Re: Is the Williamson still a good amplifier to build?



Bret Ludwig wrote:



The original design, partly because designed before modern
insulators, etc., and partly because designed before modern
bean-counting, is a pretty serious piece.

My comments are also in the context of a serious long-time
hobbyist perspective, where large stocks of most other parts
are sitting around looking for a good use. Large stocks of
ancient mil-spec transformers, oil caps, etc. are a given
for folks like Bret, and, I'd guess you too. I know that
fersure I won't outlive the iron here at the shack.

The original design had four stages. It is not easy to stabilize
unless the transformer is built exactly to RW's exact specification.
It had a FR to about 0.5MHz IIRC. I built two of them,
many years ago. Wish I still had them!

The Willy amp was a wonder which wooed woofers wonderfully with
bass wobulations.

But in 1950 the speakers of the day were all mostly 95dB efficient,
because makers knew folks hated spending the huge sums of money
for what was advanced and temperamental electronics in the form of class
A amps with lots of NFB.

But these days, 16 watts of class A sound from a pair of KT66 is mostly
woefully inadequate
because today's speakers have reduced sensitivity to about 88dB/W/M if
you are lucky.

So one should have KT88 in triode, at least, or UL. The DTNW driver amp
may still be used of course.
The driver bw IS GOOD because the concertina phase inverter acts as a
buffer between
input stage and balanced amp and Miller C is overcome, and driver bw is
wide.

The biasing of the original DTNW amp is a stupid idea, and the balance
in the balanced
amp can be bettered and made automatic. Don't tell me DTNW's ideas can'r
be bettered.

However, because there are so many stages, and all that CR coupling,
the Lp must be high, and 100H is the minimum needed with RL = 10k,
and Ra-a of 3.2k of two KT66 in triode.

The -3dB gain pole is about 3Hz.

It need not be so low, but its only this low because of stability issues
with the original DTN circuit. To avoid building what might otherwise
be a phase shift oscillator due to phase shift at LF with CR couplings
and inductive
shunting of Ra-a at LF, the 100H was required AT LOW SIGNAL VOLTAGES.
inductance varies with applied voltage amoung other things,
and Lp diminishes at low signal levels, so its possible a DTNW amp
can oscillate at LF woth a slow F which cannot build up to a large
signal amplitude
because the increasing inductance reduces phase shift.
Old DTNW *didn't know* a good easy way around this problem, and
advocated the use
of 4,400 P turns around a core Afe = 32mm x 44mm.
This certainly gave plenty of L, but leakage L increases with Lp
which are both proportional to the square of the turns used.
So if P turns are doubled, Lp rises 4 times, and LL also 4 times,
and HF bw reduced from say 50kHz to 12.5kHz for a given number of
interleavings.
So then to stop this, the interleaving must be increased to the
SPSPSPSPS arrangement, ie, 5S plus 4P sections.
This creates a lot of capacitance between windings, but with a large
window the insulation can be kept
in excess of 0.4mm between P&S and BW of about 80kHz is achieved, and
the 20dB of NFB
can be applied without HF instability after gain step networks are in
place.
Nobody working at home in 1950 to build a pair of DTNW amps had a CRO.
They'd have a voltmeter perhaps. They'd by watered down versions OPT.
The resulting amps were great oscilators with any sort of C load,
and would oscillate themselves to death with no load connected.

By 1960, many home made DTNW amps had been happily dumped.
A cause of failure was the death of tubes. As tubes die, they often
become gassy,
and the grid cannot control Ia flow, so the tubes go into thermal
runaway.
This all so very easily fused the tragically thin fucking wire used in
most
1950s OPTs. So by 1965, many owners of tube amps had witnessed their
tubes dying, and OPTs frying,
and then they went to a shop, to see what could be bought.
Usually there was onlt ONE amp to dump, because sound was only mono.
When stereo finally became desirable by 1965, just like surround sound
is now
so desirable for HT, Solid State had arrived in force with TWO cool
running amps in a single lightweight
box, but it was still expensive......

