Audio, hi-fi and car audio - Valve amp (preferably DIY) to drive apair of Wharfedale Diamond II's

In article , Patrick Turner
wrote:

I'd also be interested to know the (complex) o/p impedance as a
function of frequency and perhaps power level....

Ro = 0.5 ohms at 1 kHz, with about 7 uH inductance in series. The
leakage inductance in a tube amp is similar to the LR zobel network
fitted to the output of an SS amp to stop the transistors ****ting
themselves with a capacitive load.

Slight typo, there, I think. :-)

There is no typo.

Didn't know organic transistors were now available. :-)

In this context, I don't know what your comment actually means. It isn't
necessary to employ an output inductor with SS amps. Depends upon the
design. Can't say more unless you define your "****ting themselves" claim.

What about as you go down to 20Hz?

All my SE tube amps and other PP tube amps can make 20 Hz at no more
than 2 dB below the 1 kHz clipping levels.

My question referred to the output impedance you quoted. You gave a value
for 1 kHz, and I was asking what happens as you go down to 20Hz to the
output impedance. I'd also be interested to know how the output impedance
varies with power level.

None are used anywhere near full power, and the response is -3 at 5 Hz
if you want it that good because that is determined by CR couplings.

You say that your amps are never used anywhere near 'full power'? Is this
because you only use them yourself, and monitor them to ensure this? I had
thought from you comments that you'd made them for other people.

FWIW I tend to regard the RC shunt as 'Zobel', and any series parallel
pair of LR as being something else. In the power amps I've done the
inductor would be shunted with a resistor as well.

Its normal to have an RC zobel network of say 0.27 uF and 6.8 ohms
between the the output point off the transistors or mosfets to ground.
This provides a 6.8 ohm load to the output devices at HF, so it limits
gain, and helps stabilise the amp at F where it is never used, say 200
kHz.

I'm not clear on what your "stabilise the amp at F were it is never used,
say 200kHz" means.

Don't know a lot about MOSFET designs as I've tended to use bipolars.
However the shunt RC people use tends to be to define the load seen at RF.

The LR zobel network is placed between the active device outputs and the
speaker terminals to prevent absurdly high currents at HF from occuring
into low value loads, or capacitors which have a low value at HF. 5 uF
is only 1.6 ohms at 20 kHz.

In some cases, perhaps. In other cases it is there to serve two other
purposes. One is to define the load at HF and decouple from the speaker and
cables to help reduce RF injection, etc. These should not have too much
effect with, say, 20kHz 'squarewaves' into something like a 2m2 cap load as
the HF rolloff can be defined at the amp input easily enough, and that then
determined the required load current in such cases. In my experience is
isn't necessary in SS to use an output series inductor for that purpose.

I dunno what speakers may present 1.6 ohms at 20 kHz, perhaps it is a
the larger Mrtin Logan types; I have never been able to find a schematic
or impedance profile of those speakers.

If you want low/cap impedance around the 20kHz region then try something
like a Quad ESL57. :-)

However in my experience the output (passive) network is to do the things I
described above. It should be possible to design an amp that is
unconditionally stable without these if desired.

If you have a 1 ohm resistive load, then any amp really struggles,
unless it has an enormous output device number or has low voltage rails
or has transformer coupling. If you have enough secondaries on the OPT
of any tube amp, its possible to get your 50 watts into 1 ohm qhite
easily.

Yet it is perfectly possible to build and use SS amps that work happily
into 1 Ohm loads. FWIW the 732 I designed 20+ years ago will deliver well
over 1kW/channel into 1 Ohm loads on a sustained basis. (I found this out
by accident one day when I accidentially connected a pair of 300W 1Ohm test
loads instead of the 4 Ohm ones and only realised when a protracted test
caused the resistors to start melting their way through the bench surface.
;- ) The amp gave no sign of 'struggling' although the transformer buzz
was a little more audible than normal.

There may well be domestic audio valve amps that can do the same, but I
don't know of them. :-) I'd agree that this may not have much to do with
music, though... ;-

And if you have 3 drivers in a system each with different Z, you are
able to load match with a transformer very efficiently with a few taps.

