I know this question may open a can of worms, but I'm curious...
I've been doing a lot of modeling of different amp topologies in LTspice, to see what kind of performance it predicts. I know it's not very accurate at predicting harmonic distortion in a circuit, but I figure it can show me gross differences between circuits.
One of the things I'm finding (which should have been obvious) is that an LTP with less than perfect balance (e.g., one with a resistor in its tail) will have much more 2nd harmonic distortion than the same LTP with a current sink in its tail.
If you get a push-pull amp really well balanced, you can just about erase all 2nd harmonic distortion, leaving the 3rd (and 5th and 7th as level goes up).
My question is, if you deliberately design in just enough 2nd harmonic distortion that 2nd harmonic is let's say -60dB and 3rd harmonic is -75dB, does that sound much different from the same circuit changed so that 2nd harmonic is -90dB and 3rd harmonic is still -75dB? With numbers this low, would a careful listener notice a difference?
Another way to put it is, does adding a bit of 2nd harmonic to a PP amp give it a more 'single ended' sound?
Does a clean amp with really low 2nd harmonic (let's say -90dB) and pretty good 3rd harmonic (let's say -75dB) sound somehow 'thin' or 'bright' because of the lack of even harmonics? Or would it just sound 'clean'?
I'm curious to know if those of you who have a lot of experience have been down this road. Thanks for any thoughts on this you want to share.
--
I also noticed that as you tweak a particular push-pull amp circuit for lower distortion, 2nd and 4th harmonic go down, but it's very difficult to reduce 3rd and 5th open loop (NFB does reduce those harmonics). It seems some tubes just generate odd order harmonics, especially output tubes biased in Class AB. Does that sound valid?
--
I've been doing a lot of modeling of different amp topologies in LTspice, to see what kind of performance it predicts. I know it's not very accurate at predicting harmonic distortion in a circuit, but I figure it can show me gross differences between circuits.
One of the things I'm finding (which should have been obvious) is that an LTP with less than perfect balance (e.g., one with a resistor in its tail) will have much more 2nd harmonic distortion than the same LTP with a current sink in its tail.
If you get a push-pull amp really well balanced, you can just about erase all 2nd harmonic distortion, leaving the 3rd (and 5th and 7th as level goes up).
My question is, if you deliberately design in just enough 2nd harmonic distortion that 2nd harmonic is let's say -60dB and 3rd harmonic is -75dB, does that sound much different from the same circuit changed so that 2nd harmonic is -90dB and 3rd harmonic is still -75dB? With numbers this low, would a careful listener notice a difference?
Another way to put it is, does adding a bit of 2nd harmonic to a PP amp give it a more 'single ended' sound?
Does a clean amp with really low 2nd harmonic (let's say -90dB) and pretty good 3rd harmonic (let's say -75dB) sound somehow 'thin' or 'bright' because of the lack of even harmonics? Or would it just sound 'clean'?
I'm curious to know if those of you who have a lot of experience have been down this road. Thanks for any thoughts on this you want to share.
--
I also noticed that as you tweak a particular push-pull amp circuit for lower distortion, 2nd and 4th harmonic go down, but it's very difficult to reduce 3rd and 5th open loop (NFB does reduce those harmonics). It seems some tubes just generate odd order harmonics, especially output tubes biased in Class AB. Does that sound valid?
--
Last edited:
Very good questions. I am sorry i cannot contribute but i'm very eager to hear
others experience.
Let the answers come !
others experience.
Let the answers come !
Many questions, some answers..
A perfect LTP cancels all even order harmonics, and that's what you notice in ltspice, when using a strong ccs in the tail. Any imperfections, even small ones, throw off the balance, and even harmonics creep back in. These can be, a tail resisto, weak ccs, unmatched triodes (always the case in real life..).
Anyway, the point of balanced PP designs is to keep even harmonics low, while accepting odd harmonics in the open loop. The odd harmonics are then suppressed by nfb (which works much more effectively in case the even harmonics have already been mostly removed by the balanced open loop design).
Now to your question. It just doesn't make sense to design in some 2nd order harmonics into a PP design(*). All there is will be mostly cancelled again at latest in the OPT, so any even order harmonic will just pollute the sound. No, it will not sound more single-ended.
