Frank Berry said:
I'd be willing to bet money you're wrong on that one.
EDIT
I mean, I bet paralleled tubes amplify more linearly than a single tube.
parallel tubes
Putting more tubes in parallel will just get the deviations between the tubes to average out, but you will be left with the design average which is generally not linear. In fact, tubes are 3/2 power devices, not linear at all, triodes manage to come close to linear via plate feedback with a high impedance load, but as soon as you connect a real load they are not linear either. Pentodes require feedback networks to get any linearity.
P-P designs only get the even harmonics to cancel out, leaving the odd harmonics, which many people seem to consider the most damaging to the sound.
This leaves one with only a few choices: feedback, inherent feedback (like triode or cathode follower) , unity error feedback (like the Hawksford error correction scheme), class A with very limited signal swing (or conversely very high idle current) or complementary predistortion.
Don
Putting more tubes in parallel will just get the deviations between the tubes to average out, but you will be left with the design average which is generally not linear. In fact, tubes are 3/2 power devices, not linear at all, triodes manage to come close to linear via plate feedback with a high impedance load, but as soon as you connect a real load they are not linear either. Pentodes require feedback networks to get any linearity.
P-P designs only get the even harmonics to cancel out, leaving the odd harmonics, which many people seem to consider the most damaging to the sound.
This leaves one with only a few choices: feedback, inherent feedback (like triode or cathode follower) , unity error feedback (like the Hawksford error correction scheme), class A with very limited signal swing (or conversely very high idle current) or complementary predistortion.
Don
Yes, PP cancels the second order nonlinearities. Third order, and all other odd errors for that matter, are still present however. Theoretically, you might be able to preamplify a signal to give it inverse distortions which cancel with those in the power amp, but that might make IMD up the ying-yang or something.
Oh. You just mentioned that. You people are phasing me out!!! 🙄 😛
Tim
Oh. You just mentioned that. You people are phasing me out!!! 🙄 😛
Tim
serengetiplains said:
Come by and measure Hept'AU7. Especially with bias backed off, I dare you to find any linearities for large signals in any half of the PP stage.
The simplest way to put it is "paralleling nonlinearity with more of the same nonlinearity is still going to produce the SAME DANG NONLINEARITY". Sorry if you can't comprehend that, but the universe is what it is due to simple logic as this.
It'll average out inconsistencies between tubes but it isn't known if those alone are detrimental or advantageous. (Statistics and averages. Some will fall above and below the average value.) Heck, you may well have several exceedingly linear tubes in the bunch that are being wasted!
Tim
Damn Statistics
No dang need to be sorry, Tim, it's I don't think any two tubes have the *same* non-linearity, which leaves the question whether averaging operational differences using a multi-tube setup produces greater linearity. I'm betting it does. Think: throw ten darts onto a dart board; now average the location of each dart relative to the other .... seems to me the "average" gets you closer to the bulls-eye which, in our case, is best-case non-linearity. Just my view of the statistics of things.
Sch3mat1c said:
No dang need to be sorry, Tim, it's I don't think any two tubes have the *same* non-linearity, which leaves the question whether averaging operational differences using a multi-tube setup produces greater linearity. I'm betting it does. Think: throw ten darts onto a dart board; now average the location of each dart relative to the other .... seems to me the "average" gets you closer to the bulls-eye which, in our case, is best-case non-linearity. Just my view of the statistics of things.
As the valves age, a number of them will change characteristics in a non-typical way. This will alter overall linearity and cause distortion to increase. I think you'll need to keep a valve tester next to the amp 😀
You stubbornly refuse to read or comprehend the page above and your analogy is completely off the mark. If you like the sound of paralleled tubes that's fine but trying to invent quasi statistical explanations is simply no good.
really only completely wrong
Nice use of adverbs. Using adverbs is among the first things one is taught to quit in Writing 101 as adverbs replace persuasive explanation, which requires hard work, with emphatic-ness, which doesn't. I'll show you. By *completely* off the mark, do you really mean I'm a bright, clear 100% wrong? (C'mon, is anybody ever clearly 100% wrong?)
My sense is that multi-tube amps can have a certain clarity that might be attributable, perhaps in part, to multi-tube operation. The best hypothesis I can conjure suggests that amusical quirks of a given tube blah blah blah.
analog_sa said:
Nice use of adverbs. Using adverbs is among the first things one is taught to quit in Writing 101 as adverbs replace persuasive explanation, which requires hard work, with emphatic-ness, which doesn't. I'll show you. By *completely* off the mark, do you really mean I'm a bright, clear 100% wrong? (C'mon, is anybody ever clearly 100% wrong?)
My sense is that multi-tube amps can have a certain clarity that might be attributable, perhaps in part, to multi-tube operation. The best hypothesis I can conjure suggests that amusical quirks of a given tube blah blah blah.
Hi,
IMO, if there were to exist a single tube with all the electrical properties (the good ones such as some linearity 😉 ) of those multiple parallel tubes it will quite likely sound better still.
