without some form of feedback and some global feedback is going to be required regardless of if you also have shunt feedback or not. If your OPT has screen taps I strongly recommend Ultralinear Mode.
Cheers,
Ian
Hello Ian,
sorry but I couldn't answer before.
Those power measurements you quote in the second post, are SPICE simulation results, NOT real amp performance.
I'm using pentode mode now: my OPT doesn't provide screen taps.
What confuses me is that ECC83 LTP + EL84 in pentode mode should have very big gain, thus ample margin for reducing Zout, in both ways (shunt and global feedback). Instead I can add only about 4dB of gNFB (without shunt feedback) to make the amp clip with normal sources: I have to investigate on the amp open loop gain.
So your suggestion is to "help" the shunt feedback in the reducing of Zout: low power is because some power is lost in the Zout of the amp.
With newer OPTs that are still in the developing process, I wil have the possibility to use 25% and 50% UL taps (well, they aren't UL taps, this explains the strange ratios). What do you suggest? Shunt feedback keeping pentode mode (with mosfet regulated feed) increasing gNFB, or UL with low gNFB?
I can answer myself, so I think it's an useless question. But if I can't have access to the UL taps, what do you think will be the best, keep the ECC83 LTP or convert it in a grounded cathode + cathodyna? Maybe this way there's more gain to spend in gNFB thus lowering Zout...
Thank you very much!
sorry but I couldn't answer before.
Those power measurements you quote in the second post, are SPICE simulation results, NOT real amp performance.
I'm using pentode mode now: my OPT doesn't provide screen taps.
What confuses me is that ECC83 LTP + EL84 in pentode mode should have very big gain, thus ample margin for reducing Zout, in both ways (shunt and global feedback). Instead I can add only about 4dB of gNFB (without shunt feedback) to make the amp clip with normal sources: I have to investigate on the amp open loop gain.
So your suggestion is to "help" the shunt feedback in the reducing of Zout: low power is because some power is lost in the Zout of the amp.
With newer OPTs that are still in the developing process, I wil have the possibility to use 25% and 50% UL taps (well, they aren't UL taps, this explains the strange ratios). What do you suggest? Shunt feedback keeping pentode mode (with mosfet regulated feed) increasing gNFB, or UL with low gNFB?
I can answer myself, so I think it's an useless question. But if I can't have access to the UL taps, what do you think will be the best, keep the ECC83 LTP or convert it in a grounded cathode + cathodyna? Maybe this way there's more gain to spend in gNFB thus lowering Zout...
Thank you very much!
Giaime,
If you use Ultralinear Mode you can generally get away with not using any global NFB at all which has benefits in stereo imaging and the spacial presentation of the "sound stage". It also sounds more dynamic and "live".
There should be adequate gain for either 25% or 50% Ultralinear - I would try both and choose by listening.
One thing to check - does the current source for the ECC83 LTP have enough compliance voltage to work properly.
With no global feedback, the grid of the second triode is at 0v AC AND the cathodes of the triodes are at 1/2 of the input signal AC voltage. The ECC83 requires such low (DC) bias voltages that CCS MUST be returned to a negative voltage rail to be able to accommodate that 1/2 input signal swing on the cathodes. Returning the CCS to 0V will almost certainly result in severe clipping as the CCS runs into saturation (collector volts fall below the base voltage) and possible cuts off the input triode (on negative going peaks of the input signal). I would recommend at least -5V as the supply to which to return the CCS. Of course it can be as high as you like.
Hope this is of help,
Cheers,
Ian
If you use Ultralinear Mode you can generally get away with not using any global NFB at all which has benefits in stereo imaging and the spacial presentation of the "sound stage". It also sounds more dynamic and "live".
There should be adequate gain for either 25% or 50% Ultralinear - I would try both and choose by listening.
One thing to check - does the current source for the ECC83 LTP have enough compliance voltage to work properly.
With no global feedback, the grid of the second triode is at 0v AC AND the cathodes of the triodes are at 1/2 of the input signal AC voltage. The ECC83 requires such low (DC) bias voltages that CCS MUST be returned to a negative voltage rail to be able to accommodate that 1/2 input signal swing on the cathodes. Returning the CCS to 0V will almost certainly result in severe clipping as the CCS runs into saturation (collector volts fall below the base voltage) and possible cuts off the input triode (on negative going peaks of the input signal). I would recommend at least -5V as the supply to which to return the CCS. Of course it can be as high as you like.
Hope this is of help,
Cheers,
Ian
Hi Ian,
that's not my case. I use the existing DC heater supply (12V) as a negative rail for the input CCS: I grounded the positive output and I'm using the negative output of the PSU to power ECC83 filaments and the CCS. I will double check it, but I doubt that the CCS has problems, because when I removed the shunt feedback I didn't remove the CCS, and it works well up to now.
I will try with the new OPTs when they're ready. Thank you very much!
that's not my case. I use the existing DC heater supply (12V) as a negative rail for the input CCS: I grounded the positive output and I'm using the negative output of the PSU to power ECC83 filaments and the CCS. I will double check it, but I doubt that the CCS has problems, because when I removed the shunt feedback I didn't remove the CCS, and it works well up to now.
I will try with the new OPTs when they're ready. Thank you very much!