The main rival to the DTNW amp was the Mullard range of amps, the 510
and 520.
leak based their ideas around Mullards.

No matter who made or designed them, a large number were just dumped,
and good riddance I say!


The GOOD parts are in Asia, have been for years.


?????


Oil caps that are
old and of uncertain provenance are best gotten rid of as many had
Askarel (PCB) filling and if they leak and the enviro people get wind
of it YOU can be in big trouble economically. There are a lot of old
power transformers but they tend to be odd voltages, and with copper
where it is a lot of big stashes of them have met Mr. Scrap Man in
recent years.

The DTN Willianson opt design was not that good from some
standpoints, most of the commercial ones were a lot better.

Don't you mean commercial samples were a lot worse?

Most commercial designs dumbed down DTN's original ideas.

But it depends what "better" means.

The DTNW like all the Mullard muck and Quad-II and other junk from the
late 1940s and 1950s could always easily be bettered, in terms of
ensuring maximal closed loop bw at full power without any one stage
saturating.

But for 1/2 a watt average power into sensitive speakers, a
pair of 6V6, EL84 in pentode or UL, or a pair of KT66 in triode
were quite adequate, even with crummy OPT.

At low power levels, one simply does not need to have the best OPT
available.
So bean counters ensured nearly all OPT in commercial amps were crummy.

I refuse to honour and respect the mediocre attempts at hi-fi
during the 1950s. It was largely a sham.

My father's generation mainly thought hi-fi and stereo to be frivolous
extravagance,
and most demand was for mainstream rubbish.
I occasionally try to render such junk less awful sounding than when
presented to me for
singing lessons.
Usually the OPTs are a complete joke, tubes operate in AB pentode,
and the circuit paramters are all wrong, and stability margins are
woeful,
so NFB applied is low, and sound is typically brackish, or HF edgy,
sibilant,
and bass is boomy or just missing.
Such horrors often had the appropriate tine control stage to truly maul
the music,
and many integrateds had phono amps with extremely wrong RIAA......

The "tube sound" of the 1950s was often that of excessive IMD,
very poor damping factors, and poor frequency responses.

A very tiny minority of ppl were devoted to true hi-fi, and none liked
to spend,
but were forced to, but they were not anywhere near enough top create a
mainstream market
for routinely REAL hi-fi components which were also reliable, and which
followed all RDH4 rules.


Several
vendors in the US and UK made suitable ones but Partridge in the UK
and Peerless in the US were considered the best. I don't think any of
the commercial ones followed the Wireless World "wickelschema" as the
pros got a lot better fill factor and balanced copper and iron losses
much better.

Just follow the design ideas best expressed at my website to improove on
what
DTNW first designed.



Partridge was a big believer in putting some air gap in all his push-
pull transformers to linearize them at the expense of a substantially
larger core than his competitors, particularly in the US.

The air gap in a PP OPT is a darn good idea.

I have one in my OPT for class A PP mosfets. It stops severe
current spikes at saturation just below 15Hz.
The gap forces the transformer to act like an inductance,
and not a defacto saturable reactance

However, the Lp is much reduced because the effective iron µ
is well reduced, as it is in any air gapped core for an SE design.

Follow my ideas at my website on iron cored OPT and choke design to
figure out why.

Hence it would be quite possible to build a DTNW OPT with a gapped core,
but since the original lamination size is not easy to find,
50 tongue wasteless E&I must be used instead of the original 32 tongue
which had a non standard window size of 76 x 25.
So a DTNW OPT would now need a 50 tongue wasteless core stack of 28mm at
least to give the same Afe as the same as the original
of 32mm x 44mm. The result would be heavier than the original, and bean
counters
would shoot me for saying all this.

