SE amps - some straight talking, hopefully.

[Roy Briggs]

Firstly, I would like to say that I am new to the net. I had been hearing an
awful lot about the single-ended debate that is going on, but I had no idea
there was such staunch and aggressive feeling about it among music lovers. I
was not prepared to discover how much hatred (and if it's not hatred, then
it certainly sounds like it) can be inspired by an exchange of differing
opinions. Some of the postings in this group on SE amps don't fall too short
of being malicious. Is that really necessary to make a point?
Anyway, I guess if I have to be categorised into a side, it will be with the
'for single-endeds' because I own and enjoy a single ended amplifier I
designed and built myself, and yes, I do believe they are more accurate (and
I have owned a lot of commercial and self-designed push-pull amplifiers). I
hope that I can give my opinions about the technology and why I think it is
better without the venom I have read in some of the pieces posted here - and
without receiving that kind of venom in reply!! ;-)
The idea is pretty simple, or so the theory goes: an amplifier presents at
its output a magnified replica of its input. You could argue that there are
a number of topologies or standard component arrangements for achieving
accurate voltage and current amplification, and that this representation of
an analogue in the form of current and voltage must then translate correctly
into a transducer to provide accurate reproduction of the source, assuming
0% distortion is created and translated to the air mass by the transducer.
So where does it all go wrong?
There are now probably tens of thousands of push-pull amplifiers, and save
for very badly designed ones, most will have negligible distortion and
measure as 'perfectly' as would seem reasonable. I actually read one opinion
that amplifiers that measure the same sound more or less the same, which
makes me wonder about the validity of that person's objective assessments,
because this clearly is not the case. An example where this is very starkly
demonstrated is in coupling capacitors. I know engineers who place no
importance on their quality, simply because unless they are of particularly
poor quality, they do not detract from measured performance - yet in a good
quality amplifier the differences between capacitors are very audible, as a
lot of you who 'tweak' will know! (And it doesn't always turn out that the
expensive ones are perceived aurally to be 'better', I hasten to add).
So does it seem reasonable to assume that a no-capacitor-coupled DC design
would be less 'coloured', because of the absence of inter stage capacitors,
than an AC coupled one, assuming an identical arrangement of components was
possible?
Well I think yes. In fact, it is my deep-seated belief that every component
in an amplifier detracts from true replication of the original, more often
than not in ways that are not measurable - but certainly audible. For me,
it's a fact: fewer components in the signal path, all modifying the wave
form (although supposedly 'invisible' to it) in some way or another, is
better. Take the DPA 1024 as an example, a CD converter that truly redefines
the capability of the dreaded CD medium because once the signal is
translated to an analogue, it does not have to pass through a myriad of
components. I have not heard this converter, but I know that it blew up a
lot of skirts and crucifies all other competition - including the Wadias and
Mark Levinsons etc. And crucifies! It measures very badly, incidentally,
both in terms of harmonic distortion and noise floor.
I would say, at the risk of being crucified myself, that measured
performance is something of a smoke screen for those who quiff at amplifiers
possessing a handful of watts, like the idea that single ended amps should
take a swipe because most speakers are unnecessarily inefficient and won't
work with them.
Amplification, and other audio components that measure badly, does not
always sound 'bad' when it measures badly. Let's face it, be honest with
ourselves, unless something is horrendously wrong, measured performance
doesn't really give any useful insight into how a component is going to
sound at all. Sure, with response curves you can get an idea whether a
component might be 'bright', 'boomy', 'light', and with spectrum analysis
you can judge whether the component will sound 'harsh', 'brittle',
'forward', 'warm', etc. - but whatever your cup of ideal-measurement tea,
you can't really tell whether a violin is going to sound like a violin until
you sit back and listen to the equipment. There are no known measurements
for the most important facets of musical reproduction, facets that help make
reproduced music sound real: stage depth, width, and height; focus; timing;
perceived 'presence' within a sound stage. I could add accuracy, but I think
we'd get back to a sine-wave is a sine-wave is a sine-wave, ad nauseum.
Single-ended working is the embodiment of ultimate simplicity. There, more
or less, rests the case why SE amps reveal so much more musical information
than push pull - they have fewer components to colour sound, they don't
manipulate the wave form in the way push-pulls have to, and they don't
necessarily have to rely on shunt or series feedback to work correctly (and
although it could be argued that push pull doesn't have to either, all too
often it is used to mask poor inherent signal manipulation - a lot of people
would be surprised just how good push-pull can sound when it is designed
correctly). This in itself is a good argument for the case that single-ended
commands good design practice from designers, and a more natural, considered
approach to circuit topology.
Arguments (which I have read) over whether to call a phase splitter a 'phase
splitter' or a 'phase inverter' seem superfluous to me, to be honest. The
essence is this: the 2VBE or diode arrangement usually adopted to perform
this task divides, halves, separates, whatever you want to call it (and all
reputable electronic publications I have read call it a phase splitter, by
the way) - they all 'chop the wave up' into its positive and negative cycle
components and feed the correct half to the appropriate BJT or whatever.
Not in single ended, of course. PNP devices are more difficult to
manufacture, and they never behave in a true inverse way to their partnering
NPN. Depending on topology, correction may have to be applied if distortion
is to be avoided (assuming, of course, that it is deemed that distortion is
not acceptable, as is usually the case, and usually it's applied in totally
overkill proportions). In my experience, high levels of feedback always
produces a sterile sound, one that makes an amplifier easily recognisable as
a member of that genre (a little like most valve amps have a signature, and
also the different valves themselves). It is more or less impossible to
perfectly match an NPN to a PNP device, and usually at some point in the
cycle, even in heavy class A bias, one transistor is in its 0-.6V conduction
state, where its transfer characteristic is not linear. Depending on the
bias arrangement, there will be various intervals of the cycle where the
amplifier is operating in this apparently very cock-eyed way. Perhaps this
is why class A operation, where it's arguable that less of the wave form
could be transferred by BJTs in this state, depending on the design, seems
to sound clearer and more tangible - and more accurate, you could argue.
The cut line?
Whatever aspects of measured performance you place importance on, you can't
tell how 'real' an amp is going to sound until you listen to it. If an amp
measures badly without feedback, then my experience is that it won't sound
better with it - the design, in my opinion, should be binned. Every
component adds a signature to the wave form - period. Therefore the fewer
components, the closer the output must be to the original (assuming this is
intended in the design, of course!). In fact, I believe this in itself goes
a long way to explaining why SE amps have a more natural, realistic
presentation compared with push pulls. Amplifiers loose their simplicity of
signal handling when feedback is applied to them, especially when it is
time-dependant (listen to pre-amps that use feedback to give RIAA
equalisation, and then listen to one that uses a passive arrangement).
For those who are interested, I have designed and intend to patent an active
speaker arrangement that uses transformerless SE amplifiers with zero shunt
feedback and almost zero series feedback which produces an SPL equivalent to
a 100W amp driving average efficiency speakers. Only 1 device modifies the
shape of the incoming wave form in any measurable way. This it does very
accurately and very cleanly, which happens to result in a very good (!)
measured performance.
And how does it sound?
Fans of single-ended amps, can you imagine?

