If amplifier fidelity---the ability of an amplifying device to faithfully amplify an input signal---depends on linearity in every respect of the device's operating parameters, and if any given single device (here, tube) has what might be called a non-linear finger-print---ie, departs from linearity differently than any other given single device---would paralleling not reduce, by the law of averages, any non-linearities peculiar to a given individual device? And wouldn't these "finger-print" non-linearities be greater reduced the larger number of devices operated in parallel?
Conrad-Johnson uses massive parallel operation in their ART preamplifier to seemingly excellent effect. Perhaps part of the OTL magic is explained by this massive paralleling effect?
Here's an idea (forgive my tendencies to SE design): massively parallel small-signal tubes on the output to achieve what is effectively a massive-PSE?
Ideas anyone?
Conrad-Johnson uses massive parallel operation in their ART preamplifier to seemingly excellent effect. Perhaps part of the OTL magic is explained by this massive paralleling effect?
Here's an idea (forgive my tendencies to SE design): massively parallel small-signal tubes on the output to achieve what is effectively a massive-PSE?
Ideas anyone?
I wouldn't call it that at all. A single tube always sounds better to my ears than two in parallel. If paralleling has a sonic signature it is one of blurring and obscuring details.
If commercial amps indeed use parallel tubes it is usually for a particular purpose like higher output power/lower output impedance/ noise but seldom because of better perceived sound.
EC8010 said:
Hmmm, let's say I was assuming the valve was designed to be as close to perfectly linear as possibly could be, which is to say I was assuming no valve can be or is perfectly linear, which is to say the target point for the best linearity achievable is some value less than perfect.
But there's less than perfect and there's less than perfect. If Tube 1 departs from linearity in a way Tube 2 does not and Tube 2 departs from linearity in a way Tube 1 does not---we can fairly assume this happens---and if certain of those operating differences between Tubes 1 and 2 are themselves variable---variable or non-linear non-linearities---paralleling Tubes 1 and 2 will linearise the overall operation of a parallel-tube-stage closer to the operationally achievable ideal.
> 2 Tubes
I've heard opposing opinions on the sonic effect of paralleling two tubes. Some like the effect, some don't. One view of this difference of opinion---difference in value judgment?---is that those who like two-tube paralleling feel more is gained by paralleling than lost, while those who don't like it feel more is lost than gained. But what *is* gained, if anything? And how might one maximise benefit vs detriment in this regard?
My question thus goes beyond paralleling-a-deux to ask whether paralleling three, four, five ... ten tubes---a menage a dix!---might hold ground to single-tube operation. OTL amplifiers running multiple tubes on the output are anything but blurry and obscuring, so I don't think parallel operating per se is the culprit. Two-tube paralleling might be an exception.
analog_sa said:
I've heard opposing opinions on the sonic effect of paralleling two tubes. Some like the effect, some don't. One view of this difference of opinion---difference in value judgment?---is that those who like two-tube paralleling feel more is gained by paralleling than lost, while those who don't like it feel more is lost than gained. But what *is* gained, if anything? And how might one maximise benefit vs detriment in this regard?
My question thus goes beyond paralleling-a-deux to ask whether paralleling three, four, five ... ten tubes---a menage a dix!---might hold ground to single-tube operation. OTL amplifiers running multiple tubes on the output are anything but blurry and obscuring, so I don't think parallel operating per se is the culprit. Two-tube paralleling might be an exception.
serengetiplains said:
Averaging only works if the errors are random. This is not the case here where the tubes are nonlinear in a similar way.
You will average the small differences in nonlinearity but this will most likely lead to higher order distortions instead. You can't get rid of the fundamental nonlinearity.
Jax said:
I have no aspirations to eliminate fundamental non-linearity. But let's be clear: tubes are only ever similarly non-linear, not identically so. I'm questioning whether averaging differences as do exist tube to tube---being presumably combinations of random departures from ideal practical linearity, or in other words departures from fundamental non-linearity constituting "random errors" or second-order non-linearity---renders an overall more faithful sonic presentation as a result .... call it the OTL effect? I mean, I've heard enough comments that this or that many-paralleled tube amp sounds *great* to think part of any such greatness claimed, such as it may be, derives from the parallel design.
