Bob Cordell Interview: Error Correction

The point was that it doesn't, as long as it has reasonable frequency and phase response--I brought up the issue initially in criticizing that advertisement. My next post was in response to your confusion about points in my initial one.

I find it amusing how concerned people are about ppm distortion in playback electronics when the sound was recorded with equipment that has higher distortion, when, worse, the sound drivers they use have far, far higher distortion, and not to mention the even bigger problems with recording and playback geometry. But, of course, people tend to prefer to work on the easier problems, even if they make the smallest difference in quality in the overall recording-playback system.
 
syn08 said:


Thanks, now that you put this as an encoding problem, it starts making sense.

But then what has a power amp (and in general negative feedback) to do with the process of encoding the audio information? Would it be possible to define a set of objective criteria to optimize the audio equipment in reproducing the encoded audio information?

Obviously, for a lot of people, it is not sub 1ppm distortion.

In fact, long story short, adding the right distortion can increase the
'image depth' and 'width'. Tubes and analog tape have been
doing it for years in recording studios.

Even our beloved ultra minimalist, super fidelity Chesky discs were
recorded with, for the most part, an all tube front end before ADC.

Considering that 99.9% of people listening to playback have
no idea what the original recorded image sounded like, then they
have no idea which playback system 'is correct' as opposed to
'sounds correct' or 'sounds best'.

As Edmond has so truely stated, musical instrument an amp may be
- but - if one amp sounds like there are a bunch of musicians playing
in your lounge room and the other doesn't, it's a no brainer which
amp most non technical people are going to like. It's as simple as
that. Listening to music, for most people, is purely a right brain
activity. They put a disc on, turn it up and go "wow, that's
great"... 🙂

Note:

Apologies for slight thread derailment.

I have been enjoying this thread immensely and have been
simulating my own amps and OP stages in LTspice, comparing a lot
of what has been discussed here.

I actually got one OP stage down to around 0.002% / 20kHz /
800W / 1 ohm load. :hot:

Yes, this would definitely be used for PA applications and I'm sure
real life measurements would be somewhat worse.

cheers

Terry
 
Anny "euphonic" distortion can be added by DSP or an analog filter. Relying on the amplifier's innate distortion for sound effects gives up flexibility. With processing, if one's tastes change when they hear another equipment, they can simply adjust the processing to match their changed taste. If the sonic signature is fixed by the amplifier, however, one would have to get a new one. Better for the sellers, but not the consumer.
 
Anny "euphonic" distortion can be added by DSP or an analog filter

been there

Triode Emulator The design for a low-noise amplifier is presented. The amplifier has a triode-like transfer characteristic and produces harmonic distortion components that are similar to triode preamplifier. It can be used as a building block for microphone preamplifiers, active front-end electronics for various pick-ups and aural stimulator units. Paper Number: 6008 Convention: 116 (May 2004)
 
nonlinear VAS loading

Bob Cordell said:
Hi Edmond,

This is very interesting. Could you post the schematics of the blameless amp that you simulted, both the straight DEF version and the one wherein you added the EC?

Thanks!
Bob

Hi Bob,

Interesting? I don't think so. It is rather obvious that the main source of distortion does NOT stem from a nonlinear loading of the VAS, as Hugh was wondering.

You need a schematic of the blameless amp? Come on, you can get it everywhere. All you need to know is that I spiced it with quite linear drivers (KSC2690/KSA1220) and output devices (KSC5200/KSA1943). (D. Self too is using highly linear devices in his latest versions)

As for the EC circuit, as I said before, it is an ideal EC thingy, nothing special. It consists of a voltage source that senses the difference between input and output of the OPS and it is rolled off by a simple RC network so that the Ft of the EC-NFB loop is set to 3MHz.

Since the distortion of the blameless amp mainly consists of the 2nd and 3rd harmonic, it's quite plausible that the THD20 is reduced 27 times, because the EC-NFB loop gain at 40kHz is 75x and at 60kHz it is 50x. Why the THD is 'only' reduced by a 'disapointing' factor of 27 might be explained by the distortion of the front-end, which is (of course) not reduced by the EC circuit.

Cheers, Edmond.
 
Re: nonlinear VAS loading

Edmond Stuart said:


Hi Bob,

Interesting? I don't think so. It is rather obvious that the main source of distortion does NOT stem from a nonlinear loading of the VAS, as Hugh was wondering.

