"What's your reasoning?" and not "What's your belief?".

[andy_c]

Hi Mike,

I'm tied up right now, but I'll try to respond after work (about 7 hours from now).

 

[jan.didden]

Yes, I agree with youir paragraph 1.

Well, not quite I think, for para 2. You introduce the VAS terminology, which doesn't relate to the AD844, a current feedback type. The internal Z is referenced to GND (or the supplies), and is high. Part of it is the high Z of the current sources feeding it, part of it the Cc in parallel. The external R shunts this Z, which effectively lowers the dynamic Z (assuming Cc is constant). The total OL gain is lowered, and the OL break frequency is pushed up. Gain has been traded for BW. The phase now is relatively constant with freq., since the OL gain is flat in the audio band.

As for your para 3, I don't think I quite understand it. Remember that you can't change the Gain*BW with this trick, so if you lower Gain (by the shunt R), the OL BW must go up, to maintain the same G*BW.

Walt Jung

Walt,

Sorry, I used the term Vas stage too loose, meaning the volt amplification stage. As you say, it is really the current output stage, which becomes a Vas with the addition of the load R.

But the other thing, I wrestle with that because it seems so counter-intuitive. You say, you can't change GBW product by adding a resistor at the current source output. But the resistor DOES lower the gain below the fc, no?
So at the lower freq range you have a lower gain with essentially the same BW, so your GBW IS lower, I would think. As I understand this, the GBW product is constant below fc. Since by this method you don't chnage the freq response above fc, but you slice off the part below fc, it lowers the gain part of the GBW product so it MUST lower the GBW product. Or am I interpreting it wrong?

Jan Didden

 

[andy_c]

Hi Jan,

Just a real quick note.

For VFB and CFB

Gain(DC) = K * R
BW = 1/(RC) rad/sec

GBW = k/C independent of R

For VFB, K = gm
For CFB, K = current gain of current conveyor (usually 1), so gain = transimpedance

Try with SPICE looking at VCCS driving parallel RC. Step R. Look at family of gain vs freq curves. Compare with jcx's previous post (gain vs freq with two different R values).

Sorry to be so terse.

 

[andy_c]

But now another thing ! As i stated in a previous post, PM seems to behave like even harmonics

That's what I originally thought too. Then I went through Gilbert's analysis where he shows mathematically the cause of a variable phase shift from input to output that depends on signal level (and gets worse at higher frequencies). This turns out to be associated with the third harmonic. So now I'm confused. I'm thinking of two possibilities.

1) My original thinking on PIM may have been completely flawed. This is certainly possible, since I had only seen third-party accounts of what Otala was saying, and had been only using heuristic arguments to try to figure out which harmonics were involved.

2) What Otala describes as PIM may be somehow different from what Gilbert describes.

Both could be true I guess. Later this week I'll be getting some AES articles including all of Otala's articles and will read through those to try to figure out what's going on. Feel free to email me for details of these.

...means it should go with a fully complementary design ?

I don't think this will help or hurt either way according to Gilbert's analysis. Both the complementary diff amp and the single diff amp have similar tanh()-type nonlinearities, and that's one major source of the signal-level-dependent phase shift that Gilbert describes.

I observed this PM only in asymetrical circuits, i called it "asymetrical slewrate"

I've seen this in simulation with asymmetrical circuits as well. In the case I saw, it was due to the constant-current load of the VAS, the two-transistor feedback current source that Self uses. The fast-changing collector voltage causes its base voltage to "bounce", which then reduces its current temporarily. Using a cascode current source seems to fix this. Since its base is AC ground, its current doesn't change much even with fast transients of collector voltage.

I guess this is a similar idea to PIM, but Gilbert's analysis assumes the time rate of change of the output is much less than the slew rate, such that the nonlinearity of the input stage is small. For example, he approximates atanh(x) as (x + x³ / 3), where x is the change in the output current of the input stage divided by its DC value. This only works when x is small, so the circuit must be far away from slew rate limiting.

 

[john curl]

AndyC, I think that you are on the right track. I would not attempt to equate the phase modulation with asymmetry, as it happens with symmetrical circuits as well, if the mechanism is simply non-linearity modulating the bandwidth with voltage level. If you have not read it yet, also look at Barrie Gilbert's article: 'The Multi-tanh Principle: A Tutorial Overview' IEEE Journal of SSC, Vol 33, Jan 1998. Walt may have it, but I also found it on CD rom, if you have trouble locating it.