How do you do that "very efficiently" when the speaker impedance swings
about from below 4 Ohms to over 20 in various places over the frequency
range and the user has to choose just one tap/ratio for the entire band?
Also, what does it do to the distortion behaviour of a SE amp to do this?

But with the tube amp the leakage L is included in the FB loop,
wheras the L on the SS amp isn't, and many SS amps give a worse
peaked response into capacitive loads.

Afraid I don't have the statistics to know what your 'many' might mean
here. Can't say I am concerned much if a passive LC resonance occurs
at HF well above 20kHz and them amp is unconditionally stable.

Making tube amps to be unconditionally stable is a careful exercize.
Imagine taking the feedback from after the LR zobel network fitted to
most SS amps, but without the R in parallel with the L . This would be
difficult to do with SS amps, and its never done, because it would make
the amount of gain and feedback impossible due to stablility issues at
HF.

Indeed. Fortunately, it isn't necessary with most SS amps I know of and
the o/p network can be treated as an entirely passive 'out of the loop'
network to define the o/p impedance seen by the am at RF. The difficulty
with valve amps is - I presume - the distortion, etc, that tends to arise
in the o/p transformer. Hence to keep the distortion down and get a flat
response, etc, you tend to need to include the transformer inside the loop.

But an SS amp needs to have 10 times lower thd at the same
levels of the tube amp, say 0.003% because they mainly operate
in the middle of the switching region of the output transistors,
and although the thd is low,

I note the "mainly" in your statement. :-)

Well, they do.

All that does is repeat your initial claim, but without giving any
actual supporting evidence.

In a class B amp, or an AB amp where only one side of the PP circuit is
conducting, the PS caps which are being charged by the rectifiers, and
which have a varying level of 100 Hz saw tooth wave on them are in
series with the transistors, valves, etc, and the load.

In principle, yes.

Some of the 100 Hz modulates the tones of the music, and NFB is used to
clean the mess up.

This is where you assume any such 'modulation' is significant in level. It
is easy enough to design a SS amp so that even without feedback it tends to
ignore rail fluctuations. In the context of comparing with valve designs
the obvious example being to use a long-tailed pair on a current source to
operate in a differential manner.

Once you design with this in mind in the first place, then any feedback
does not really have much of a problem to deal with. Probably no larger (if
not smaller) than that due to lack of perfect symmetry in a classic PP
valve arrangement.

TBH these points don't seem to me to have much to do with transistor versus
valve, but are simply a matter of decent design using whatever methods suit
the designer.

Can you give me some references that establish this on a statistical
basis for a representative sample of commercial designs? I'd also be
interested to see your definitions for some of the terms you use like
"middle of the switching region". Also for why it has to be "10 times"
as opposed to some other value, etc....

You may do your own research on the issues raised. I picked arbitary
figures.

I was assuming you claim was based upon evidence. Do you not have any
evidence for what you said? You have given no real reason so far for the
values you now say are "arbitrary" should be taken as validating the claim
you made.

FWIW my own work (nearly all 20+ years ago) gives me no reason to simply
accept your assertion on the above point. However I am out-of-date in
various ways, so I was assuming that if you made such a specific claim you
were doing so on the basis that you had specific statistical evidence.

Some SS designs are better, some are worse.

I suspect the same could be said about valve amps... :-)

The switching region for many amps is at a voltage less than needed for
1/4 of a watt, but that's where a lot of folks listen.

Again, I'd be interested in the statistical basis for your "many" here.
What you say sounds plausible, but I have no idea if it is really correct.

1/4 of a watt produces 84 dB SPL into 90 dB sensitive speakers. Most
wives would tell you to turn it down. Are you deaf? they'd ask.

Wouldn't matter (until later!) if you could not hear them at the time. 8-]

However the speakers I use mostly (e.g. ESL988's/63's) seem to have
sensitivities somewhat lower than the 90dB/'watt' you seem to assume. Don't
know what is typical with modern speakers, but looking quickly through some
of the old review collections I have seem to indicate that values around
80dB/'watt' are common. Couldn't see one that was 90dB/'watt' although I am
sure they exist.

