I am looking for different options of OPS. I am more interested in push pull configuration.
1) Comparing Ultra Linear vs Triode push pull. Which one has lower distortion?
2) I understand the advantage of using triode for push pull because triode has mainly even harmonics and they cancel out in push pull if it is AC balanced. But how much power can I get with a pair of EL34 if the power supply is designed for 40W when connected as penthode?
3) Bob Cordell gave very detail in the crossover distortion of a pair of complementary transistor pair and how to optimize them. Can anyone give me a link that analyze crossover distortion of push pull tube with OT?
4) Does running class A a lot better?
4) Any other OPS configuration that is better than Ultra Linear and Triode push pull?
1) Comparing Ultra Linear vs Triode push pull. Which one has lower distortion?
2) I understand the advantage of using triode for push pull because triode has mainly even harmonics and they cancel out in push pull if it is AC balanced. But how much power can I get with a pair of EL34 if the power supply is designed for 40W when connected as penthode?
3) Bob Cordell gave very detail in the crossover distortion of a pair of complementary transistor pair and how to optimize them. Can anyone give me a link that analyze crossover distortion of push pull tube with OT?
4) Does running class A a lot better?
4) Any other OPS configuration that is better than Ultra Linear and Triode push pull?
Last edited:
Why not compare the options yourself? As an OPT made for UL connections also has every possibility for triode or pentode connection, you can compare the options yourself. What option is best, depends on many variables, such as tubes used, load impedance, B+ voltage and bias current.
With class A, there is NO crossover distortion. The commonly used bias is class AB for low to medium output power stages in PP, that is normally class A up to a certain power, then more like class B. Again no crossover distortion. You can´t compare transistor topologies to tubes.
The output power with EL34 triode connected I´m a bit unsure about, but 15-20W should be obtainable with the optimum load. The load has to be calculated according to tubes, B+ and choosen type of bias, that is fixed vs cathode bias...
As an example, EL84 likes an impedance around 8-10 kohms, 6C33S an impedance around 1,2 kohm, thats in push-pull. The latter tube is a triode.
With class A, there is NO crossover distortion. The commonly used bias is class AB for low to medium output power stages in PP, that is normally class A up to a certain power, then more like class B. Again no crossover distortion. You can´t compare transistor topologies to tubes.
The output power with EL34 triode connected I´m a bit unsure about, but 15-20W should be obtainable with the optimum load. The load has to be calculated according to tubes, B+ and choosen type of bias, that is fixed vs cathode bias...
As an example, EL84 likes an impedance around 8-10 kohms, 6C33S an impedance around 1,2 kohm, thats in push-pull. The latter tube is a triode.
Last edited:
Some info here:
European Triode Festival and Crossover Notch Distortion and New OTL Design
OTL Amplifier Design: the Broskie OTL
It should be possible to do the de-notch technique with a conventional center-tapped P-P OT too. Just put the current sensing resistors in the cathodes of the output tubes and run the sense points back to the cathodes of a differential driver stage.
Using CFB windings, with equal internal winding resistances, should work also, and in fact some McIntosh Amps do have feedbacks from the CFBs back to the driver stage.
European Triode Festival and Crossover Notch Distortion and New OTL Design
OTL Amplifier Design: the Broskie OTL
It should be possible to do the de-notch technique with a conventional center-tapped P-P OT too. Just put the current sensing resistors in the cathodes of the output tubes and run the sense points back to the cathodes of a differential driver stage.
Using CFB windings, with equal internal winding resistances, should work also, and in fact some McIntosh Amps do have feedbacks from the CFBs back to the driver stage.
Last edited:
I am kind of wondering about this. The crossover distortion in SS output stage is mainly caused by the voltage divider effect of the output impedance of the power amp and the impedance of the speaker. Distortion is caused by the changing of the output impedance of the power amp with the output voltage.
For tube amp, the output impedance looking back into the OT is the step down of the output impedance of the power tubes. Assume no global feedback, the output impedance looking into the plate of the power tube is quite high, it's more like a current source. Is this the reason why crossover distortion is insignificant with tube power amp? If it is true, this is a strong point of tube power amp.
