In article , Trevor Wilson
wrote:
"Jim Lesurf" wrote in message
...
In article , Trevor Wilson
wrote:
"Jim Lesurf" wrote in message
...
In article , Trevor Wilson
So the person throwing the switch did know which choice (A or B)
they had made at each time?
**They knew whether it was A or B, but not which amp was which.
That means the result isn't double-blind. The point of
double blind is that *noone* involved in either running or taking
the test has *any* idea which - A or B - is being presented at the
time.
**True enough. I never claimed that the test was perfect.
But the above means that the test can't be established as blind.
**Like I said: It was not a perfect test.
That does seem rather an understatement. Given that your answers indicate
that the test may well have been worthless, certainly so as evidence for a
claim that - as a class of devices - MOSFETs and BJTs 'sound different'.
So the
results could easily be for reasons other than the one you asserted.
i.e. nothing to do with 'MOSFET' being any different to 'Bipolar' as
classes of o/p device. Thus your evidence isn't reliable as a basis
for your belief.
**It is not a belief.
I'm afraid that simply re-asserting a belief does not make that belief
correct as a description of reality.
The test was constructed in an attempt to verify
what I and others had already informally noted as having heard.
But from your own answers since, clearly cannot have done that if what you
wished to verify was if MOSFETs and BJTs sounded different as a class of
devices. So by your own comments since, the test results were worthless for
that purpose.
, how were they recruited, how much did they know about the
purpose of the test? etc.
**They only knew that there might be a change in the system when
the light changed from green to red.
Which "light" controlled how?
**The light in the listening room.
Controlled *how*?
**By the person throwing the switch. He also controlled the lights. The
lights merely indicated a POSSIBLE change in amplifier.
So they also operated this light as often when there was no change as when
there was? To have not done so would be a methodological flaw unless you
dealt with established 'difference' bias in some other way. cf the
Stereophile test that fell foul of this problem and thus produced a false
claim about amps being shown to sound different.
Raw data of results?
**Lost in the mists of time. The results were 100% anyway.
That has no assessable meaning without knowing how many times the test
was done, if the same material was used, conditions of tests, etc, etc.
**I understand and have acknowledged the imperfect nature of the test.
But not that the results are actually of no worth in supporting the view
you initially expressed re MOSFET versus BJT.
Statistical analysis and outcomes in terms of levels
of confidence, etc?
**The participants picked which amp was playing 100% of the time.
Problems as above.
I could easily say "I spent 2 mins listening to one amp, then 1 min
listening to another. I could hear they were different. 100% right."
Alas, as evidence that would be worthless.
**Indeed. Except that the listeners were able to correctly identify each
amplifier perfectly.
Which is meanignless in this case for the purposes of your initial claim.
How did you establish that *current* limiting never occurred?
**I didn't. In fact, when using the Quad ESL63 speakers (I also used KEF
104.2 speakers), I suspect that both amps would have experienced
current limiting at some point.
In that case any difference could quite easily come from differences in the
current limiting behaviours of the two specific designs using the specific
devices. Nothing to do with MOSFET versus BJT as a class of device.
My experience
with o/p fets from days of yore is that they were far more prone to
this than bipolars.
**Indeed. However, the current limiting of MOSFETs has always tended to
be more benign than BJTs (depending on how the BJT current limiting is
employed).
That depends on how you define 'benign'. I define it as being 'never
happens in the normal use because the saturation level is above the peak
demanded'. You may be defining it to mean 'don't blow up the amplifier'.
But I take that as a safety requirement. :-)
In the above context I regard BJT behaviour as more 'benign' for music as
the devices tend to allow very high peak currents *without* saturation or
failure. Whereas FETs tend to dumbly limit at a specified level, for peaks
or for continuous. That is OK for test sinewaves, but not ideal for music.
How did you monitor for any current limiting or clipping *during the
tests*?
**No. As I previously stated though, I checked the music, prior to the
test with my CRO and did not note any significant issues. Yes, I am
well aware of how transient the nature of current limiting can be and,
thus how current limiting could have occured.
Thus meaning your test can't support the original claim you made.
The problem here is that real-world speakers can be more demanding
of current that you might be aware when playing music.