 

[mikeks]

Re: more high gain reasoning

..............the high gain approach will always have less differential input signal amplitude at low frequencies which hugely reduces the distortion amplitude in the bjt diff pair..............

I agree.....

Indeed, this is readily apparent with simple THD+N measurement.......

The moral of the story really is that there is no 'new' audio amp. impairment out there that does not also manifest itself as an increase in THD.....

viz: In my view, if a design's THD is less than 1000ppm across the audio band, for any conceivable load driven to rated voltage swing, then there can be no doubt that the design is, for practical purposes, free of ALL known error.

 

[ALW]

In my view, if a design's THD is less than 1000ppm across the audio band, for any conceivable load driven to rated voltage swing, then there can be no doubt that the design is, for practical purposes, free of ALL known error.

But the audibility of that error to the human ear is inextricably linked to it's spectral makeup, so I'd take issue, in principle, that an arbitary error of any level is rendered inaudble by virtue of it being measurably low in a THD test.

Now you didn't talk about specifically about audibility and I'm not saying your wrong, since I haven't actually performed any test to correlate constant THD (at the ppm level you quote) with varied spectral content, but my gut feel, based on actual experience, is I'd be surprised if you're right, without further qualification.

Andy.

 

[ingrast]

System wide distotion sources

Can someone point me to a readily available (public) up to date discussion of system wide distortion contributions? I mean, at least from preamplifier input to actual sound pressure?

I am afraid there could be a hot debate regarding the importance of 1000 ppm THD in some link, while happily oblivious of a 2% contribution from the driver's mechanical nonlinearities just to name something.

Or, may be this contribution is less offensive to the ear? I cannot tell by personal experience but surely there are here qualified individuals to cast an opinion.

Anyway I at least should be much better served with results backed up by actual quantifiable magnitudes (your reasoning) than perceptual (your belief).

Rodolfo

BTW, Walt's 4 part article series http://home.comcast.net/~walt-jung/wsb/html/view.cgi-resources.html-.html is an authoritative, enlightening discussion on op amp audio performance.

 

[mikeks]

But the audibility of that error to the human ear is inextricably linked to it's spectral makeup....Andy.

The spectral constituents at the given level are completely inaudible.....

This is not speculation, but cold, hard, unyielding fact...

There are good reasons why THD measurements are accused of bearing 'no correlation with what is heard'....

Spectral make-up at this level cannot rationally be one of them...

One obvious example:

Douglas Self presents what he considers to be an excellent amplifier with the 'load-invariant' design...

Indeed, the worthy presents THD+N AP graphs as proof of its excellent performance...

But it takes only elementary arithmetic to demonstrate that these graphs could not have been produced with his choice of SOA protection in-situ....(with 40V rails)

This is because his protection would be invoked even in the absence of a nominal load....

So....if this design sounds nasty, one can hardly blame Self's THD specs. for the problem...

It merely means that low THD was not gauranteed for all specified supply rails and nominal loads.

 

[traderbam]

dither

It might be worth considering dither noise in CD when considering the lowest audible noise floor. The absence of dither noise is audible. I don't know the details of dither, perhaps someone else has this knowledge to hand. It is at least 1 bit or 30ppm.

 

[Swedish Chef]

Me too would like an article that discusses cascaded non-linear stages which really is what this is all about.

In the meantime, could someone please explain why PIM, TIM or any other non-linear behaviour would not manifest itself in a simple THD/IMD measurement? I can't figure it out.

/Magnus

 

[mikeks]

............In the meantime, could someone please explain why PIM, TIM or any other non-linear behaviour would not manifest itself in a simple THD/IMD measurement? ............

/Magnus

Indeed!!!

 

[mikeks]

Re: System wide distotion sources

I am afraid there could be a hot debate regarding the importance of 1000 ppm THD in some link, while happily oblivious of a 2% contribution from the driver's mechanical nonlinearities just to name something.

Loudspeaker systems in general comfortably contribute greater than 2% THD....which is why it is absurd to suggest that an amp. with say less than 1000ppm across the audio band into any given 'speaker could have 'audible defects'...

 

[traderbam]

Rodolfo wrote: I am afraid there could be a hot debate regarding the importance of 1000 ppm THD in some link, while happily oblivious of a 2% contribution from the driver's mechanical nonlinearities just to name something.

