Beneficial or detrimental? Can't wrap my head around if it acts as a form of pre-correction or not.
The context is a two stage amp with about 18 dB of feedback fronted by a Pimm style CSS loaded 12GN7 driver. It swing 120 volts p-p at 0.5% distortion predominantly second. At half voltage it's ~0.2% with 3rd and 4th harmonics 80 dB or more below the fundamental, 5th and above in the noise floor.
Is it worth chasing a driver with less 2nd as well? Thanks.
The context is a two stage amp with about 18 dB of feedback fronted by a Pimm style CSS loaded 12GN7 driver. It swing 120 volts p-p at 0.5% distortion predominantly second. At half voltage it's ~0.2% with 3rd and 4th harmonics 80 dB or more below the fundamental, 5th and above in the noise floor.
Is it worth chasing a driver with less 2nd as well? Thanks.
Well you can measure the open loop distortion and find out? My feeling is cancellation is unlikely to be very good. Silk purse from two pig's ears isn't likely!
That was done, open loop is the only way to determine the driver's harmonic profile and output stage interaction into load. It's interpreting the closed loop result that is anything but clear. Loop feedback from the OPT secondary winding suppresses all harmonics as far as I can tell. FWIW this single ended UL mode amplifier is doing 25 watts at ~0.5% thd, harmonics above the 3rd at least 80 dB down.
Simple feedback from the OPT secondary winding at this stage.
2nd harmonic cancellation?
Let me show you a different method:
Suppose the loudspeaker driven by 25 Watts has 0.5% 2nd harmonic distortion.
The amplifier putting out 25 Watts is also 0.5% 2nd harmonic distortion.
Just connect the loudspeaker one way. Listen for the 2nd harmonic distortion.
Then swap the loudspeaker + and - leads. Listen for the 2nd harmonic distortion.
One of those two loudspeaker cable connections will cancel all the 2nd harmonic distortion.
Suppose the loudspeaker 2nd harmonic distortion at 25 Watts is 1%, there will be a connection that causes some reduction of the 2nd harmonic distortion.
Suppose the loudspeaker 2nd harmonic distortion at 25 Watts is only 0.2% . . . I want to know the make and model number of that loudspeaker.
Let me show you a different method:
Suppose the loudspeaker driven by 25 Watts has 0.5% 2nd harmonic distortion.
The amplifier putting out 25 Watts is also 0.5% 2nd harmonic distortion.
Just connect the loudspeaker one way. Listen for the 2nd harmonic distortion.
Then swap the loudspeaker + and - leads. Listen for the 2nd harmonic distortion.
One of those two loudspeaker cable connections will cancel all the 2nd harmonic distortion.
Suppose the loudspeaker 2nd harmonic distortion at 25 Watts is 1%, there will be a connection that causes some reduction of the 2nd harmonic distortion.
Suppose the loudspeaker 2nd harmonic distortion at 25 Watts is only 0.2% . . . I want to know the make and model number of that loudspeaker.
Ya, I've posted on that often and generally target a predominantly 2H distortion profile with a magnitude in the range and below that anticipated from speakers. Unfortunately the measurements are lost but the frequency band of effectiveness depends on loudspeaker and type. A 2-way with a phase inverse tweeter obviously doesn't offer partial cancellation across the entire spectrum. At first surprisingly though obvious in hindsight, neither do full range drivers. Their phase rotations become sporadic above the piston range.
My curiosity here though is driven by the trade offs of using the approach inside a feedback loop. Is retaining some driver 2H to cancel output stage 2H worthwhile or do the additional consequences of an H2 distorting input stage offset the cancellation benefit? Is it different than for example Benchmark's approach with their power amps?
My curiosity here though is driven by the trade offs of using the approach inside a feedback loop. Is retaining some driver 2H to cancel output stage 2H worthwhile or do the additional consequences of an H2 distorting input stage offset the cancellation benefit? Is it different than for example Benchmark's approach with their power amps?
Last edited:
I don't think you can detect 1% THD of any spectral composition if the applied power is 25W, even with 80dB/1W/1m inefficient speakers. It's physiologically below the threshold of audibility for 100% of the people. 25W means 96 dB at 1 m and there will be room effects that will prevail. If you try to move away even worse. Actually one might need to get closer than 1m......
