The goalposts keep shifting in this discussion. We now appear to be mixing up CCS bias shift, P-P common mode feedback, possibly re-entrant distortion. We started with simple (but incorrect) statements from smoking-amp about OPT CT signals, but having admitted that they were incorrect we now seem to be piling complication on complication. My head is spinning so I give up. It is not an output stage architecture I am ever likely to use so I have decided I don't really care what other people believe, rightly or wrongly, is its method of operation.
No, you have doubled the fundamental frequency by repeating the waveform.wahab said:
I didn't pull your phrases out of context. I systematically disagreed with them all. It's unlikely that I misunderstood: your points are conventional and in widespread currency.
My view is not so conventional, and you may not have considered it before. If you dismiss it as misunderstanding, then you have missed something.
On the issue of synthesised sound I subsequently made my point more explicit. Perhaps I could clarify it a little more.
I contend that "distortion" as a function of a synthesiser or guitar amp is not distortion. In the context of a discussion about the presentation of recorded music, that meaning of "distortion" is a red herring.
All musical instruments produce an output that is "distorted" with respect to some notional source within the machine. They all incorporate sound-shaping functions. From the point of view of the listener, however, the sound is just how it is. There is no real distortion of the music arising from that notional "distortion" inside the machine.
Synthesisers and electric guitars are no different in that respect from any other instruments. They are, however, different in that the sound-shaping is done in software or firmware, electrically rather than mechanically. Consequently there are electrical signals within, and some of these may be "distorted" in relation to others, in other parts of the machine. This fact opens the door to a significant logical error.
Put simply, the sound of an electric guitar is not distorted, no matter how much the musician turns the knob marked "distortion". If I plug the same device into my stereo and turn the same knob, the sound is distorted. I guess you agree with that, more or less. With respect to the music, the first effect is not distortion, whereas the second is.
I realise my point seems pedantic. All the same it is necessary, because the common guitar amp argument, frequently dragged into discussions by reproductionists defending themselves against the champions of what they call "subjectivism", is a trojan horse.
If I accept that "distortion" in synthesisers or guitar amps has anything in particular to do with discussions about audio engineering, then I would also accept that electrical signals are in some way special: that it is electrical signals, rather than sound or music, that lie at the focal point of audio engineering.
As an exponent of authentic presentation, rather than reproduction, I cannot allow that confusion into my scheme of thought. Electrical signals can be reproduced, but not authentically presented. Music, on the other hand, can be authentically presented, but not reproduced. If I focus on signals, I will slide inevitably backwards into reproductionism.
From the point of view of authentic presentation, the commonly identified division between "subjectivists" and "objectivists" is a false dichotomy. With respect to the basic arguments, they are both quite right. In taking sides rather than seeking reconciliation, they are both quite wrong. From a third viewpoint, they are quite obviously building the same bridge from opposite banks. Between them, over history, they have in part defined the meaning of authentic presentation.
Within that history, especially when it was not possible to banish nearly all distortion, it was widely agreed that low order harmonics (rather than even order, as the OP postulated) in declining small proportions, were desirable. That is because the result sounds like a modification of timbre, rather than distortion, and the music generally retains its integrity unless it is very dense. This was desirable as an alternative to modifications to the sound that were intrusively not musical, in that they sound like distortion or noise.
Consequently, if you want to listen to Little Richard, a Dansette record player or an early American car radio is high fidelity. That's the historical truth. I remove the relative distortion due to my system by using a valve amp and Celestion speakers to replace the lost harmonics and noise. Not perfect though...it still sounds a bit middle-class-sat-on-the-sofa. Perhaps I need a scratch synthesiser.
Naturally I am frequently ambushed by exponents of both sides of the old reproductionist divide. Occasionally I get a little grumpy. I remind myself that Copernicus had the same problem.
That s no argument , if i can say so..
In the doubt i did look at an engineering orientated manual.
I was wrong for the exact formulation , confusing sinus and
cosinus in the formula form ,though...
y = cosx , 0 < x < pi
y = 4/pi ( 2sin2x/3 + 4sin4x/15 + 6sin6x/35 + .................