(*) There is a trade-off whether to accept some even order harmonics if you can achieve less odd order in the pre-stages.
Rgds,
GB
A perfect LTP cancels all even order harmonics, and that's what you notice in ltspice, when using a strong ccs in the tail. Any imperfections, even small ones, throw off the balance, and even harmonics creep back in. These can be, a tail resisto, weak ccs, unmatched triodes (always the case in real life..).
Anyway, the point of balanced PP designs is to keep even harmonics low, while accepting odd harmonics in the open loop. The odd harmonics are then suppressed by nfb (which works much more effectively in case the even harmonics have already been mostly removed by the balanced open loop design).
Now to your question. It just doesn't make sense to design in some 2nd order harmonics into a PP design(*). All there is will be mostly cancelled again at latest in the OPT, so any even order harmonic will just pollute the sound. No, it will not sound more single-ended.
(*) There is a trade-off whether to accept some even order harmonics if you can achieve less odd order in the pre-stages.
Rgds,
GB
Assuming single-ended sound is a similar concept to "tube sound" then ...yes.... no... maybe?
It would definitely distort more so it would sound fuzzier and less detailed. Add to that, the "order" of the harmonics begins to loose more of its importance when practical signals begin to introduce any intermodulation distortion. Less ANY distortion - > Less IM distortion.
If you assume that the reason people like SE is the 2nd order distortion (which many of them deny) then adding some to a PP will make it sound more SE-like.rongon said:
No. All active devices generate distortion at all orders, almost always falling in level with increased order. PP cancels even order distortion generated in the output stage (but not, of course, any generated earlier in the amp).
Like DF96 noted, it's not exactly like that.
Distortion will practically always generate harmonics of all types, but the underlying issue is akin to what harmonics have highest amplitude, "dominance".
Then you have circuit concepts like "push-pull" where each half of the push-pull can greatly decrease, or "cancel", even order harmonic distortion produced by the other half.
And then you have devices like beam tetrodes (e.g. 6L6 tube), which reputedly were devised to produce lesser amounts of odd distortion than their successors. (The developers didn't devote so much effort for minimizing even order distortion because they knew push-pull circuit could be used).
So, distortion is largely generated in all orders, but some circuits, tubes devices, etc. definitely have a clear "pattern" in their harmonic distortion and it can be easily observed that there is a tendency for them to produce distortion most prominently at specific harmonics.
And like I said, throw in the concept of intermodulation distortion and you have to re-evaluate what "even" or "odd" order harmonic even mean.
Last edited:
You could insert a trimmer potentiometer into your long tailed pair input stage, and adjust it to get perfect balance. Theoretically this will reduce even order harmonics to their lowest possible value. Measure the "distortion profile" (per the title of this thread). Listen to the amplifier. Make careful listening notes.
Now, you can adjust the potentiometer again, and deliberately set it for more even order harmonics. Measure the "distortion profile" again. Listen to the amplifier again. Make careful listening notes.
You can repeat this as often as you like, with increasingly more and more even order harmonics. Maybe you'll find a sweet spot "X", where X amount of even order harmonics sounds better to you, than any other amount of even order harmonics. In particular, X amount sounds better than Zero amount. Wouldn't that be a delightful result?
If you use Google to find the circuit schematic of the very well regarded B&K ST-140 power amp (here is Stereophile's glowing review, written by J. Gordon Holt), you will find that it includes this very potentiometer in this very position. Perhaps the secret to its good sound is/was, that the factory adjusted it for X amount of even order harmonics (where X > Zero) ?? It might be true!
Now, you can adjust the potentiometer again, and deliberately set it for more even order harmonics. Measure the "distortion profile" again. Listen to the amplifier again. Make careful listening notes.
You can repeat this as often as you like, with increasingly more and more even order harmonics. Maybe you'll find a sweet spot "X", where X amount of even order harmonics sounds better to you, than any other amount of even order harmonics. In particular, X amount sounds better than Zero amount. Wouldn't that be a delightful result?