But the thd was modulated by the rail swings from the mains jitter,

"Mains jitter"? Do you mean ripple?

No. the mains constantly varies in level due to streetfuls of ppl
turning things on and off and thus the voltage "jitters" up and down,

The term "jitter" tends these days to normally be used for timing
variations. Hence I'd have described what you describe here as amplitude or
voltage variations.

with some spikes of 50 mV.

Interfererence.

So the DC voltage of the rails also tries to follow a fraction of the
mains variations. It many be only a few mV in the case of a transistor
amp, but its effect will make the amplitude of the distortion vary like
an AM wave envelope.

As per the comments I made earlier. Yes *if* the design is sensitive to
rail variations. However by design you can deal with this.

You may have seen some recent work published by Keith Howard in 'Hi Fi
News' where he looks at variations in performance like this. Some amps do
such things to a clearly measurable extent. Others do not. Design.

Its clearly visible on the CRO whilst taking measurments, along with any
hum within the output signal.

It can be, depending upon design, etc.

FWIW as I have commented elsewhere, when designing SS amps 20+ years
ago I used to deliberately modulate the power rail(s) of a working power
amp and then check to see if this appeared directly, or cause distortion,
at the amp output. Standard technique, I assumed.

Again, though, this is a matter of design and fabrication, not - so far as
I can see - one of transtors versus valves....

so the thd level changed dynamically.. Under normal conditions,
class B amps have somewhat large swings on their supply rails and
these modulate all other frequencies. Unless one regulates the rails
or uses 100,000 uF caps, the thd is far worse than what one measures
with a sine wave.

Partly, I'd agree. That was why I tended to design amps with care
w.r.t. being good at rejecting powerline fluctuations. (Must admit to
also being a fan of big caps... :-) ) Also why I used to test for the
kinds of non-sine and asymmetric waveforms implied in what you say.
That said, I didn't make any Class B designs, just AB ones.

AB amps suffer all the bothers of class B amps, except that thier rail
caused IMD bothers begin at threshold of output level higher than most
"class B" amps, which in fact do have some quiescent current, but its
often only 25 mA per output transistor, allowing only a 50 mA peak
current swing in class A, and not a very linear bit of class A at that.

Again, you seem to be making sweeping general claims, but have not yet
provided any real evidence I can see. You also, again, seem to be assuming
that problems you may have encountered in the above respects are endemic
and unavoidable. My experience leads me to doubt this - unless you have
some actual evidence?

Class A tube amps have the advanatage ofr common mode rejection in
the CT OPT output stage. SE class A tube designs have a continuous
drain of power, and in any case huge supply caps are used, say 1,000
uF.

All class B amps which include nearly all SS amps used today

My understanding is that most are AB, not B. Hence I would not agree
with the statement you make above.

Nearly all class AB amps made with SS are AB in name only. The techs I
know all refer to them as class B because the quiescent current is so
tiny, and virtually no class A power is made.

In that case they you/they may be employing your own definition of the
term which may not be the same as the standard definition in each case.

My understanding of the usual distinction is not quite as you describe.

The point is not that "virtually no class A power is made" (slightly odd
wording, but I think I understand).

The point is that the AB bias and arrangement allows the design to
effectively remove crossover distortion by appropriate biassing. i.e.
to obtain a situation where in normal use the distortion level falls
monotonically with reduction in signal level. Thus avoiding the
'classic' symptom(s) of something like a discontinuity.

I would agree that if the quiescent current is set too low, then
this might not occur. But this is not the same as how much 'class
A power' is available, but of what occurs over a range of signal
sizes both larger and smaller than this (implied) level.

really need all the NFB they can muster because of the nature of
their intermodulation production.

I'd guess that Class B would be difficult. However this is why I've
avoided it, and assume that other have as well.

Bryston apparently don't. They use both npn and pnp on both sides of PP
circuit. The low bias current operation is very good indeed.