No. If crossover distortion was simply due to impedance variation at the output then a valve amp might have much worse crossover than SS.
Crossover distortion is caused by a number of things, and potentially different things in different output stage topologies. For example, in some cases it may arise partly from kinks in the input impedance of the output stage interacting with the output impedance of the driver stage.
Crossover distortion is "better" in valve output stages for two reasons (imo)
1: The crossover region is very broad and low order because valves operate according to a 3/2 power law as opposed to the exponential law of BJTs. With a bit of math you can see that the series expansion of x^(3/2) falls off quicker than that of e^x. This means that it is subjectively less audible for a given THD reading.
2: Because the crossover region is so broad, there is no "gm doubling" to speak of. The more idle current the better. Most of the classic PP amps went all the way to Class A, so crossover distortion could certainly be called insignificant since there was no crossover.
1: The crossover region is very broad and low order because valves operate according to a 3/2 power law as opposed to the exponential law of BJTs. With a bit of math you can see that the series expansion of x^(3/2) falls off quicker than that of e^x. This means that it is subjectively less audible for a given THD reading.
2: Because the crossover region is so broad, there is no "gm doubling" to speak of. The more idle current the better. Most of the classic PP amps went all the way to Class A, so crossover distortion could certainly be called insignificant since there was no crossover.
Last edited:
I don't see why crossover distortion is worst in valve amp. This is limited to the effect of the power tube. As I was saying, output impedance much higher, so the voltage divider effect between the output impedance and the speaker is a lot less. So the crossover distortion is lower.
Yes, the speaker impedance variation have a lot more effect on the amp with high output impedance, but that is not crossover distortion. Driver stage does not contribute to the crossover distortion. You need different method to lower the distortion on the driver stage, so it's a different subject.
That's what I think, so I don't understand why people keep saying valve power amp is inferior. If you connect the power valve as triode, it's most have only 2nd harmonics, this is easily cancelled in push pull stage if you do AC balance. Transistor is more like pentode that has high order of odd harmonics that cannot be cancelled out. This together with no crossover distortion should make valve power amp superior than SS.
This logic extends to differential PI stage also. Triode mostly has 2nd harmonics. The 2nd harmonics in the PI should also be cancelled in the push pull stage also.
This just as true for pentode, UL, and the other topology zoos. You also have to consider gain and source impedance of the tubes as part of the overall design. It's just not as simple as "triodes are best" or other useless mantras.
I didn't say that. I said that if it was caused by impedance then it would be worse for valve outputs.Alan0354 said:
Opposite way round. To the extent that crossover distortion comes from output impedance variations then the higher the output impedance the worse the distortion (because the voltage divider effect is a lot more).
In reality valve amps often suffer less from crossover distortion. They usually suffer more from other distortions, as much less global NFB is used than in typical SS.
Power tubes are generally optimized for good gm, so they have more like square law V to I transfer curves, like Mosfets (from grid wire to cathode proximity effects). Their power law gets more linear at higher current too, like Mosfets.
Since square law transfer (V to I) has linear law gm, the overlapped class A region from class AB should have constant gm. So one should end up with a gm versus I curve that is nearly flat in the middle (rounding up a little in the middle from gm law droop at higher current) and ramping up outside that region but becoming assymptotic to a constant level eventually, then finally drooping off badly near saturation.
(ie., the sum of two overlapped rolling off "linear" ramps)
One of Broskie's articles mentioned above say he saw a factor of 10 improvement with the "de-notcher" circuit (in simulation), which corrects for gm doubling. He doesn't mention the absolute crossover distortion however.
NOTE:
That may have been his simulation of the SS circuit, not clear. Also, I'm not real convinced that his tube version is doing the right function, since it does not take the difference of the two current sense feedbacks into consideration for control for each output tube directly. The far simpler approach of taking the current sense feedbacks back to the cathodes of a differential driver should work, and work for a conventional center tapped P-P OT configuration besides, or a CFB setup like the Mac. (at least one of the Mac designs used CFBs back to driver cathode connections, as long as the CFB windings had equal internal winding resistance.)