If you use Google to find the circuit schematic of the very well regarded B&K ST-140 power amp (here is Stereophile's glowing review, written by J. Gordon Holt), you will find that it includes this very potentiometer in this very position. Perhaps the secret to its good sound is/was, that the factory adjusted it for X amount of even order harmonics (where X > Zero) ?? It might be true!
Thanks for your answers, much appreciated.
The reason I bring this up is that I'm playing with an old beater integrated amp EICO ST40 that has ruined controls and switches, but the chassis and transformers are still good. About 15 years ago I'd built a simple integrated amp into it. I got it back to take a look and was horrified at the dumb mistakes I'd made. It appears I've learned a lot in the last several years.
After posing lots of questions and doing lots of simulations, I came up with a circuit I think will sound nice. It's a 12AX7 common cathode stage RC coupled to a 12AU7 cathodyne phase splitter, RC coupled to push-pull 6L6 triodes. Since I'm not going for max power, and the driver stage pulls a paltry 7mA of plate current per channel, I've dropped the B+ down to about 400VDC, used cathode bias on the output stage (-30V g-k), and raised the output stage plate current to 60mA per 6L6. The bonus is that the 6L6 triode with 370V plate volts and 60mA plate current looks like it will yield low odd order harmonics, especially compared to what it looks like at 450V plate and 48mA current (-40V g-k). My goal is to make a clean 10 watts per channel (Class A).
In simulation (at least), this particular circuit makes 12 watts max at very low distortion, and relatively low 3rd harmonic in particular. 2nd harmonic is higher than 3rd at lower levels (1W to 5W), then 3rd harmonic goes higher than 2nd harmonic above 5W.
I've also cooked up a "Mullard" style driver that makes a surprisingly low level of THD. It makes about the same level of 3rd harmonic distortion than the cathodyne-based circuit, but with an active CCS in its tail it makes almost zero 2nd harmonic, staying very low at all levels up to clipping. 3rd harmonic and 5th harmonic rise as output level goes up.
Of course all these glowing numbers will degrade in real life, due to mismatched or weak tubes, etc.
I've wondered why you rarely see a cathodyne phase splitter directly driving a pair of 6L6 or EL34 outputs. But then again, commercial amps are designed for max output power. Who would want a 12 watt amp that burns up 45 watts in plate current alone?
Mullard-style drivers are common, probably because it's easy to make one that supplies lots of gain which can be used for negative feedback around the whole circuit. I'm planning to use only -6dB of NFB, so I don't need quite that much gain.
I think I'll go ahead and build this circuit with the cathodyne phase splitter, just so I can hear it. If it's horrible I'll rip everything out and start over again. 😀
--
The reason I bring this up is that I'm playing with an old beater integrated amp EICO ST40 that has ruined controls and switches, but the chassis and transformers are still good. About 15 years ago I'd built a simple integrated amp into it. I got it back to take a look and was horrified at the dumb mistakes I'd made. It appears I've learned a lot in the last several years.
After posing lots of questions and doing lots of simulations, I came up with a circuit I think will sound nice. It's a 12AX7 common cathode stage RC coupled to a 12AU7 cathodyne phase splitter, RC coupled to push-pull 6L6 triodes. Since I'm not going for max power, and the driver stage pulls a paltry 7mA of plate current per channel, I've dropped the B+ down to about 400VDC, used cathode bias on the output stage (-30V g-k), and raised the output stage plate current to 60mA per 6L6. The bonus is that the 6L6 triode with 370V plate volts and 60mA plate current looks like it will yield low odd order harmonics, especially compared to what it looks like at 450V plate and 48mA current (-40V g-k). My goal is to make a clean 10 watts per channel (Class A).
In simulation (at least), this particular circuit makes 12 watts max at very low distortion, and relatively low 3rd harmonic in particular. 2nd harmonic is higher than 3rd at lower levels (1W to 5W), then 3rd harmonic goes higher than 2nd harmonic above 5W.
I've also cooked up a "Mullard" style driver that makes a surprisingly low level of THD. It makes about the same level of 3rd harmonic distortion than the cathodyne-based circuit, but with an active CCS in its tail it makes almost zero 2nd harmonic, staying very low at all levels up to clipping. 3rd harmonic and 5th harmonic rise as output level goes up.