So the waveform look like a cosine restricted to its first half
and then repeated at the discontinuity point indefinitly ,
thus including a periodical jump from minimal to maximal value
of the function.
PlasticsGood; I don't participate in semantics' games: English is not my native language. Sorry. And I don't play "Objectivists/Subjectivists" game. I use objective measures to fool subjective perception, and have great results in this game. You may compare them with Celestions, American car radios, and what ever you want, and draw your own conclusions.
That s no argument , if i can say so..
In the doubt i did look at an engineering orientated manual.
I was wrong for the exact formulation , confusing sinus and
cosinus in the formula form ,though...
y = cosx , 0 < x < pi
y = 4/pi ( 2sin2x/3 + 4sin4x/15 + 6sin6x/35 + .................
So the waveform look like a cosine restricted to its first half
and then repeated at the discontinuity point indefinitly ,
thus including a periodical jump from minimal to maximal value
of the function.
What kind of transfer function would you use to get such waveform?
(Except some kind of synchronous modulator)
What kind of transfer function would you use to get such waveform?
(Except some kind of synchronous modulator)
Question was if a symetrical around zero and periodic function
could contain only a said frequency and its even multiples.
The polynome above is the serial develloppement of the said
function , nothing else.
"The goalposts keep shifting in this discussion. "
Its not that the goalposts are shifting, its a different question altogether. And an interesting one I think, if you will stop to consider it. Let me re-state the question.
For both the CCS tail case (class A P-P in all cases) and the optimum adjusted common mode neg. feedback case (CMFDBK), the summed tube currents should become constant. (the CMFDBK requires adjustment of the B+ current sensed feedback level to both grids to achieve this constant current, while the CCS case is automatic) My first point was that these two cases are very similar in that the common mode cathode to both grids voltage is being manipulated to maintain constant current, just that one is using the cathode end and the other is using the grids end.
Since the CMFDBK case is directly removing the even order current distortion to do this, my first surmise is that the CCS case is doing the same. Both cases thus removing the even order dist. effectively --in the tubes--, since no residual current variation occurs in the tail or B+ center tap. ( really by modifying the input signal)
Now lets look at the more common, grounded cathodes P-P class A case (GCPP). Here the OT removes the even order dist. component, and it flows off as varying current in the B+ center tap. The tubes are still generating the even order distortion in this case, but the OT removes it. This means that the tubes are operating over a different curve (versus output signal) here than in the CCS or CMFDBK cases.
So the question arises as to the relative level of the non-directly-removed odd harmonics between CMFDBK or CCS versus GCPP. For a tube with say a >1 power law, the GCPP case would have higher currents at the signal peaks than the current sum constrained CMFDBK/CCS cases. This means that the gm varies more in the GCPP case toward the peaks (versus the zero crossing), and this symmetrical type of gm variation should produce more odd harmonic distortion. (ie, no gm variation would lead to zero odd harmonic generation for example, but here we have more gm variation than the other versions)
So my second surmise is that the CMFDBK or CSS cases both have less odd harmonic distortion than plain GCPP. Yet the direct mechanism they use is by removing the even harmonics in the tubes, rather than later in the OT. It would seem that one could interpret this as some internal even harmonic dist. in the GCPP case gets converted to odd harmonic dist., but not directly by the OT. More of a cross modulation type effect thru gm modulation.
So one could then say that CMFDBK or CSS reduces -some- of the odd order distortion as well (over plain GCPP), but not completely. (a CCS tail still has the inherent odd order tube distortion present, the CMFDBK should be similar, just that the even harmonic removal mechanism is preventing some additional cross mod odd order dist.)
Hope I was clearer this time.
-----------------
Finally, a proposal to remove even more odd harmonic in the CMFDBK case. By adjusting the common mode feedback level higher to --over-- compensate, we may be able to make the current sum reduce some at the signal peaks rather than just stay constant. (this would be the opposite of GCPP where the sum current increases at peaks) This would have the effect of reducing the gm sum at signal peaks, making the tubes (with >1 power law) effectively more linear. So less odd order distortion maybe........... speculation at this point.
Its not that the goalposts are shifting, its a different question altogether. And an interesting one I think, if you will stop to consider it. Let me re-state the question.