Now that's a heap of reasoning in a single sentence. ALW is also correct that the nature of the distortion is paramount to the listener's sensitivity.
Just imagine the plethora of noises that our ears pick up during a normal day, simultaneously, mixed together. To make sense of it we must process it and ignore some of it and select that which is most important. To "hear" everything at once with equal weighting would surely be major overload for our ape brains!

 

[john curl]

Folks, I have to say it once again: Many of you have NEVER studied the subjective sensitivity of the human ear to distortion artifacts. It was shown about 65 years ago, by German scientists, that you must WEIGHT the harmonic distortion products, IF you want accurate evaluation of the MAGNITUDE of harmonic that is generated.
This was clearly written in the 'Radiotron Designers Handbook' of 1941.
There have been MANY weighting factors: The latest being: N(2)/4 for each harmonic. You can see that this will give: 2nd=1, 3rd=9/4, 5th=25/4, 7th=49/4 (or than 12 times the 2nd harmonic), 9th=81/4 etc.
I look primarily at 7th harmonic. Why? Because tubes have a difficult time generating it, as well as loudspeakers, BUT solid state can make it easily.
Measure for yourself, if you don't believe me.
Are there other distortions, other than harmonic that elude us? YES!
One is Hirata distortion, shown by Dr. Hirata about 25 years ago, and FM distortion, which like Doppler distortion in loudspeakers, hides in a harmonic measurement. Serious designers must address EVERY type of distortion, in order to make a successful audio product. I certainly do.

 

[destroyer X]

Finally, there you are, some ligth to us

This make sense, just follow the track.

regards,

Carlos

 

[andy_c]

Me too would like an article that discusses cascaded non-linear stages which really is what this is all about.

Hi Magnus,
I've got a copy of the Gilbert article that John sent me. If you'd like a copy, just send me an email.

In the meantime, could someone please explain why PIM, TIM or any other non-linear behaviour would not manifest itself in a simple THD/IMD measurement? I can't figure it out.

I don't think anyone has made that claim in this thread. In fact, going through the math of the Gilbert article, it becomes clear that the level-dependent phase shift can't occur without also having a third harmonic. Does it matter? Is it any worse than distortion that doesn't change the input/output phase relationship as a function of level? I don't know. But it is interesting to think about.

 

[ingrast]

... Serious designers must address EVERY type of distortion, in order to make a successful audio product. I certainly do.

Gentlemen,

In a post I placed some days back and went unresponded, I proposed / asked for a sweeping test designed to uncover and charaterize each and every type of distortion, and to try to identify peculiar signatures as much as to try to correlate and make sense of perceived good and bad sounding with actual perturbation data not restricted specifically to harmonic, PIM, Hirata etc.

An input - output difference analysis will unmask each and every departure from "wire with gain", both known and perhaps (unlikely) yet undiscovered modes.

Should love to hear from results if this has been done, for the idea is certainly unremarkable in originality I guess.

Rodolfo

 

[Mike Gergen]

I'd like to ask, "What does it all mean from a theoretical and practical standpoint?"

I guess what I'm wondering is;
A complementary diff amp as a front end is good at reducing (or not creating) distortion X but not so good for distortion Y?

Then the same for Fet diff amp vs bipolar, single diff amp, any variuos second stage sections or output sections that are used, etc.

I know this is kinda vague and this question could turn into a complete thesis paper... I also understand that there are a whole lotta variables in all these configurations; parts used, bias levels, compensation, and on and on.

I think it would give some insight into why a certain topology is used by different people. The design goals and priorities of design A may be different for design B for a different sound.

Anybody willing to bite?

Thanks
Mike

 

[mikeks]

Folks, I have to say it once again: Many of you have NEVER studied the subjective sensitivity of the human ear to distortion artifacts.

Rather presumptious and patronizing don't you think?

It was shown about 65 years ago, by German scientists, that you must WEIGHT the harmonic distortion products, IF you want accurate evaluation of the MAGNITUDE of harmonic that is generated....................

No one disputes this....

Serious designers must address EVERY type of distortion, in order to make a successful audio product. I certainly do.

Not really.....merely taking care that THD+N is low across the audio band for all conceivable loads is sufficient.

John, do you have any evidence that this is not true?

http://www.diyaudio.com/forums/showthread.php?postid=491144#post491144

http://www.diyaudio.com/forums/showthread.php?postid=491152#post491152

http://www.diyaudio.com/forums/showthread.php?postid=491090#post491090