Of course all these glowing numbers will degrade in real life, due to mismatched or weak tubes, etc.
I've wondered why you rarely see a cathodyne phase splitter directly driving a pair of 6L6 or EL34 outputs. But then again, commercial amps are designed for max output power. Who would want a 12 watt amp that burns up 45 watts in plate current alone?
Mullard-style drivers are common, probably because it's easy to make one that supplies lots of gain which can be used for negative feedback around the whole circuit. I'm planning to use only -6dB of NFB, so I don't need quite that much gain.
I think I'll go ahead and build this circuit with the cathodyne phase splitter, just so I can hear it. If it's horrible I'll rip everything out and start over again. 😀
--
It seems to be fairly generally accepted that some low level 2nd harmonic can sweeten up "simple" vocal music, but gets into IMD trouble with complex music.
Putting the adjustable pot into one P-P splitter load would allow adjusting the 2nd harmonic, but it only works cleanly for a class A amp. Class AB would put a sudden discontinuity into the gain, raising higher harmonics too.
There is a means of adjusting 3rd harmonic (and other odd harmonics to some extent) out of LTP circuits using a specific value of tail resistance (at least for triodes, should work for pentodes at higher current, the V to I power law must be below 2.0) (a paper by M.V. Kiebert "System Design Factors for Audio Amplifiers") But a tail resistance can introduce 2nd harmonic if the drive signals are not already balanced.
I would suggest that the most audible improvement can be gained by just using some "local" or "partial" feedback around the output stage to lower output Z into the OT. (not necessarily UL mode though, although that gives some benefit, but can be bettered. Simple Schade or back to driver stage feedback preferred.) The Citation II did this. That cleans up the most distorting devices (including the OT magnetizing current, shunt winding capacitance, and core permeability variation) without triggering the stability problems of global feedback with the OT in the loop. Some moderate (safe level) global feedback can be added if damping factor needs further improvement. (to overcome OT winding resistance and leakage L)
Putting the adjustable pot into one P-P splitter load would allow adjusting the 2nd harmonic, but it only works cleanly for a class A amp. Class AB would put a sudden discontinuity into the gain, raising higher harmonics too.
There is a means of adjusting 3rd harmonic (and other odd harmonics to some extent) out of LTP circuits using a specific value of tail resistance (at least for triodes, should work for pentodes at higher current, the V to I power law must be below 2.0) (a paper by M.V. Kiebert "System Design Factors for Audio Amplifiers") But a tail resistance can introduce 2nd harmonic if the drive signals are not already balanced.
I would suggest that the most audible improvement can be gained by just using some "local" or "partial" feedback around the output stage to lower output Z into the OT. (not necessarily UL mode though, although that gives some benefit, but can be bettered. Simple Schade or back to driver stage feedback preferred.) The Citation II did this. That cleans up the most distorting devices (including the OT magnetizing current, shunt winding capacitance, and core permeability variation) without triggering the stability problems of global feedback with the OT in the loop. Some moderate (safe level) global feedback can be added if damping factor needs further improvement. (to overcome OT winding resistance and leakage L)
Last edited:
There are examples of musical instrument amplifiers (well, amplified "effect processors" really) where balance of phase inverter's output signals is even an user-adjustable control. With perfect balance driving power tubes the push-pull operation is theoretically "ideal", with drive cut off from the other tube in push-pull pair the operation is effectively single-ended.
Since clipping of a single-ended amplifier (especially that of a transformer coupled power pentode) is generally highly asymmetric distortion products have significant amounts of even order harmonic distortion. In practice, even order harmonic distortion is merely inherent attribute of any signal asymmetry.
A single-ended amp can be tweaked for more symmetric clipping by balancing DC bias points. With tubes this is inherently somewhat difficult because of their different characteristics to "saturate" and "cut off", and because of different reactions of surrounding circuitry to interact with such states.