For both the CCS tail case (class A P-P in all cases) and the optimum adjusted common mode neg. feedback case (CMFDBK), the summed tube currents should become constant. (the CMFDBK requires adjustment of the B+ current sensed feedback level to both grids to achieve this constant current, while the CCS case is automatic) My first point was that these two cases are very similar in that the common mode cathode to both grids voltage is being manipulated to maintain constant current, just that one is using the cathode end and the other is using the grids end.
Since the CMFDBK case is directly removing the even order current distortion to do this, my first surmise is that the CCS case is doing the same. Both cases thus removing the even order dist. effectively --in the tubes--, since no residual current variation occurs in the tail or B+ center tap. ( really by modifying the input signal)
Now lets look at the more common, grounded cathodes P-P class A case (GCPP). Here the OT removes the even order dist. component, and it flows off as varying current in the B+ center tap. The tubes are still generating the even order distortion in this case, but the OT removes it. This means that the tubes are operating over a different curve (versus output signal) here than in the CCS or CMFDBK cases.
So the question arises as to the relative level of the non-directly-removed odd harmonics between CMFDBK or CCS versus GCPP. For a tube with say a >1 power law, the GCPP case would have higher currents at the signal peaks than the current sum constrained CMFDBK/CCS cases. This means that the gm varies more in the GCPP case toward the peaks (versus the zero crossing), and this symmetrical type of gm variation should produce more odd harmonic distortion. (ie, no gm variation would lead to zero odd harmonic generation for example, but here we have more gm variation than the other versions)
So my second surmise is that the CMFDBK or CSS cases both have less odd harmonic distortion than plain GCPP. Yet the direct mechanism they use is by removing the even harmonics in the tubes, rather than later in the OT. It would seem that one could interpret this as some internal even harmonic dist. in the GCPP case gets converted to odd harmonic dist., but not directly by the OT. More of a cross modulation type effect thru gm modulation.
So one could then say that CMFDBK or CSS reduces -some- of the odd order distortion as well (over plain GCPP), but not completely. (a CCS tail still has the inherent odd order tube distortion present, the CMFDBK should be similar, just that the even harmonic removal mechanism is preventing some additional cross mod odd order dist.)
Hope I was clearer this time.
-----------------
Finally, a proposal to remove even more odd harmonic in the CMFDBK case. By adjusting the common mode feedback level higher to --over-- compensate, we may be able to make the current sum reduce some at the signal peaks rather than just stay constant. (this would be the opposite of GCPP where the sum current increases at peaks) This would have the effect of reducing the gm sum at signal peaks, making the tubes (with >1 power law) effectively more linear. So less odd order distortion maybe........... speculation at this point.
Last edited:
Reproductionist LOL yes, I'm a breeder; have children.
Sounds like you know an analog Aphex Aural Exciter from a ring modulator. The Aphex supposedly adds only (mostly??) even-order harmonics (to the treble above some frequency). Subjective descriptions say it sounds like additional "sizzle" much like treble EQ boost but without the typical corresponding increase in the treble noise floor. What I find curious is that its application moved out of the studio where it was originally used to create and blend rock and pop sounds (often vocals), to mall public paging systems where it "increases intelligibility". Supposedly there's a white paper somewhere about this...
I recenlty picked up a few Aphex Aural Exciters cheap but haven't had time to play with them yet. The Behringer Exciter supposedly achieves the same goal of adding even-order harminics but achieving it in the digital domain (a digital Aphex emulator if you will). Now I think Aphex has a digital algorithm they license to digital mixing board mfgrs too (an Aphex digital Aphex emulator).
I don't know English term, but "interruptible" function does not exist in natural sounds.
We speak about distortions of sinewave function, and waveform of the result.
Right; it may be good for breeders.
But it is more than application of specifically bent transfer curve.
Sound reproduction means restoration of the sound in listener's position in order to fool imagination as if the listeners is there. Or, as if the sound source is here and now.
You will need to introduce some phase shifts between inter-modulating signals to do so.
FT by hand..... Huh....
Hi,
Its quite easy for a square, triangular or sawtooth waves.
