"Robert Morein" wrote in message
news
SNIP
**It's OK. Most do, when I present them with that information. However,
it is the ONLY explanation which makes sense of the "softness" associated
with Haflers, Perreaux and other early, standard MOSFET amps. Even with
the level WAY below clipping, even a crappy BJT amp (like a late model
Phase Linear) the lack of dynamics in the MOSFET amps is immediately
noticable. You, yourself, have acknowledged in your statement that they
are "soft" sounding.
Even the DH-200 was specced
at something like 0.2% distortion at full power. If the amplifier
compressed a full power sine wave, the result would not be a sine wave.
**Except that I am not talking about sine waves. I am talking about fast
rise time, assymetrical transients. Just likke the stuff we get in music.
The Hafler does a fine job of reproducing sine waves, just like any other
MOSFET amp. It's music that it stuggles with.
I appreciate your effort to explain the characteristic sound, but I would
need to see a gapless explanation.
**OK. I'll try to explain later. Briefly, however, consider the effects, at
chip level, when a MOSFET is subject to the heating via a fast rising
transient (which causes lots of current to flow). The MOSFET very briefly
tends to reduce this current, via the negative tempco of Gm. The NFB loop
will tend to counteract this effect to some degree, but, IMO, not all that
successfully. Very high bias designs (ala Pass, et al) get around this
problem by operating the chip at constant high temps. Thus, the effect is
not noticable. Low bias MOSFET amps do suffer with this compression effect.
Sound reasonable?
Your conjecture might be right, but
there's so much of that in this backwater field. In order to explain why a
classic MOSFET design can reproduce high amplitude sine waves but not
transients, if in fact that's what happening, some more work has to be
done. A good explanation doesn't mix the various possibilities together;
it separates them out. Nelson Pass says, "
The Mosfet designs on the market are also Class AB designs. The transfer
curve of Mosfets reveals serious nonlinearities at low bias
**Which is exactly what I said further down the thread. The 'knee' of
MOSFETs is much higher than that of BJTs, yet MOSFETs are rarely operated
with significantly more bias current than BJTs. A critical failure IMO.
currents, resulting in crossover nonlinearity in push-pull designs. This
design flaw makes for a sonic signature that many have referred to as
"Mosfet mist", where a loss of detail is apparent."
His explanation is different from yours; it is unlikely that both
explanations would contribute in approximately equal proportions. And he
isn't necessarily right. Here's my own personal anecdote.
**Perhaps, but I happen to think he is on the money. MOSFETs operating at
elevated bias currents DO sound pretty decent. They do not have the
compression effects I note with low bias designs. They also sound notably
cleaner and more articulate. As good, in fact, as a decent BJT deisgn
operating at significantly lower bias currents.
The Hafler XL-280 is an improved version of the DH-200/220, but it looks
very similar. It has six output devices/channel; the bias current is the
same 100ma/device, it has JFET inputs, but it looks pretty much the
same -- except that there is a small air plate varicap in the output.
This is/was to be used with a comparator bridge lent by the dealer to null
the difference between the amp input and output. I've played with it,
although recently, I've left it untouched. The amp is flat far above the
audible range, with a slewing rate of 75 v/us, but it is said to have a
peak at 170 kHz, adjusted by the cap. This amplifier does not have the
liquid mist of a DH-200; at least one review, perhaps Audio, stated that
it had the greatest amount of low level detail of any amp they had heard.
Nevertheless, the amplifier does sound a little flat, or undynamic.
**Of course. It is a low bias, conventional MOSFET amp. They're all very
similar in my experience.
But when the amps are bridged, the sound changes completely. I use a pair
with the NEAR 50m's. In bridged mode, they sound like a very, very good
amplifier. What happens when an amplifier is bridged? Does the damping
factor halve?
**Yep. Along with a bunch of other problems and advantages. Personally, I
have serious issues with most bridged amps. Mainly because, at clipping, the
results are extremely unpridictable. Damping factors SEEMS to drop to zero
in many amps. Perhaps it is the fact that the amps can never be absolutely
perfectly matched and problems arise due to that. I'm not certain, but
bridging amps seems to improve the dynamic abilities, but it also seems to
introduce other problems which may be worse to many listeners.