In the SS de-notch version, the two I sense resistors only produce varying relative feedback voltage from across the emitters during crossover. So that produces a different loop gain between the crossover region and SE regions. Broskies tube version does not take the difference directly, so it is actually enforcing linear gain of each tube, which do not add up correctly for removing the overlap gm notch.
Since square law transfer (V to I) has linear law gm, the overlapped class A region from class AB should have constant gm. So one should end up with a gm versus I curve that is nearly flat in the middle (rounding up a little in the middle from gm law droop at higher current) and ramping up outside that region but becoming assymptotic to a constant level eventually, then finally drooping off badly near saturation.
(ie., the sum of two overlapped rolling off "linear" ramps)
One of Broskie's articles mentioned above say he saw a factor of 10 improvement with the "de-notcher" circuit (in simulation), which corrects for gm doubling. He doesn't mention the absolute crossover distortion however.
NOTE:
That may have been his simulation of the SS circuit, not clear. Also, I'm not real convinced that his tube version is doing the right function, since it does not take the difference of the two current sense feedbacks into consideration for control for each output tube directly. The far simpler approach of taking the current sense feedbacks back to the cathodes of a differential driver should work, and work for a conventional center tapped P-P OT configuration besides, or a CFB setup like the Mac. (at least one of the Mac designs used CFBs back to driver cathode connections, as long as the CFB windings had equal internal winding resistance.)
In the SS de-notch version, the two I sense resistors only produce varying relative feedback voltage from across the emitters during crossover. So that produces a different loop gain between the crossover region and SE regions. Broskies tube version does not take the difference directly, so it is actually enforcing linear gain of each tube, which do not add up correctly for removing the overlap gm notch.
Last edited:
Late fix:
The word "notch" above really should be changed to "bump" to be more descriptive, since the gm sum bumps up in the overlap region.
On thinking more about this gm fix scheme, I am seeing where it should work well for SS emitter followers, but not so well for Mosfets or tubes. It flattens the central gm bump (which emitter followers have big time), but does not fix the general gm valley between the opposite ramps issue. On that issue, Broskies tube circuit is working better. Looks like this "fix" needs to be completely re-thought out for tubes (and Mosfets). Back to square one...
The word "notch" above really should be changed to "bump" to be more descriptive, since the gm sum bumps up in the overlap region.
On thinking more about this gm fix scheme, I am seeing where it should work well for SS emitter followers, but not so well for Mosfets or tubes. It flattens the central gm bump (which emitter followers have big time), but does not fix the general gm valley between the opposite ramps issue. On that issue, Broskies tube circuit is working better. Looks like this "fix" needs to be completely re-thought out for tubes (and Mosfets). Back to square one...
Last edited:
Hawksford's EC paper(s) had a later addition of differential current feedback to that scheme. I think its time to dig out his paper.
Ahh..., OK. Just needed some coffee to wake up here. The differential current signal is the same as the output current. So making the differential current signal proportional to input signal (via FDBK) is always going to make gm more constant. So the two I sense signals should go back to a differential driver stage as earlier mentioned. This is so easy to do, but I never see it done except possibly in the one version of the Mac.
Ahh..., OK. Just needed some coffee to wake up here. The differential current signal is the same as the output current. So making the differential current signal proportional to input signal (via FDBK) is always going to make gm more constant. So the two I sense signals should go back to a differential driver stage as earlier mentioned. This is so easy to do, but I never see it done except possibly in the one version of the Mac.
Last edited:
While the usual V FDBK alone tends to give one infinite gm, the differential I FDBK added to V FDBK tends to give one constant finite gm. This means that the amplifier would tend to have constant damping factor. (While V FDBK alone gives a very high damping factor that changes with the FDBK versus frequency effect, quite pronounced using an OT with global FDBK.)