Which leads me to question whether the typical SE topology might be improved by paralleling a large number of tubes on the output (and by extension on the driver and in the PSU .... ah, the burden of trying to improve).
Hi,
Putting n-tubes in parallel essentially creates a new tube with n times the transconductance of the single tube, Ri is divided by n as is Req.
Cgin is multiplied by n, mu is 1 times n and for the same power output n times that tube will distort noticibly less although not in a perfectly linear relation to a single tube.
Total output power won't be n times the single tube either but it'll come close. (separate envelopes helps here but brings other problems)
Zout will also be close to n divided by Zout of the single tube.
Also current drive capability will also have increased in a similar manner.
So, yes, in theory we'll have a supertube but there's a catch:
In order for all this to work as expected all tubes need to be as identical as possible, age in the same way and of course need to share current in the same way ( if they're really identical they of course will do all that).
Having individual bias supplies and current shatring plate resistors can help greatly.
The fact that no two tubes are perfectly identical or age in the same way, plus the augmented stray interlectrode capacitances explains why we often prefer to avoid putting tubes or sections of identical tubes in parallel.
Will it be n times as linear as a single tube?
No, but it will be far more linear than a single one.
Why does one put several tubes in parallel in an outputstage?
Simply because it yields more power and lowers Zout allowing for a better adapatation to the load (speaker or xfmr) prior to the application of global negative feedback.
With OTL amps there no escaping this as there's no single triode with an internal resistance small enough to drive a load varying from, say 2-16 Ohm.
What makes the classic OTL such a linear circuit is mostly the SEPP outputstage wich, when traced, exhibits curves that are evenly spaced and almost vertical lines.
Which also puts the circuit on the brink of instability, BTW.
This linearity is further enhanced by the use of global NFB which reduces distortion further and is also necessary to reduce the output impedance of the amplifier.
That low mu (~2) triodes or penthodes are used means that the added millercapacitance won't have a great penalty on bandwidth
and the powertriodes and sweeppenthodes are well know to be capable of delivering highish amounts of current making them the ideal choice to drive a modern loudspeaker as far as tubes go.
Cheers,
Putting n-tubes in parallel essentially creates a new tube with n times the transconductance of the single tube, Ri is divided by n as is Req.
Cgin is multiplied by n, mu is 1 times n and for the same power output n times that tube will distort noticibly less although not in a perfectly linear relation to a single tube.
Total output power won't be n times the single tube either but it'll come close. (separate envelopes helps here but brings other problems)
Zout will also be close to n divided by Zout of the single tube.
Also current drive capability will also have increased in a similar manner.
So, yes, in theory we'll have a supertube but there's a catch:
In order for all this to work as expected all tubes need to be as identical as possible, age in the same way and of course need to share current in the same way ( if they're really identical they of course will do all that).
Having individual bias supplies and current shatring plate resistors can help greatly.
The fact that no two tubes are perfectly identical or age in the same way, plus the augmented stray interlectrode capacitances explains why we often prefer to avoid putting tubes or sections of identical tubes in parallel.
Will it be n times as linear as a single tube?
No, but it will be far more linear than a single one.
Why does one put several tubes in parallel in an outputstage?
Simply because it yields more power and lowers Zout allowing for a better adapatation to the load (speaker or xfmr) prior to the application of global negative feedback.
With OTL amps there no escaping this as there's no single triode with an internal resistance small enough to drive a load varying from, say 2-16 Ohm.
What makes the classic OTL such a linear circuit is mostly the SEPP outputstage wich, when traced, exhibits curves that are evenly spaced and almost vertical lines.
Which also puts the circuit on the brink of instability, BTW.
This linearity is further enhanced by the use of global NFB which reduces distortion further and is also necessary to reduce the output impedance of the amplifier.
That low mu (~2) triodes or penthodes are used means that the added millercapacitance won't have a great penalty on bandwidth
and the powertriodes and sweeppenthodes are well know to be capable of delivering highish amounts of current making them the ideal choice to drive a modern loudspeaker as far as tubes go.