The XL-280 has a larger cousin, the XL-600, with a similar
circuit, the same air varicap, and 8 devices/channel, run at 100ma/device.
It has a damping factor of 80, approximately half that of the XL-280. The
XL-600 has enjoyed considerable popularity in U.S. studios, as a very good
all-around amp. It sounds indistingushable from the bridged XL-280's. I
have both in my rack, but I use the 280's in the winter, because they have
no fan.
I wen through the above to show that the MOSFET mist, which still afflicts
certain modern designs, such as the ADCOM pieces, can apparently be
diminished or eliminated to the point where the result is considered
highly desirable, at least by some ears, while still maintaining the
traditional zero-voltage gain output stage.
The distorted
sine wave would be decomposable into a Fourier series with higher
powers. The extreme of this form of compression results in something
approximating a square wave. Please explain how an amplifier can
compress while having excellent harmonic distortion.
**Er, because it can.
Did I see your hands wave? Please say that again with your hands in your
pockets 
**Sorry. I couldn't resist.
Because of gate capacitance and on resistance, MOSFETS do compress,
**Their compression, I believe is more to do with the negative Tempco of
Gm.
but in
order to meet rated distortion at rated power, they do this at power
levels above the rated power. This is not an egregious flaw; the amp
lacks headroom in a manner similar to Class A amplifiers. But the
classic MOSFET design also provides a benefit similar to pure Class A
bipolar amps; extremely low distortion at low power.
**Only because it uses massive amounts of Global NFB. Due to the high
levels of distortion at low currents, MOSFETs need to be biased on real
hard, or be used with huge amounts of NFB (usually Global). Except for
very high bias designs, there are no (to the best of my knowledge) Zero
Global NFB MOSFET amps.
The result is an ideal
clipping characteristic.
**Sort of.
I'll admit it has a characteristic sound, but imho,
you come down hard on what some of us consider a very useful
characteristic. I can drive one of the classic Hafler amps with a
source that has a very high peak-to-crest ratio, such as certain piano
music. At high volumes, any other amp exhibits some audible clipping.
**Nope. Haflers are no different to many other amps, WRT clipping. Clip
them hard and they sound harsh and nasty. Possibly less so than many
amps, but nasty, nevertheless. BTW: There are many other schemes which
exist to keep an amp (BJT) from exhibint unpleasant clipping
characteristics. Most involve sacrificing a few Volts of headroom. NAD
is one such promoter of "soft clipping" type systems. There are others.
They cost a little more, but can provide demonstable benefits. The best
systems keep the anti-clipping system outside the feedback loop.
The Hafler, never.
I have a NAD PE receiver, and my perception is that it does not clip as
cleanly as a Hafler.
**Which model?
7225PE
**AHA! Note the PE moniker. They are sonic disasters. All of them. The
clipping characteristic can be utterly appalling in the PE designs. Try it
sometime (not connected to speakers!). Connect to a dummy load, hang a
'scope off the end and crank the volume up to clipping. When the high
Voltage power supply shuts off, the amp falls into VERY severe clipping and
the resultant sound is terrible, as is the potential for speaker damage.
ALL of the PE NADs are utter horrors. They clip badly. Soft
clipping notwithstanding. In any case, I used the NAD as an exmaple. The
NAD scheme is primitive and not all that good, but it does work. NADs
clip more gracefully than most of their competition.
When driving Acoustat panels with a DH-200 or XL-280, a
deliberate twist of the volume knob full clockwise (which the Acoustats
can tolerate, they're rated at 117db), produces no audible clipping. The
amplifier has to clip, but does not manifest a typical clipping
signature. In mentioning Hafler, I specifically mean only those based
on the original Hafler design: the DH and XL models. When Rockford
bought Hafler, and then Acoustat, they eventually replaced the original
Hafler design completely with the Transnova circuitry. As much as I like
them, these amplifiers do clip audibly, because they do drive the
MOSFETs straight to the rails.
**They use a higher Voltage rail for the drivers?
The rail voltage in the Transnova is 72V. However, because the outputs
have voltage gain, the drivers are operated from a regulated 25V rail.
**Interesting. Do you mean that the drivers are operated from a supply which
is operating 25 Volts ABOVE the 72 Volts?
--
Trevor Wilson
www.rageaudio.com.au