Draw a sine wave with amplitude 1. Draw a sine wave 1/3 amplitude,
3 x the frequency. Add them together. The centre of the the sine wave
dips, the skirts increase in slope. there are 3 ripple peaks. Plot 1/5
amplitude at 5 x frequency and add that, its easy to visually see by
hand with 1/7+7F,1/9+9F etc you will end up with a square wave.
If you subtract the harmonics rather than add them = a triangular wave.
If you add the series 1/2+2F, 1/4+4F etc you get a |\|\|\ sawtooth.
If you subtract the same series you will get a /|/|/| sawtooth.
Doing this by hand should help with Fourier understanding, and for
fairly simple repeating waveforms help you predict the level and
phase of each harmonic roughly, if not precisely.
Just to throw a spanner in the works though - any arbitrary repeating
waveform can be represented by any set of mutually orthogonal
(multiply them and the answer is always zero) harmonic functions,
i.e. a square wave and its harmonics is just as valid as sine waves.
In digital electronics you can generate sine waves from square waves
simply by doing something very the similar to the above, try it on paper.
rgds, sreten.
Guys,
Above there was some reference to adding common mode feedback in a Push Pull Output Stage to reduce odd harmonic distortion.
I would lie to draw your attention to something Sheldon posted in my Baby Huey thread.
Refer post #451 and #452 of that thread.
http://www.diyaudio.com/forums/tubes-valves/72536-el84-amp-baby-huey-46.html
His two plots do show a reduction in odd harmonic distortion but also importantly show a massive reduction in intermodulation sidebands which suggests to me elimination of higher harmonics - why? Low even order harmonics are accompanied by few IM Sidebands which are remote from the original frequencies while odd harmonics are accompanied by IM Sidebands half of which are very close to the original frequencies. Also higher order products have many more IM Sidebands.
His 2 plots suggest to me that the common mode feedback has reduced odd order products particularly higher order ones. Do you agree with this or do you think my conclusions are invalid, that is I saw what I wanted to see?
Cheers,
Ian
Above there was some reference to adding common mode feedback in a Push Pull Output Stage to reduce odd harmonic distortion.
I would lie to draw your attention to something Sheldon posted in my Baby Huey thread.
Refer post #451 and #452 of that thread.
http://www.diyaudio.com/forums/tubes-valves/72536-el84-amp-baby-huey-46.html
His two plots do show a reduction in odd harmonic distortion but also importantly show a massive reduction in intermodulation sidebands which suggests to me elimination of higher harmonics - why? Low even order harmonics are accompanied by few IM Sidebands which are remote from the original frequencies while odd harmonics are accompanied by IM Sidebands half of which are very close to the original frequencies. Also higher order products have many more IM Sidebands.
His 2 plots suggest to me that the common mode feedback has reduced odd order products particularly higher order ones. Do you agree with this or do you think my conclusions are invalid, that is I saw what I wanted to see?
Cheers,
Ian
Last edited:
"Do you agree with this or do you think my conclusions are invalid, that is I saw what I wanted to see?"
Sheldon's post does show some reduction of 3rd harmonic, but also a big boost in the 4th harmonic. The only IMD sidebands I see going away are the power supply ones around the fundamental.
---------------------------------
"What a muddle has been done! "
I suspect some new reading glasses are required in these parts. Disconnections everywhere.
---------------------------------
"Doing this by hand should help with Fourier understanding"
FFT applet:
http://www.falstad.com/fourier/
Sheldon's post does show some reduction of 3rd harmonic, but also a big boost in the 4th harmonic. The only IMD sidebands I see going away are the power supply ones around the fundamental.
---------------------------------
"What a muddle has been done! "
I suspect some new reading glasses are required in these parts. Disconnections everywhere.
---------------------------------
"Doing this by hand should help with Fourier understanding"
FFT applet:
http://www.falstad.com/fourier/
Last edited:
Discontinuous function. Discontinuity.Wavebourn said:
When I said Fourier by hand I actually meant doing the trigonometric integrals with pencil and paper, but the graphical synthesis method works well for those whose calculus is a little rusty.
sreten has raised the issue of alternative orthogonal basis sets - I suspect that will only cause further confusion, unless DIYaudio wants a tutorial on inner product spaces etc. Lets stick with sine waves; they create enough perplexity already.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.