Afraid I don't know the circuit you are referring to so can't comment on
it. I have no reason to doubt your comment about the low bias operation
of this specific design, but do not regard that as evidence of the
other points you are making in much more sweeping generalisations.

Usually, because SS amps measure 10 times less thd than a tube amp,
they often cannot be distinguished from a decent tube amp which
measures well enough. But I think the dynamic distortion mechanism
within SS class B amps is 10 times

Well, in measurement terms, the valve power amps I've seen in
magazines sometimes seem to give THDs somewhat more than x10 the
values for the SS designs.

An SET 300B amp at full power might make 5% thd. This is ten thousand
times the thd of an Halcro at 8 watts.

Be interesting to see the relative figures with reactive loads, and/or at
low/high frequencies...

Both amps at 1/2 a watt has sufficiently low enopugh thd to allow great
listening.

Fair enough. Can't comment for the same reasons as I give above. But
subject to the same caveat re not using this as a basis for sweeping
generalisations. :-)

However I have no statistics for this. Can't say that it bothers me
much once the nonlinearity is reasonably low. (And, of course,
avoiding Class B.)

worse than the mainly class A tube amp, so the ten times greater NFB
amount leaves the two genres somewhat similar sounding to many
people. But not to all ppl, and some hear a lot more though a decent
tube amp, and it ain't got much to do with measurements if the
buyers of SET 300B amps are witnessed.

Regulated rails are one solution for class B amps.

Again, I can't actually recall designing or using a Class B amp. I've
used/built A and AB, but not B. That said, I'm not clear why Class B
amps should require regulated rails specifically.

AB and B amps all benefit with regged rails.

Despite the sweeping assertion behind your "all" I'm afraid that my
specific experience is to the contrary. I decided that for power amps that
are well designed w.r.t. rail rejection that regulated rails caused more
problems than they 'solved'. After working on this for a few years I
concluded that it was better to cure such amplifier sensitivies by
redesigning the amp rather than trying to protect it.

Hence my own experience is that *some* class AB designs show (in my
judgement) 'benefits' from *not* using regulated rails. Can't comment
on "all" except to point that even one such case makes your "all" claim
unreliable. :-)

Above said, I am a fan of nice big rail caps.

To make class A in SS, you need to have idle currents in amps, not mA,
and that causes large ripple voltages and PS charge currents in the OV
rail circuits. Using CLC filters or reulators like one would in a tube
amp reduces the need for so much NFB, and a total of 25 dB is usually
enough to get to 0.2% at 40 watts,

with very small amounts of thd at 2 watts, where most listening is
covered by. Most SS class B amps have perhaps 40 dB of voltage nfb in
their emitter follower output stage, as well as 6 dB of current nfb in
the current sharing emitter resistors, then another 60 dB of global
voltage nfb is applied.

That makes a total of 106 dB of NFB.

Once again you make all kinds of assumptions. :-)

[snip more assumptions]

0.01% of 4 vrms = 0.4 mV. Now if we banished the wanted signal and
played the 0.4 mV of distortion only, then who is going to tell me 0.4mv
is audible at 3 metres away with 90 dB speakers?

Assuming 90dB/'watt' speakers... :-)

Also, of course, your "play only the distortion" neatly sidesteps the
issue that the speakers when playing such signals as caused such amp
distortion may well be distorting as well. Albeit in a different
manner. Also ignores the nonlinear and masking properties of
human hearing. Hence even given all your assumptions I am far from
sure that you can draw the implications you seem to wish from what
you describe.

Maybe with horn speakers of 100 dB one could hear the thd in two watts,
but then we'd only need 0.2 watts for the same level, and the thd
voltage at 0.01% would have fallen to 0.04 mV.

A lot of what you say is based on a series of assumptions.
Unfortunately, you still have not really given any reliable evidence that
these assumptions are all applicable in the "many" or "all" cases you
either claim, assume, or imply.

Some of the things you say may be so in your personal experience. However
in some of the things you say, my own experience is quite different to
yours. Hence without more reliable statistical/general evidence I
don't see that you have established that your sweeping claims are
as reliable as you assert, I'm afraid.

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html