With effect processors like guitar amps it is quite customary to use other than perfect "center balance" bias because the resulting asymmetric clipping not only creates higher amount of even order harmonic distortion, but also causes intentional dynamic shifts to signal's DC offset level, and hence dynamically and interactively affects bias / clipping thresholds of various stages within the signal processing chain.
Since push-pull scheme effectively cancels or decreases its own even order distortion then by principle nature of its distortion must also be more of symmetric type. Symmetric clipping does not upset signal's DC offset level in similar manner to asymmetric clipping.
But what's even and odd in bigger context? Concept of harmonic distortion is easy: Take, say, a 1kHz sinusoidal wave. 1 kHz is the fundamental frequency, so 2nd order distortion is 2kHz, 3rd order 3kHz and so on. "Dubious" harmonic distortion products like 7th and 9th harmonic, and so on, are, by common theory, considered to have no pleasant musical interval with the fundamental frequency and sound "off tune" and "offending".
Ok. ....But how about when we discuss intermodulation distortion of a tone consisting of strong, say, 1kHz and 9 kHz components. You can calculate the wealth of intermodulation distortion products any "even" or "odd" order distortion in either component would produce and it's not a nice thought. But yet we sense IM distortion marvelously better than harmonic distortion. And we never listen to single frequencies in practice. So what gives?
In regard of making -an effect processor- a single-ended amplifier will generate more of ALL order harmonic distortion than a push-pull amp. Generated distortion is both odd and even order but SE can't reduce even order distortion like PP. They will archetypically also clip in distinctly asymmetric manner, which means even order harmonic distortion products have higher amplitude. This is an easy task to accomplish signal-processing wise.
Any increased harmonic distortion will, naturally, make IM distortion higher and in large extent distinctly degrade signal quality. So by recreating the SE performance you give more opportunity for overall IMD to "mush" up the signal into unintelligible mess.
Since clipping of a single-ended amplifier (especially that of a transformer coupled power pentode) is generally highly asymmetric distortion products have significant amounts of even order harmonic distortion. In practice, even order harmonic distortion is merely inherent attribute of any signal asymmetry.
A single-ended amp can be tweaked for more symmetric clipping by balancing DC bias points. With tubes this is inherently somewhat difficult because of their different characteristics to "saturate" and "cut off", and because of different reactions of surrounding circuitry to interact with such states.
With effect processors like guitar amps it is quite customary to use other than perfect "center balance" bias because the resulting asymmetric clipping not only creates higher amount of even order harmonic distortion, but also causes intentional dynamic shifts to signal's DC offset level, and hence dynamically and interactively affects bias / clipping thresholds of various stages within the signal processing chain.
Since push-pull scheme effectively cancels or decreases its own even order distortion then by principle nature of its distortion must also be more of symmetric type. Symmetric clipping does not upset signal's DC offset level in similar manner to asymmetric clipping.
But what's even and odd in bigger context? Concept of harmonic distortion is easy: Take, say, a 1kHz sinusoidal wave. 1 kHz is the fundamental frequency, so 2nd order distortion is 2kHz, 3rd order 3kHz and so on. "Dubious" harmonic distortion products like 7th and 9th harmonic, and so on, are, by common theory, considered to have no pleasant musical interval with the fundamental frequency and sound "off tune" and "offending".
Ok. ....But how about when we discuss intermodulation distortion of a tone consisting of strong, say, 1kHz and 9 kHz components. You can calculate the wealth of intermodulation distortion products any "even" or "odd" order distortion in either component would produce and it's not a nice thought. But yet we sense IM distortion marvelously better than harmonic distortion. And we never listen to single frequencies in practice. So what gives?
In regard of making -an effect processor- a single-ended amplifier will generate more of ALL order harmonic distortion than a push-pull amp. Generated distortion is both odd and even order but SE can't reduce even order distortion like PP. They will archetypically also clip in distinctly asymmetric manner, which means even order harmonic distortion products have higher amplitude. This is an easy task to accomplish signal-processing wise.
Any increased harmonic distortion will, naturally, make IM distortion higher and in large extent distinctly degrade signal quality. So by recreating the SE performance you give more opportunity for overall IMD to "mush" up the signal into unintelligible mess.
- Status
- Not open for further replies.