At 90dB/1W/1m the threshold is 3% THD for most people, some might detect a small fraction less. I think accurate physiological tests when SPL gets higher than 80-85 dB are done with headphones exactly to avoid room acoustics effects. At 80 dB SPL or less the threshold is much lower and room acoustics effects too....
If I had already 18 dB GRND feedback I would not bother with cancellation at all. That's already a big amount of feedback in my book.
At 90dB/1W/1m the threshold is 3% THD for most people, some might detect a small fraction less. I think accurate physiological tests when SPL gets higher than 80-85 dB are done with headphones exactly to avoid room acoustics effects. At 80 dB SPL or less the threshold is much lower and room acoustics effects too....
If I had already 18 dB GRND feedback I would not bother with cancellation at all. That's already a big amount of feedback in my book.
A 2nd harmonic distortion audibility test was conducted.
The double blindfold test used a pure 100Hz test tone; a 2A3 SE amplifier, and a DPDT switch between the 2A3 amplifier output and the Spica TC-50 loudspeaker.
The switch was unmarked.
The test was done with medium acoustic volume, not low levels, and not at high distortion levels
One person changed the switch position back and forth multiple times, and the other person tried to hear any difference in the timbre of the two switch positions.
Then the switch operator and listener swapped roles, and ran the test again.
Success (depending on what you wanted the test result to be). Gee, did I detect a bias here?
In both cases, a difference was heard in the timbre of the sound, as the the switch position was changed.
It was not noticed which sounded better or worse, only that the difference could be heard.
If you do not trust the test results, then conduct your own test using that method.
The double blindfold test used a pure 100Hz test tone; a 2A3 SE amplifier, and a DPDT switch between the 2A3 amplifier output and the Spica TC-50 loudspeaker.
The switch was unmarked.
The test was done with medium acoustic volume, not low levels, and not at high distortion levels
One person changed the switch position back and forth multiple times, and the other person tried to hear any difference in the timbre of the two switch positions.
Then the switch operator and listener swapped roles, and ran the test again.
Success (depending on what you wanted the test result to be). Gee, did I detect a bias here?
In both cases, a difference was heard in the timbre of the sound, as the the switch position was changed.
It was not noticed which sounded better or worse, only that the difference could be heard.
If you do not trust the test results, then conduct your own test using that method.
Yeah, actual blind tests are humbling. (At least when you're young - when you get old you realize you can't hear anything anymore anyway).
All good fortune,
Chris
All good fortune,
Chris
I can just detect 7th harmonic of 100Hz at -65dB down (about 0.06%), done by generating a WAV file with intermittent harmonic turning on and off - when you can't hear when the changes are, you're at the threshold. With headphones of course, and various volume levels by experiment.
Here's my sensitivity for various harmonics of 100Hz:
2nd: -34dB
3rd: -52dB
4th: -61dB
7th: -65dB. <- best result
10th: -63dB
For 200Hz:
3rd: -57dB
5th: -63dB
500Hz
3rd: -40dB
4th: -40dB
Of course adding in speaker distortion on top of this would make it harder to detect as there would already be substantial harmonic content.
The measurements are rough and ready, as I had to generate files on the fly for each level.
Would be interesting to do the same for IMD. (I already know I can pick up IM between 19kHz and 20kHz tones as my high frequency hearing has lost those frequencies and all I hear is the IMD!)
Physiological BLIND tests have been done professionally to the death. No need to reinvent the wheel every time. Have you checked the distortion or alteration introduced by the switch? Have you checked room acoustics? Just to say that it's easy to introduce uncontrolled errors or make mistakes. I don't believe you can detect 0.5 % 2nd harmonic distortion at 90 db SPL. At 96 dB, as you suggested earlier, is even worse and impossible unless you are Batman. "Medium acoustic volume" doesn't mean anything.
Last edited:
Sensitivity to distortion is a function of SPL.
Here we are talking about 0.5% 2H distortion at 25W basically, because the rest is below 80 dB. You can't detect it.
Generally,
With all other things being equal:
A small amount of harmonic distortion caused by an amplifier, with a pure sine wave test tone applied at that amplifier input, is easier to hear . . .
Versus the same small amount of harmonic distortion caused by an amplifier, when a note from a music instrument is applied to the amplifier input.
When 2 or more instruments are playing at the same time, it maY becomes harder to hear the same low level of amplifier harmonic distortion.
Complexity tends to mask the problem.
Musical instruments create their own harmonics, and that can mask small amounts of amplifier harmonic distortion.