On the other end of the scale, if we all had half decent loudspeakers with healthy effeciency, low colouration and acceptable levels of distortion we wouldn't be needing those huge power mongers anyway.
It seems to me that the mistake one tends to make is that two wrongs make a right when adding tubes in parallel.
Unfortunately nature is cruel and in a bunch of paralleled tubes the weakest link is going to determine what you'll have, after that things will only get worse as time goes by.
In my mind, the less tubes one needs to put in parallel for a preset result the better.
If we'd return to the OTL analogy, a 100W class AB amp using 4 6C33-Cs is likely to yield better results than say, 10 6080s disregarding the fact that the two tubes are of course not the same.
Cheers,😉
IMO, if there were to exist a single tube with all the electrical properties (the good ones such as some linearity 😉 ) of those multiple parallel tubes it will quite likely sound better still.
On the other end of the scale, if we all had half decent loudspeakers with healthy effeciency, low colouration and acceptable levels of distortion we wouldn't be needing those huge power mongers anyway.
It seems to me that the mistake one tends to make is that two wrongs make a right when adding tubes in parallel.
Unfortunately nature is cruel and in a bunch of paralleled tubes the weakest link is going to determine what you'll have, after that things will only get worse as time goes by.
In my mind, the less tubes one needs to put in parallel for a preset result the better.
If we'd return to the OTL analogy, a 100W class AB amp using 4 6C33-Cs is likely to yield better results than say, 10 6080s disregarding the fact that the two tubes are of course not the same.
Cheers,😉
Frank, I meant to thank you for your previous post above, so do so now.
Re your comment, North America was long ago bitten by the bigness bug: big amps, big cars, big houses, big power cords, big capacitors, big burgers, big buildings, big companies, big computer screens, big bath tubs, big foreign policy, big big big big big. Check out the new Chrysler 300 automobile: 350hp hemi in a car weighing about a billion pounds. My god.
We have almost forgotten the fine art of knowing the importance of all that is either not or is before big, like the first watt coming out of an amp.
So now you're talking down the "big complex" most of us have, yet you're the one going on about superparalleled tubes? 😉
Going back to statistics, let me try it this way. Put linearity on a scale of 0 to 10, with 0 being perfectly linear and 10 being well, nothing in particular but let's just say, really crappy. You might have five tubes under test land at 1, 4, 3, 1 and 2. The two which landed 1 are very linear, but the average (which is what you would get by paralleling them, assuming they are weighted equally, which is another can of worms) is 2.2, not as linear is it?
To produce a perfectly linear result, half would have to land at a negative number (which can be considered inverse distortion, alone a degredation but cancels with a positive distortion; and the scale would now be -10 to +10) such that the average tends towards zero.
The problem with tubes is THEY DO NOT PRODUCE AN INVERSE DISTORTION. Current must always be positive, so along that 3/2 power law, you will always have exactly ONE phase of distortion. Higher or lower current or different tube design will shift this a bit and can have good results, but no matter what, you can't change the fact that it's a 3/2 power device and will ALWAYS have this same distortion. It's just a matter of degree; you'll never find an inverse distortion because there are no positron or proton-conducting tubes.
Tim
Going back to statistics, let me try it this way. Put linearity on a scale of 0 to 10, with 0 being perfectly linear and 10 being well, nothing in particular but let's just say, really crappy. You might have five tubes under test land at 1, 4, 3, 1 and 2. The two which landed 1 are very linear, but the average (which is what you would get by paralleling them, assuming they are weighted equally, which is another can of worms) is 2.2, not as linear is it?
To produce a perfectly linear result, half would have to land at a negative number (which can be considered inverse distortion, alone a degredation but cancels with a positive distortion; and the scale would now be -10 to +10) such that the average tends towards zero.
The problem with tubes is THEY DO NOT PRODUCE AN INVERSE DISTORTION. Current must always be positive, so along that 3/2 power law, you will always have exactly ONE phase of distortion. Higher or lower current or different tube design will shift this a bit and can have good results, but no matter what, you can't change the fact that it's a 3/2 power device and will ALWAYS have this same distortion. It's just a matter of degree; you'll never find an inverse distortion because there are no positron or proton-conducting tubes.
Tim
Methinks there's a point that has yet to be made...the improvement, or lack thereof, will depend on the topology.
And on the ear of the beholder.
There's another point I don't recall seeing yet, which is that input capacitance will climb as more tubes are paralleled. Whether this becomes a problem will also depend on the topology and how much current is available upstream.
This thread shows signs of non-linear behavior, e.g. it could easily degenerate into another semantic mud-slinging contest. Whether the nature of the tube "improves" is going to depend on who is defining "improvement." I, for one, have no problem with parallel tubes, although I find the arguments for non-parallel operation interesting and may conduct a few experiments along those lines at some point in the future.
Grey
And on the ear of the beholder.
There's another point I don't recall seeing yet, which is that input capacitance will climb as more tubes are paralleled. Whether this becomes a problem will also depend on the topology and how much current is available upstream.