You need a schematic of the blameless amp? Come on, you can get it everywhere. All you need to know is that I spiced it with quite linear drivers (KSC2690/KSA1220) and output devices (KSC5200/KSA1943). (D. Self too is using highly linear devices in his latest versions)

As for the EC circuit, as I said before, it is an ideal EC thingy, nothing special. It consists of a voltage source that senses the difference between input and output of the OPS and it is rolled off by a simple RC network so that the Ft of the EC-NFB loop is set to 3MHz.

Since the distortion of the blameless amp mainly consists of the 2nd and 3rd harmonic, it's quite plausible that the THD20 is reduced 27 times, because the EC-NFB loop gain at 40kHz is 75x and at 60kHz it is 50x. Why the THD is 'only' reduced by a 'disapointing' factor of 27 might be explained by the distortion of the front-end, which is (of course) not reduced by the EC circuit.

Cheers, Edmond.


I wanted to SPICE what you did myself to make sure you were not overlooking something and to understand better what is going on. Just send me the two asc files you actually used. No need to speculate and argue. I'll probably end up confirming your results.

Cheers,
Bob
 
Re: Re: nonlinear VAS loading

Bob Cordell said:
I wanted to SPICE what you did myself to make sure you were not overlooking something and to understand better what is going on. Just send me the two asc files you actually used. No need to speculate and argue. I'll probably end up confirming your results.

Cheers,
Bob

Hi Bob,

As you know I don't use LTSpice, so I can't send you the asc files.
Indeed, no need to speculate and argue. I like to go even one step further: always run a sim before making any comment.

But I really don't understand why you need these files or a schematics.
Actually, it is a bad idea, because running the same application on the same files yields the same result, which proves of course nothing.

As far as I know, you have already captured a version of your amp without EC, so it is quite easy to use that one with and without ideal buffer between the VAS and OPS.

Most likely, the results (86 vs 87ppm) will not exactly be the same, but that doesn't matter. The point is (ie Hugh's point) in how far unloading the VAS from the nonlinear input impedance of the OPS contributes to the distortion reduction by means of HEC.

As I have shown, this contribution is minuscule. Even in the event that I have overlooked something and it is ten times higher, it's still far less than the overall THD reduction.

Cheers, Edmond.
 
Re: Re: Re: nonlinear VAS loading

Edmond Stuart said:


Hi Bob,

As you know I don't use LTSpice, so I can't send you the asc files.
Indeed, no need to speculate and argue. I like to go even one step further: always run a sim before making any comment.

But I really don't understand why you need these files or a schematics.
Actually, it is a bad idea, because running the same application on the same files yields the same result, which proves of course nothing.

As far as I know, you have already captured a version of your amp without EC, so it is quite easy to use that one with and without ideal buffer between the VAS and OPS.

Most likely, the results (86 vs 87ppm) will not exactly be the same, but that doesn't matter. The point is (ie Hugh's point) in how far unloading the VAS from the nonlinear input impedance of the OPS contributes to the distortion reduction by means of HEC.

As I have shown, this contribution is minuscule. Even in the event that I have overlooked something and it is ten times higher, it's still far less than the overall THD reduction.

Cheers, Edmond.


Edmond,

Sorry, I forgot that you don't use LTSPICE.

Why don't you just post the two schematics?

Else, OK, I give up, it's not worth the trouble.

Bob
 
Re: Re: Re: Re: nonlinear VAS loading

Bob Cordell said:



Edmond,

Sorry, I forgot that you don't use LTSPICE.

Why don't you just post the two schematics?

Else, OK, I give up, it's not worth the trouble.

Bob

Bob,

I did a comparison of self blameless amp with double and triple EF
OP stages on Ltspice.

Edmond is right - to a point - the OP stage loading has less effect
than you would think. However it does have an effect, especially
when you want an all BJT OP stage to deliver high power.

My findings were at higher powers, buffered OP stage can
have around 1/2 distortion - but still I would have expected
more than this.

If you would like the asc files I can send them to you to have a play
with.

cheers

Terry
 
Re: Re: Re: Re: Re: nonlinear VAS loading

Terry Demol said:
Bob,

I did a comparison of self blameless amp with double and triple EF
OP stages on Ltspice.

Edmond is right - to a point - the OP stage loading has less effect
than you would think. However it does have an effect, especially
when you want an all BJT OP stage to deliver high power.

My findings were at higher powers, buffered OP stage can
have around 1/2 distortion - but still I would have expected
more than this.