Hi Giaime,
generally it is no good idea to use current trickler & high rp tubes as drivers in a (balanced or not) shunt FB setup. Compute the Z poor ECC83 is facing and forced to deal with and you will see
I said "generally" - obviously there are good sounding amps like Baby Huey and the ECL86PP from Yves that use balanced shunt FB with ECC83 (triode section of ECL86 being identical to ECC83 section except max.Pd).
Patrick Turner has some interesting notes on balanced shunt FB available at his site, see here. He uses mu-followers to deal with the Z problem.
Tom
generally it is no good idea to use current trickler & high rp tubes as drivers in a (balanced or not) shunt FB setup. Compute the Z poor ECC83 is facing and forced to deal with and you will see
I said "generally" - obviously there are good sounding amps like Baby Huey and the ECL86PP from Yves that use balanced shunt FB with ECC83 (triode section of ECL86 being identical to ECC83 section except max.Pd).
Patrick Turner has some interesting notes on balanced shunt FB available at his site, see here. He uses mu-followers to deal with the Z problem.
Tom
Thank you very much Tom for the link and for the explanation.
The most interesting thing is that I'm not investigating on shunt feedback in general, but specifically I built a "Baby Huey" clone (well, sort of it) and it didn't work as expected.
As I said I will try UL, and will study what Patrick says.
Thank you all very much!
The most interesting thing is that I'm not investigating on shunt feedback in general, but specifically I built a "Baby Huey" clone (well, sort of it) and it didn't work as expected.
As I said I will try UL, and will study what Patrick says.
Thank you all very much!
Hi guys,
Here is the OPT I've used in the original ECL86 prototype:
and the file for OPT_DA:
http://www.dissident-audio.com/PP_ECL86/DA17.MOX
Link to the ECL86 prototype:
http://www.dissident-audio.com/PP_ECL86/Page.html
Ciao, Ivo
Here is the OPT I've used in the original ECL86 prototype:
and the file for OPT_DA:
http://www.dissident-audio.com/PP_ECL86/DA17.MOX
Link to the ECL86 prototype:
http://www.dissident-audio.com/PP_ECL86/Page.html
Ciao, Ivo
Hello!
I'm resurrecting this topic to let you know that I finally managed to try ultra-linear connection. But with the old transformers, since new ones have errors.
Well, nothing has changed. Output impedance of the amp is lower with shunt feedback than plain ultra-linear (2.5ohm with sFB, 3.5ohm without), but also output power is reduced at the same THD (was 8-9W, now is 3W).
Those measurements are made with 33k shunt resistor: with the suggested 15k, output power couldn't even reach 1W.
Frankly I don't know. All is strange in this amp: I expected lower distortion, lower output power, and wider bandwidth when changing from pentode mode to UL, obviously this wasn't so. Bandwidth remained the same (with my limited measurement setup, I can only test from 20Hz to 30kHz, and the output is pretty flat in this range in both modes, even with sFB applied).
Now, with shunt feedback applied, it seems like the driver tubes are clipping too early, but it's strange, I followed the schematic part's values.
Any ideas?
I'm resurrecting this topic to let you know that I finally managed to try ultra-linear connection. But with the old transformers, since new ones have errors.
Well, nothing has changed. Output impedance of the amp is lower with shunt feedback than plain ultra-linear (2.5ohm with sFB, 3.5ohm without), but also output power is reduced at the same THD (was 8-9W, now is 3W).
Those measurements are made with 33k shunt resistor: with the suggested 15k, output power couldn't even reach 1W.
Frankly I don't know. All is strange in this amp: I expected lower distortion, lower output power, and wider bandwidth when changing from pentode mode to UL, obviously this wasn't so. Bandwidth remained the same (with my limited measurement setup, I can only test from 20Hz to 30kHz, and the output is pretty flat in this range in both modes, even with sFB applied).
Now, with shunt feedback applied, it seems like the driver tubes are clipping too early, but it's strange, I followed the schematic part's values.
Any ideas?
Hello Yves,
what you say does happen
but that's not the case, I suppose.
I replaced the Sylvania 12AX7WA that I'm using with Siemens ECC83, same results. Also tried brand new JJ/Tesla ECC81, a bit less gain but similar results, so since there's a measurable reduction in gain with ECC81, I guess mine are real ECC83
About my amp... well, I cheated, I must admit. I wired the EL84 in triode mode, adjusted global negative feedback, and now I have the cleanest 5W I ever heard. Haven't tested the bandwidth yet, I guess it's not worse than the one in UL (that was already enough). No compensation needed, even with 11dB of global negative feedback.
Maybe I'll try the "Baby Huey" approach some other time, with "trusted" OPTs, those ones just seem strange to me.
what you say does happen
but that's not the case, I suppose.I replaced the Sylvania 12AX7WA that I'm using with Siemens ECC83, same results. Also tried brand new JJ/Tesla ECC81, a bit less gain but similar results, so since there's a measurable reduction in gain with ECC81, I guess mine are real ECC83
About my amp... well, I cheated, I must admit. I wired the EL84 in triode mode, adjusted global negative feedback, and now I have the cleanest 5W I ever heard. Haven't tested the bandwidth yet, I guess it's not worse than the one in UL (that was already enough). No compensation needed, even with 11dB of global negative feedback.
Maybe I'll try the "Baby Huey" approach some other time, with "trusted" OPTs, those ones just seem strange to me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.