Multiple instruments . . . well you can imagine what happens.
"All Generalizations Have Exceptions"
Oh, is that quote a generalization?
Oxymorons anonymous
With all other things being equal:
A small amount of harmonic distortion caused by an amplifier, with a pure sine wave test tone applied at that amplifier input, is easier to hear . . .
Versus the same small amount of harmonic distortion caused by an amplifier, when a note from a music instrument is applied to the amplifier input.
When 2 or more instruments are playing at the same time, it maY becomes harder to hear the same low level of amplifier harmonic distortion.
Complexity tends to mask the problem.
Musical instruments create their own harmonics, and that can mask small amounts of amplifier harmonic distortion.
Multiple instruments . . . well you can imagine what happens.
"All Generalizations Have Exceptions"
Oh, is that quote a generalization?
Oxymorons anonymous
Last edited:
I was assuming single tones at physiological level. No instruments, no music. If we start talking about music most, if not all, measurements become irrelevant (wrong) because the listener must be included in the chain and Fourier and several other things are not valid anymore.
My concern is less 2H than the potential ramifications at higher harmonics or IM from leveraging 2H cancellation inside a feedback loop. From playing around in LTSpice with a pair of 12ax7 in series the concern might be exaggerated.
BTW, I discount all tests of distortion audibility that doesn't include accurate acoustic measurements of the test subject stimulus. So pretty much all of them to date. There are just too many confounding variables.
2nd harmonic distortion seres cancellation of an open loop 2 stage amplifier is not always what it seems.
(The driver 2nd harmonic distortion is 180 degrees different than the output stage 2nd harmonic distortion; that allows for cancellation similar to push pull cancellation).
Suppose the first stage is 0.5%, and the output stage is 5% when the speaker is 8 Ohms.
That can provide a pretty good 2nd harmonic distortion cancellation, even when the power changes up by 6 dB, and down by 6dB from the signal level that gives 0.5%.
(As long as the + 6dB and - 6dB is within the linear range of the driver and the output stage, the distortion of the two tracks quite well), so the cancellation is very good.
But what happens to the output stage 2nd harmonic distortion when the speaker load is 6 Ohms at some frequency, and 24 Ohms at another frequency, The load line steepness changes 4:1, the 2nd harmonic distortion changes with it too.
Worse yet, often the speaker load is not even resistive at some frequencies, instead it is reactive (elliptical load line).
How good is the cancellation then when the driver 2nd harmonic distortion and output 2nd harmonic distortion are no longer 180 degrees different.
Good Luck!
(The driver 2nd harmonic distortion is 180 degrees different than the output stage 2nd harmonic distortion; that allows for cancellation similar to push pull cancellation).
Suppose the first stage is 0.5%, and the output stage is 5% when the speaker is 8 Ohms.
That can provide a pretty good 2nd harmonic distortion cancellation, even when the power changes up by 6 dB, and down by 6dB from the signal level that gives 0.5%.
(As long as the + 6dB and - 6dB is within the linear range of the driver and the output stage, the distortion of the two tracks quite well), so the cancellation is very good.
But what happens to the output stage 2nd harmonic distortion when the speaker load is 6 Ohms at some frequency, and 24 Ohms at another frequency, The load line steepness changes 4:1, the 2nd harmonic distortion changes with it too.
Worse yet, often the speaker load is not even resistive at some frequencies, instead it is reactive (elliptical load line).
How good is the cancellation then when the driver 2nd harmonic distortion and output 2nd harmonic distortion are no longer 180 degrees different.
Good Luck!
Last edited:
The only way to know in my opinion is: try and see for real. Forget simulations.
One signal tube which is very good for this kind of trick is the ECC82: make it work at lower current and you will get more 2H without worsening the rest significantly. You could try with 83 by decreasing its operational current and adjusting the load to get the same swing with higher H2.
Last edited:
The term 'cancellation' implies totality and should probably be replaced with 'reduction'. As I understand it, consider a pair of series stages A>B. 'A' is ideal with no distortion, 'B' has some arbitrary level of pure 2H. Call it 10%. Now start dialing 2H into stage 'A' some small step at a time from 0%. With each increase stage 'A' will offset, or reduce, some amount of stage 'B' 2H until it reaches the point of near or total cancellation. Increase stage 'A' 2H injection past this point and 2H reappears on the A>B output, my guess is in inverse phase. As long as the stage 'A' 2H is still cancelling, from a 2H reduction perspective the effect is beneficial. Maybe a pointlessly tiny benefit, maybe an easily measured one but a beneficial reduction if that's what you're after.