If you would like the asc files I can send them to you to have a play
with.

cheers

Terry

Hi Terry,

>a buffered OP stage can have around 1/2 distortion.

That's considerable more than I have observed. Probably due to less linear trannies, a slightly different topology and pushing the amp to near its maximum output level, while I've spiced it at only half of its max. power.

The effect seems also frequency dependent as Douglas Self has figured out (see picture below). Also notice the marginal improvement above 10kHz, far less than 50%

Anyhow, the conclusion remains unaltered: the distortion reduction by means of EC has very little do with loading effects of the VAS.

Thanks for your contribution.

Cheers, Edmond.
 

Attachments

  • vas.png
    vas.png
    51 KB · Views: 520
Re: Re: Re: Re: nonlinear VAS loading

Bob Cordell said:
Edmond,

Sorry, I forgot that you don't use LTSPICE.

Why don't you just post the two schematics?

Else, OK, I give up, it's not worth the trouble.

Bob

Why I don't just post the two schematics? Because:

1. I can't imagine that you don't already have a copy of the blameless amp schematic.

2. My schematics are too large and it isn't worth the trouble to redraw them.

3. You can use any amp with a Miller compensated VAS and BJT OPS for your simulations.

4. A man of your stature is supposed to have enough skills to insert an ideal buffer between the VAS and OPS in the correct way. You certainly don't need me (nor my schematics) to figure that out.

Cheers, Edmond.
 
Re: Re: Re: Re: Re: Re: nonlinear VAS loading

Edmond Stuart said:


Hi Terry,

>a buffered OP stage can have around 1/2 distortion.

That's considerable more than I have observed. Probably due to less linear trannies, a slightly different topology and pushing the amp to near its maximum output level, while I've spiced it at only half of its max. power.

The effect seems also frequency dependent as Douglas Self has figured out (see picture below). Also notice the marginal improvement above 10kHz, far less than 50%

Anyhow, the conclusion remains unaltered: the distortion reduction by means of EC has very little do with loading effects of the VAS.

Thanks for your contribution.

Cheers, Edmond.

Yes, it was 30V peak into 4 ohms at 20kHz. I wanted to give areal
stress test.

Thinking about it now, I probably should have done 1kHz as the
miller c would be loading down the VAS somewhat at 20kHz.

I also varied the IP degen from std 100ohms down to 5
ohms, comparative results same, although with less degen
distortion is reduced from higher OLG.

WRT trannies, no I used very linear ones, 3601/5171/2922 etc.

This whole scenario is very complex. Another limiting factor may be
the actual finite impedance of the VAS itself.

So I also tried a cascoded vas, and again the improvements were
there but only minor.

I will look into this further but just don't have a heap of time
presently.

cheers

Terry
 
Re: Re: Re: Re: Re: nonlinear VAS loading

Edmond Stuart said:


Why I don't just post the two schematics? Because:

1. I can't imagine that you don't already have a copy of the blameless amp schematic.

2. My schematics are too large and it isn't worth the trouble to redraw them.

3. You can use any amp with a Miller compensated VAS and BJT OPS for your simulations.

4. A man of your stature is supposed to have enough skills to insert an ideal buffer between the VAS and OPS in the correct way. You certainly don't need me (nor my schematics) to figure that out.

Cheers, Edmond.


I don't think I lack the necessary skills, Edmond, but I do lack the surety of duplicating properly your results. I'm wondering if all your resistance here has to do with reluctance on your part to have your results truly checked and analyzed. I thought your results were interesting and non-intuitive, and wanted to pursue them further in a true apples-apples way. I am not going to waste a bunch of time trying it on some other topology with possible other assumptions only to find out we are not doing apples-to-apples. I suspect that there is an "ah but..." in what you did, but I cannot pin it down.

I give up. Its not worth arguing about.

Bob
 
Re: Re: Re: Re: Re: nonlinear VAS loading

Terry Demol said:


Bob,

I did a comparison of self blameless amp with double and triple EF
OP stages on Ltspice.

Edmond is right - to a point - the OP stage loading has less effect
than you would think. However it does have an effect, especially
when you want an all BJT OP stage to deliver high power.

My findings were at higher powers, buffered OP stage can
have around 1/2 distortion - but still I would have expected
more than this.

If you would like the asc files I can send them to you to have a play
with.

cheers

Terry


Hi Terry,

Thanks for doing this. An interesting result.

Cheers,
Bob
 
Thank you Bob, Edmond, Terry, for interesting results.