This thread shows signs of non-linear behavior, e.g. it could easily degenerate into another semantic mud-slinging contest. Whether the nature of the tube "improves" is going to depend on who is defining "improvement." I, for one, have no problem with parallel tubes, although I find the arguments for non-parallel operation interesting and may conduct a few experiments along those lines at some point in the future.
Grey
Hi,
Hi Grey,
Was mentioned already...by yours truly...Admittedly not in an explicite way, but still.
That too was mentioned...by yours truly.
What else is new? 😀
My rule of thumb regarding this is simple: if you don't have to, don't.
Cheers,😉
Hi Grey,
Was mentioned already...by yours truly...Admittedly not in an explicite way, but still.
That too was mentioned...by yours truly.
What else is new? 😀
My rule of thumb regarding this is simple: if you don't have to, don't.
Cheers,😉
Actually, I was thinking of paralleling small-signal tubes for a total output wattage of, say, 10. Big brings in too many considerations for my liking.
Actually, I was thinking more along the lines of paralleling to reduce small, variable non-linearities---call them deviations from text-book non-linearity, which would be the negative numbers to which you refer here:
😀
Whelp, that was my best attempt at telling you the tube-to-tube variations are above the zero line. Still hasn't sunk in, so, see ya in another thread.
Tim
Tim
Sch3mat1c said:
Tim, I *do* appreciate your responses and advice, and do consider what you say. So, thanks.
🙂
Here's my attempt then...
if you look at the picture, the straight line represents a perfectly linear device, all the other colours are actual real life non-linear devices. hopefully it's fairly clear that no matter how many real ones you have, and average, it is impossible to get to the straight line. hope this helps.
Steve
if you look at the picture, the straight line represents a perfectly linear device, all the other colours are actual real life non-linear devices. hopefully it's fairly clear that no matter how many real ones you have, and average, it is impossible to get to the straight line. hope this helps.
Steve
Attachments
Guys, I'm not being stubborn. I just don't share certain of assumptions I'm seeing in your replies. Steve, I know it's impossible to average non-linear lines into a straight line in a graph like the one you posted. I just don't think tube distortion is best modelled on a 2-factor (ie, 2D) graph, where any conceivable deviation from linear must be in one of two directions. Tubes have multiple properties and characteristics any or all of which can operate non-linearly. Many of these non-linearities are dynamic in nature and depend on other operating factors including the non-linearity of such. I'm not intuitively convinced that the sum total non-linearity of all these properties and characteristics, which "total" seems to me almost hopelessly complex, is best graphed on a 2D chart. But as soon as one admits greater than two factors into the equation, the balancing-toward-greater linearity idea begins to make some sense. This idea I attempted to picture by my bull's-eye analogy above. Actually, a better analogy would be not a bull's-eye, but a 3D cone with 4D space-time warping, if you will, thrown in for good effect.
The idea also attempts, by reach of hypothesis (ie, best-guessing), to account for my and others' observations that multi-tube amps can (not necessarily do) sound clear in a way single-tube designs don't (or have a clarity of a different type than single-tube designs attain, for lack of adequate descriptive language). As the thought then runs: I associate clarity with linearity, so how, I ask, might a multi-tube amp achieve higher linearity in a way single-tube designs cannot? The first answer that springs to mind is that multi-tube operating can (not necessarily will) linearise, somehow, and to some degree, or in some manner perhaps not otherwise attainable, the net overall operation. All I am doing by this thought string is inducing theory from this latter, known observation.
None other than John Broskie, if the attribution is correct, called the "sound" of multi-tube operating paradoxical. Actually, it was Broskie's comment, quoted in an AA thread and attributed to him, that prompted me to post this thread. Here's the comment:
hello,
i think the only reason the 6sn7's would be paralleled would be to get sufficient power, but the tubes may be fundamentally more linear than the normally used power output tubes. the only way that parallel tubes could ever end up working overal linearly would be if the tubes linearity was random, which it isnt. they are all non-linear in the same direction.
i think the only reason the 6sn7's would be paralleled would be to get sufficient power, but the tubes may be fundamentally more linear than the normally used power output tubes. the only way that parallel tubes could ever end up working overal linearly would be if the tubes linearity was random, which it isnt. they are all non-linear in the same direction.
Serengetiplains: Can I give you a practical example that may help your conceptual understanding?
It is possible to make a triode having identical values of mu, gm and ra using coarsely-spaced thick grid wires, or closely-spaced fine grid wires. The resulting anode characteristics can be overlaid, and at reasonable currents, the curves are identical. At low currents, the curves for the triode with coarsely-spaced thick grid wires show a significant deviation from linearity. This shows a design failing, and paralleling lots of these valves will only yield a closer approximation to the non-linearity caused by this design failing.
It is possible to make a triode having identical values of mu, gm and ra using coarsely-spaced thick grid wires, or closely-spaced fine grid wires. The resulting anode characteristics can be overlaid, and at reasonable currents, the curves are identical. At low currents, the curves for the triode with coarsely-spaced thick grid wires show a significant deviation from linearity. This shows a design failing, and paralleling lots of these valves will only yield a closer approximation to the non-linearity caused by this design failing.
- Status
- Not open for further replies.