So the special version of the Mac may have had a very constant damping factor. (of course this still does not take into account the varying Z-input of the speaker)
So the special version of the Mac may have had a very constant damping factor. (of course this still does not take into account the varying Z-input of the speaker)
Back to one of the original questions:
The UL.
Examining test results will indicate that with advancing the screen tap in the UL configuration, shows that rp and distortion drop fairly quickly to triode-region distortion and impedance while the efficiency and maximum output are still hardly down from pentode operation. (There are some handy KT88 graphs showing how rp, gm and distortion behave when the screen tap is advanced from zero (pentode) to 100% (triode operation). (Sorry, unable to find those right now.)
A typical practical example (for EL34s) is that maximum triode output is some 15W when in UL the figure is 30W. The distortion for UL at maximum output is somewhat higher than that for triodes, but at the triode maximum of 15W the UL topology distortion is only some 75% of that for triodes, while leaving another 3 dB of output for headroom.
The UL.
Examining test results will indicate that with advancing the screen tap in the UL configuration, shows that rp and distortion drop fairly quickly to triode-region distortion and impedance while the efficiency and maximum output are still hardly down from pentode operation. (There are some handy KT88 graphs showing how rp, gm and distortion behave when the screen tap is advanced from zero (pentode) to 100% (triode operation). (Sorry, unable to find those right now.)
A typical practical example (for EL34s) is that maximum triode output is some 15W when in UL the figure is 30W. The distortion for UL at maximum output is somewhat higher than that for triodes, but at the triode maximum of 15W the UL topology distortion is only some 75% of that for triodes, while leaving another 3 dB of output for headroom.
Here are some papers Claus Byrith wrote, about a traditional amplifier based on Mullard 5-1 design. Together with the theory there are exellent graphs showing the effects on output and distortion with varying UL ratio. Reading all will give a good sense on how different connections of output tubes affect power and distortion. And the transformers he used are Lundahls, one of the best available nowadays, with a good pricetag compared to Tango and other in the same class of quality...
30W Push Pull amplifier designed by Claus Byrith, Royal Academy of Music in Aarhus, Denmark. | Lundahl Transformers
30W Push Pull amplifier designed by Claus Byrith, Royal Academy of Music in Aarhus, Denmark. | Lundahl Transformers
The reason, as I understand it is the gap with no conduction of the active elements in the output circuit. This gap are when the input voltage for silicon transistors are lower than about 1 volt. Then the both transistors cease to conduct, and there is a hump in the waveform around zero volts out. To overcome this, the transistors are biased to have a level on the bases more than the lowest voltage at which they conduct, so overcoming this hump in the transfer function, and this gives a current thru the transistors at zero signal, measured as the bias current.
Tubes have a much more subtle cease of current, and the crossover distortion is not as severe as with transistors. But also tubes have to have a bias, to set the operation at the most linear portion of their transfer curves. This can be made visually obvious when plotting loadlines on tube transfer curves, which are available in tube datas.
Tubes have a much more subtle cease of current, and the crossover distortion is not as severe as with transistors. But also tubes have to have a bias, to set the operation at the most linear portion of their transfer curves. This can be made visually obvious when plotting loadlines on tube transfer curves, which are available in tube datas.
Thanks, you answer a lot of my questions. I just want to verify:
1) UL tapping is just a way to vary between panthodes ( screen connect to the CT of OT) and triode ( screen connect to plate). So if it is 70% tap, then it learn more to triode. 50% is in between.
2) The more the taps are towards triode, the lower the harmonic distortion, but the power is getting lower. The more the taps are towards the panthode, the more harmonic distortion you get BUT you get higher power.
3) Even at full power, UL has higher distortion than triode, but because UL provide higher power, when you back off the power output to match the triode mode, UL is actually has lower distortion.
Am I correct? So I should buy OT with UL tap? If I get UL tap, should I get 70% or over?
Also, why you loose power if you connect the pentode in triode mode? Voltage is the same, PT is the same, why the power is so much lower?
Thanks
Thanks, I need some time to read this. I am interested in Lundahl OT, seriously thinking about buying a pair for EL34.
- Status
- Not open for further replies.