The µ of the best 1947 GOSS available was about 5,000 max, and with the
original DTNW OPT
max Lp was about 600H at Va-a of say 200Vrms, but only 100H at 5Va-a.

With a gain stepping network to curtail open loop gain below 20Hz where
the less gain there is, the better,
the requirement for 100H can be greatly reduced, and all we have to
worry about
is that saturation at full power is not likely above 20Hz.

So the P turns could be reduced to 2,200, wire size increased about 1.35
times in Dia,
and LL is reduced 4 times, Lp also from 100H to 25H at low Va-a, and
if Ra-a = 2k with a pair of KT88 in triode then the LF pole is at
12.7Hz.

There isn't "much room" to allow an air gap.
But usually, the air gap used isn't a real full gap, the effective µ is
reduced by
placing Es and Is into the bobbin in groups or bundles of say 10 at a
time,
and alternated in direction.

One can adjust the wanted µ from maximum available to anything one
wants.

Modern GOSS material I use now has maximum µ when fully interleaved =
17,000
which is a 3 fold improvement over the days of 1947.
Therefore there is considerable scope to instal an air gap
and reap the benefits.

But really the gapping isn't a big deal where levels are low, say 1/2 a
watt average
out of a possible max of 30watts AB from say a pair of KT90 in triode.
Core saturations do NOT manifest themselves providing dc bias is
adjusted
properly.

See my design pages for my design for OPT No1 etc to see
a better alternative to DTNW's ideas.

Instead of 4,400t of dangerously fragile thin wire with high winding
losses
around 32 x 44 Afe, I use 0.4mm dia wire, about 2,200 turns, and core 44
x 50 at least, but could be 75mm
for those wishing for best performance in line with fine print
recomendations in RDH4.

Chicago Standard built decent transformers for Williamsons, as did
Triad, but the Peerless was the preferred one for several reasons.
Ercel Harrison's big compulsion was full symmetry of the copper path.

There is absolutely NO need for two bobbins each with 1/2 the PP primary
and secondaries.
I have tried all this and found the performance was not any better at
all to
the designs shown at my website.

You can do a heck of a lot better than 99% of the bean counter designed
efforts
which were presented to us and which are still being presented to us
by mainstream amp makers and OPT makers such as Hammond.

Most of my living this last year was devoted to re-wiring many existing
amps
and OPT quality varies from utter rubbish in all asian/chinese amps, and
"made in UK " CR Audio amps,
and in many US made low brow amps from the 1950s.
Manley and ARC make pretty decent OPT, but don't know how to load them
properly,
and use far too low a load than should be used to maximize class AB
power so
the quoted power figures impress stupid buyers led astray because
they think they *must have* 120 watt capability.

The OPT is one thing amoung the many things that contibute to final
sound quality!!

If I had a dollar everytime I heard BS about OPT, i'd be rich.

I cannot say all that needs to be said here, but if you want a blameless
OPT,
then follow RDH4 to the letter, or follow my website.

Punters should feel free to experiment with partial air gapping,
and core material mixtures, such as using 50% nickel laminations
interleaved
with GOSS lams.
C cores are also very nice, and then there are amorphous cores.
Lundahl use them.

However, having said so much positively about GOSS, it IS POSSIBLE
to get very fine sound with much lower µ iron than modern GOSS.
There is such stuff called non grain oriented silicon steel ( NOSS )
which has a µ of around 3,000 max
and it is 1/2 the price of GOSS.

The Fsat is about the same, and distortion generated by the iron
is far higher than GOSS, but still much less than the tube distortions
except when F gets really low near saturation at high levels.

A sample of the 8585 amp I made has 62mm stack of 44tongue NOSS
material,
2,080 P turns, and gives utterly blameless sound quality.
The sample now has KT90, and gives 100W from each channel.
OPT weight is 50% larger than in ARC VT100.

You can't beat having a shirtload of iron!!

But its only ever any good when followed up with careful design
for minimal LL and Csh.

Patrick Turner.
.