BTW, that cancellation has a side consequence: increasing 3H . No free lunch. From long ago: https://www.diyaudio.com/community/threads/cancelling-harmonics.99125/post-1172271
Attachments
I did but the output tube's distortion was confounding the result. Given the first stage only drives the second's grid while the second also drives the feedback return, benching a pair of 12au7s also isn't as straightforward as first glance suggests.
Though precise distortion results haven't been typical for me simulations have been useful for revealing trends. Hence the Frankenstein sim below. Thanks for the tube suggestion, a 12au7 model wasn't at hand so a 12bh7 did service instead. The output stage mosfet buffers the second tube from the load of driving the feedback loop. It's distortion is far below the tube's. The first stage mosfet is there simply to match as close as possible the operating conditions of the second stage. Both tubes are driven by the same grid voltages and operate into the same loads.
At some point I could post the details but the results are pretty simple. Open loop the even harmonics at node02 are greater than those at node03. The odd harmonics are greater at node03 than node02. More surprising, closing the feedback loop doesn't generate new harmonics. Increasing feedback reduces all harmonics with increasing effectiveness as frequency increases.
Now this does reflect what I see on the bench, a harmonic profile emphasizing 3H. For me the take away is 2H cancellation is a tool for minimizing total measured THD or shifting the balance of harmonics from evens to odds. Looks like I'm in the hunt for a different driver tube.
Attachments
rdf,
Good point!
There is no 'cancellation' of 2nd harmonic distortion. It is a 'reduction' of 2nd harmonic distortion.
If the word cancellation is used, it should be 2 words: partial cancellation.
2nd harmonic distortion, before any negative feedback is applied or employed:
With a typical 2 stage single ended amplifier (if there is a typical):
The reduction of 2nd harmonic distortion from the partial cancellation due to the opposite phase of serial stages is that the amount of 2nd harmonic distortion tracks reasonably well, but only as long as the output stage is loaded by a resistor.
Unfortunately, many loudspeakers only "act" like a resistor only at a very few frequencies across the audio band.
That means that at some different frequencies there is partial cancellation, and at some other frequencies there is little or no cancellation.
This variable partial cancellation of second harmonic distortion that is not consistent versus frequency, might cause some undesirable audible effects.
This effect is a possibility with single ended circuits.
With a typical push pull amplifier output stage (if there is a typical):
The variable impedance versus frequency of many loudspeakers is reflected back through the output transformer.
If the output stage is Class A, that varying impedances is equal on both the push and the pull tubes.
That seems more consistant with push pull operation, versus the serial distortion reduction of single ended operation.
A fairly well established generalization says that technically: push pull is able to work better with elliptical loads, versus how well single ended amplifiers work with elliptical loads.
(technically, but not necessarily audible).
Happy hunting, no matter whether the Game is single ended, or push pull.
Good point!
There is no 'cancellation' of 2nd harmonic distortion. It is a 'reduction' of 2nd harmonic distortion.
If the word cancellation is used, it should be 2 words: partial cancellation.
2nd harmonic distortion, before any negative feedback is applied or employed:
With a typical 2 stage single ended amplifier (if there is a typical):
The reduction of 2nd harmonic distortion from the partial cancellation due to the opposite phase of serial stages is that the amount of 2nd harmonic distortion tracks reasonably well, but only as long as the output stage is loaded by a resistor.
Unfortunately, many loudspeakers only "act" like a resistor only at a very few frequencies across the audio band.
That means that at some different frequencies there is partial cancellation, and at some other frequencies there is little or no cancellation.
This variable partial cancellation of second harmonic distortion that is not consistent versus frequency, might cause some undesirable audible effects.
This effect is a possibility with single ended circuits.
With a typical push pull amplifier output stage (if there is a typical):
The variable impedance versus frequency of many loudspeakers is reflected back through the output transformer.
If the output stage is Class A, that varying impedances is equal on both the push and the pull tubes.
That seems more consistant with push pull operation, versus the serial distortion reduction of single ended operation.
A fairly well established generalization says that technically: push pull is able to work better with elliptical loads, versus how well single ended amplifiers work with elliptical loads.
(technically, but not necessarily audible).
Happy hunting, no matter whether the Game is single ended, or push pull.