[Stefano Rovetta]

Roy Briggs (Roy.B...@mommys.demon.co.uk) wrote:
: There are now probably tens of thousands of push-pull amplifiers, and save

: for very badly designed ones, most will have negligible distortion and
: measure as 'perfectly' as would seem reasonable.

The heart of the problem is, "do we really know what to measure?".

It is hard to believe that an objective measurement cannot detect an
objective quantity such as distortion. But we lack an insight on what
distortion is important (the same holds for other parameters).

Anyone who tries to use a synthesizer is well aware that
dynamical aspects of sound (such as the envelope) are more important
than statical aspects (such as the harmonic content). The point is:
most of the measured parameters have a meaning only in steady-state
or in particular cases of transient phenomena. An example: the
generalized use of averages (e.g. RMS quantities).

I believe that, if a deeper knowledge on perception could be achieved
and put in practice, the war "measurement vs. perception" will
eventually come to an end.

---

Il cuore del problema e': "sappiamo davvero cosa misurare?".

E' difficile credere che una misurazione oggettiva non riesca a rilevare
una grandezza oggettiva come la distorsione. Ma ci manca una conoscenza
approfondita su quale distorsione sia importante (l;o stesso vale per altri
parametri).

Chiunque provi ad usare un sintetizzatore si rende ben conto che
gli aspetti dinamici del suono (come l'inviluppo) sono piu' importanti
degli aspetti statici (come il contenuto armonico). Il punto e':
la maggior parte dei parametri misurati hanno significato solo in
condizioni stazionarie o in casi particolari di fenomeni transitori.
Un esempio: l'uso generalizzato delle medie (p.es. le grandezze RMS).

Sono convinto che, se una conoscenza piu' profonda della percezione
potra' essere raggiunta e messa in pratica, la guerra "misure contro
percezione" finira' per terminare.

--
------------------------------------------------------------
Stefano Rovetta, PhD Phone: +39-10-353 2268
DIBE -- University of Genova Fax: +39-10-353 2175
Via all'Opera Pia 11a E-mail s...@dibe.unige.it
16145 GENOVA (Italy)
------------------------------------------------------------
The opinions expressed here are neither mine
nor those of my department