**Indeed. OTOH, the KEFs were chosen for a couple of reasons. One is the
extensive use of Zobels throughout the crossover and the consequently
relatively smooth 4 Ohm load presented. The ESL63 was not as easy to
drive. In fact, the differences between the two amps was far more
profound with the Quads, than with the KEFs.
If so, that supports what I am saying.
How did you establish the results *were* a basis for conclusions
about one form of transistor versus another rather than being a
problem with some specific designs or devices?
**I didn't, though I've noted (informally) that most (all?) Class A/B
MOSFET amps exhibit similar sonic issues to those I heard several
decades ago.
Afraid that is an opinion, not assessable evidence.
**Indeed. Hence the inclusion of the word: "informally".
And means you can't then quote the results to try and use them to support
your beliefs when making claims about them to other people. As you seem to
have been doing... :-)
Given all the work, where did you publish the results?
**It was for me and to prove a point to some of my sceptical clients.
As you have already discovered, the test was hardly rigorous enough
to publish.
Since your results
seem to run contrary to all the published results I've seen
**Do they? Can you cite where a DBT was performed using a Phase
Linear vs. a Perreaux?
Sorry. I thought you were claiming that *MOSFETS* sounded different as
a class of device to *Bipolars*. Are you now *only* saying that the
Phase Linear sounded different to 'a Perraux'?
**Nope. Those are the two amps I tested. I have not performed any tests
since. All listening has been done on a much more informal basis.
So your claim now is simply about those two amps? Any previous statements
you made should not now be taken to mean you are claiming that MOSFETs
sound any different to BJTs as classes of device?
**Indeed. Perhaps you may care to relicate my test?
You are offering to send me a Phase Linear and 'a Perraux' so I can do
this with no costs to myself?
**Nope. Both amps are common and cheap.
Which should make it cheap and easy for you to obtain them for me.
Afraid I can't recall the last time I saw either amp. Indeed the name
'Perraux' does not mean anything to me as I can't recall an amp of that
name at present.
If they are so cheap, would you pay the price if I found one of each amp
and told you the asking prices? You are probably safe here as I suspect
no-one in the UK will have both amps for sale at any price!
If the latter, it is something I have done many times in past decades.
The results in general were that neither I - nor others I tried them
on - could relaibly tell one amp from another. i.e. they/I showed no
ability to do so with statistical reliability.
**You have directly compared Class A/B MOSFET amps with Class A/B BJT
amps?
Yes. I repreatedly did that in the past. Remember I was involved in
developing a new amp back in the late 1970s and early 1980s. As part of
that I both 'borrowed' a range of designs from others and built a number of
them myself. Then tested them in listening comparisons as well as on a test
bench. At that time Hitachi and others were bringing in their new FET
designs.
During the late 1980s and into the 1990s I also did some similar things
with a few friends, one of whom is actually a very good amp designer in my
view.
The results were as I have said. That provided we avoided simply and
obvious problems like level mismatch, gross distortions, ensured stability,
avoided clipping or saturation, etc, the amps gave no reliable sign of
producing audible differences.
But then we were careful to allow for the way many other factors can affect
perception. So avoided trying to jump to conclusions.
And as I have said, all the well-performed comparisons I've seen reported
come to the same conclusion.
Some reports initially concluded otherwise - e.g, the Stereophile test. But
when examined the results turn out not to reliably do what was initially
believed.
And I'm afraid that these days I don't think it would be fair to do
such a test simply using my own ears. Afraid I am now too old for that
to be reliable. So I'd need to line up a set of listeners generally
younger and more alert than myself. Would you be paying for any of
this?
**Not a chance. If you are sufficiently motivated you will do what I
did. If you are not, then you won't.
I see no reason to bother at the moment since you haven't actually provided
any evidence that you *did* hear a difference due to MOSFET versus BJT as a
class of device. Instead you may simply have found that currently limiting
can be audible. Which does not surprise me as it agrees with my own
experience and would be so for perfectly understandable reasons. :-)
But if you can find an ultra-cheap pair of the amps you used and are
willing to buy them, I'd enjoy trying them. If only to measure the current
limiting behaviours. :-)
Slainte,
Jim
--
Please use the address on the audiomisc page if you wish to email me.
Electronics
http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio
http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc
http://www.audiomisc.co.uk/index.html
Opinion needed re power amp building
Hybrid Mode