In light of the fact that EC is very complex and difficult to precisely set up, and that comparing apples and oranges is a real risk, it's emerging that the only definitive way to test the viability of EC is to build the circuit, measure it carefully vis a vis the PSpice/LTSpice results, then start the listening tests. Argument here proves nothing, and serves only to bruise egos.

When experts roundly condemn the Dartzeel NHB108 for its adverse distortion performance (presumably having never heard it) and yet this is an amp which won a Stereophile award, then it is clear to me that the gulf between PSpice, measured results and listening tests is wider than ever and all the technical progress of decades past remains inconclusive. Marketing, perhaps, has done this, but I suspect with the complicity of experts who still cannot correlate the objective with the subjective.

This is not to condemn anyone here, but it does draw attention to the veracity of strongly held expert belief. When the highly sensitive, attuned listener has the final word, and this is the basis for many high end purchases, then I ask whether an expert opinion of the product schematic has any relevance at all.

I was interested to see Terry's comment that while reduction in local degeneration worsened stage performance, global feedback improved it since OLG increased. This indicates a very fine balance, and leaves the way open to some very careful tweaking 'for best sound'.

Cheers,

Hugh
 
Blind listening tests should be the ultimate arbitration, not measurement and not non-blind listening (measurements can only be a substitute once a metric has been found that fully correlates with perception, but despite the work of Geddes and others this seems not very close at this point).
 
Re: Re: Re: Re: Re: Re: nonlinear VAS loading

Bob Cordell said:
I don't think I lack the necessary skills, Edmond, but I do lack the surety of duplicating properly your results. I'm wondering if all your resistance here has to do with reluctance on your part to have your results truly checked and analyzed. I thought your results were interesting and non-intuitive, and wanted to pursue them further in a true apples-apples way. I am not going to waste a bunch of time trying it on some other topology with possible other assumptions only to find out we are not doing apples-to-apples. I suspect that there is an "ah but..." in what you did, but I cannot pin it down.

I give up. Its not worth arguing about.

Bob

Hi Bob,

Give it up. Indeed, the best you can do. There's nothing you can add to this topic (except confusion), as my sims just confirm the measurements of Douglas Self.

In the mean time I have also simmed the blameless amp under highly stressed conditions (4Ohm/20kHz/1V below clipping) and the results were pretty close to Terry's findings: a buffered OP stage can have around 1/2 distortion.

Next, I simmed the amp at 20Hz, and, not surprisingly, the results were again in close agreement with Self's measurements.

If you still suspect that there is an "ah but...", than I'm really sorry for you. There ain't no "ah but...", there's nothing to pin down. This is one more reason why I'm so hesitant to give the schematics to you. Just see it as an attempt to prevent that this debate degrades into a real soap.

Cheers, Edmond.

NB: Please, don't forget to look at this graph:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1403999&stamp=1200657887
 
Re: Re: Re: Re: Re: Re: Re: nonlinear VAS loading

Edmond Stuart said:


Hi Bob,

Give it up. Indeed, the best you can do. There's nothing you can add to this topic (except confusion), as my sims just confirm the measurements of Douglas Self.

In the mean time I have also simmed the blameless amp under highly stressed conditions (4Ohm/20kHz/1V below clipping) and the results were pretty close to Terry's findings: a buffered OP stage can have around 1/2 distortion.

Next, I simmed the amp at 20Hz, and, not surprisingly, the results were again in close agreement with Self's measurements.

If you still suspect that there is an "ah but...", than I'm really sorry for you. There ain't no "ah but...", there's nothing to pin down. This is one more reason why I'm so hesitant to give the schematics to you. Just see it as an attempt to prevent that this debate degrades into a real soap.

Cheers, Edmond.

NB: Please, don't forget to look at this graph:
http://www.diyaudio.com/forums/attachment.php?s=&postid=1403999&stamp=1200657887

Edmond,

You have contributed some interesting ideas to this thread. That makes it extra regrettable that you decided on this course of action.

I'm sure you are well aware that the greatest advances always come from interactions and mutual checking of results, conditions and often subtle differences. It is often the subtle differences that provide the insight. Indeed, the best scientists and researchers go through great lenghts to painstakingly describe how and what they do, including full disclosure of methods, circuits etc. This type of peer review greatly increases confidence in results.

Without this intense interaction and review, any reported results remain at the level of anecdotes. Your work deserves better.


Jan Didden