Cheers,
paralleling technique
A common technique used for paralleling devices in solid state designs (although not generally seen in audio designs) is to use a center tapped inductor connected between the emitters, sources, or in this case cathodes. The center tap point is then used as the common terminal. This has the advantage of providing some local feedback to equalize the currents in the two devices by tipping the cathode voltages in such a way as to equalize currents drawn. No or only a small airgap is required in the inductor usually, due to the balanced currents. The inductor must be able to handle the grid drive voltage swing necessary at the lowest frequency to correct the imbalances without saturating. These are usually small toroid inductors. The winding inductance should be higher than 1/gm of the tubes at the lowest frequency, the higher the better.
More than two devices can be handled by building a pyramid of pairs. For example, for four tubes, each set of two tubes gets one center tapped inductor, then a third center tapped inductor is connected between the center taps of the first two inductors. Its center tap then becomes the common terminal.
Don
A common technique used for paralleling devices in solid state designs (although not generally seen in audio designs) is to use a center tapped inductor connected between the emitters, sources, or in this case cathodes. The center tap point is then used as the common terminal. This has the advantage of providing some local feedback to equalize the currents in the two devices by tipping the cathode voltages in such a way as to equalize currents drawn. No or only a small airgap is required in the inductor usually, due to the balanced currents. The inductor must be able to handle the grid drive voltage swing necessary at the lowest frequency to correct the imbalances without saturating. These are usually small toroid inductors. The winding inductance should be higher than 1/gm of the tubes at the lowest frequency, the higher the better.
More than two devices can be handled by building a pyramid of pairs. For example, for four tubes, each set of two tubes gets one center tapped inductor, then a third center tapped inductor is connected between the center taps of the first two inductors. Its center tap then becomes the common terminal.
Don
Well if anyone should know anything about superparallel amps it's me... any of you guys in the southern WI/northern IL area can come by with your favorite speakers and listen to Hept'AU7, drop me a line. Personally it's not very remarkable but I can probably fix that by dropping the IST for a 6SN7 CF driver and maybe a better OPT. And better speakers. And a better room. And...
Tim
P.S. Frank, mu remains constant.
Personally it's not very remarkable but I can probably fix that by dropping the IST for a 6SN7 CF driver and maybe a better OPT. And better speakers. And a better room. And... Tim
P.S. Frank, mu remains constant.
"wouldnt paralleling lots of push pulls approximate to linearness? But then people hate push pulls let alone paralleling them so i guess absolute linearity isn't one of the bigger aims in getting a nice sound"
If you have a non-linear tube in a circuit, you can not make the circuit linear by simply adding more of the same non-linear tubes.
You can improve the linearity by carefully adding the proper amount of inverse feedback.
If you have a non-linear tube in a circuit, you can not make the circuit linear by simply adding more of the same non-linear tubes.
You can improve the linearity by carefully adding the proper amount of inverse feedback.
as i understand it (most probably wrong ) is that in pp the nonlinearity of one tube is cancelled by the opposite tube which is working in the opposite direction (and therefore in the opposite side of its non-linearness...cant think of any other ways of putting it, so the transformer would effectively see a linear device), but this assumes that the tubes work identically, though non linearly. the idea of paralleling devices to achieve a closer approximation to a single nominal specification ought to make the top and bottom halves of a pp circuit closer to identical (i think). might take more than just a few pairs though. i thought that was why pp amps had lower 2nd harmonic distortion in general. i think i may be getting all mixed up here though
Steve
) is that in pp the nonlinearity of one tube is cancelled by the opposite tube which is working in the opposite direction (and therefore in the opposite side of its non-linearness...cant think of any other ways of putting it, so the transformer would effectively see a linear device), but this assumes that the tubes work identically, though non linearly. the idea of paralleling devices to achieve a closer approximation to a single nominal specification ought to make the top and bottom halves of a pp circuit closer to identical (i think). might take more than just a few pairs though. i thought that was why pp amps had lower 2nd harmonic distortion in general. i think i may be getting all mixed up here thoughSteve
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.