[Patrick Turner]

Thanks to Roger Rosenbaum for sending me the 1956 Crowhurst ideas on OPT
design.

Crowhurst expresses some basically solid ideas for OPT design and
particularly
he promotes GOSS C cores instead of E&I lams because the core
permeabiity ( µ )
is 3 times higher than old E&I lams, therefore 3 times the inductance
can be had
for "free" with the same size of core leg area and turns for each type
of core.

This was in the days when good GOSS had a max µ of about 5,000, and
ordinary non oriented steels went about 1,000.
But now the GOSS even in E&I lamination form can have µ = 17,000 max,
and old NOSS which is 1/3 the price can be 3,500.

So today a decent OPT can be made from either GOSS or NOSS.

The glaring omision by Crowhurst is that for good OPT, the design is
deteremined by the Fsat, frequency of saturation
which is about the same for all iron, ie, at about 1.3 Tesla when
distortion currents
begin to exceed roughly 3%, and so to get decent low distortion at bass
F
one should design with Fsat in mind from the outset.

Therefore at full PO of the amplifier, you are doing well with an OPT
design if
Bmax = 0.3 Tesla at 50Hz, which means 0.6T at 25Hz, and 1.2T at 12.5Hz.

If this condition is met, then there usually always is enough primary
inductance for whatever
series source resistance and shunt load resistance exists.
So a pentode amp with Ra-a of 40k with a load of 5k has a total terminal
source resistance of
40k//5k, approximately 4.44k.
NFB may drastically reduce the series source resistance as does the use
of UL, CFB or triode connections.

Once you base one's design on saturation, then Crowhurst's
recommendations with what cores were available in 1956
don't add up to a huge amount.

In my 300W amps with E&I lams, the Z ratio is 1.3k : 6 ohms, for use
with 12 x KT88.

The maximum primary inductance is many times what it needs to be because
the iron µ is 17,000 maximum.

In fact, I could afford to just butt all thr Es against all the Is and
reduce
the µ much less than 17,000, and still have enough primary L.

What matters is the saturation, and in my design I have a core of 110mm
stack of 51mm tongue,
and 1,060 P turns, thus giving 482H max.
If the load is 1.3k, then -3dB is at 0.43 Hz, and so L could be 48H max
and still be plenty.
But 300W into 1.3k = 624Vrms a-a and Fsat with B = 1.5T for the GOSS
occurs at 16Hz, and
thus bass response is fine at full PO down to a low enough F.

If I used Crowhurst ideas, and said 48H was enough, then the turns would
be 1,060 x 1 / sq,rt 10 = 1,060 / 3.16 = 335,
but Fsat would move up 3.16 times to 50Hz, which is way too high for my
standards.
Of course with only 335t for the P winding, BW would increase a lot
because leakage L also
reduces with primary inductance.

But with my 1,060 P turns, and an interleaving pattern with 6 S sections
and 5 P sections,
I manage to get 16Hz to 270kHz with the rated 6 ohm load connected.

C-cores could be used for a design like mine, and many C-cores
have a much larger window area to core leg area ratio, so the volume of
the winding
can be greater than the volume of the core. In my designs its always the
other way around,
core volume winding volume.
If the core volume is halved, winding volume has to double for the same
PO handling.
If you halve the core centre leg area, then you must use twice the P
turns
for the same Fsat. This also affects the P inductance, and with the 1/2
x Afe, and 2P,
you get twice the Lp, and more than 3 times the leakage L because LL
varies as the square of the turns.
So the increased winding volume may not be so good at HF as where you
have a huge core and small coil
to reduce Np to a smallest mumber of turns for lowest Fsat, thus gaining
enough Lp,
and also getting a huge BW.

Sorry, folks, but youse ain't ever gonna get lightweight transformers
out of me!

So after finally reading Crowhurst, I doubt I see much to gain from.
His diagrmatical way of explaining leakage inductance is fine, but
readers of 1956
would be utterly be-fuddled to try to decode what on earth all those
fancy smancy graphs he's drawn
might possibly mean.
See the simple formula for leakage inductance at my website; its far
easier to apply that
trying to use graphs by Crowhurst, or use anything in RHD4.
All of what was said by Crowhurst and others was later interpreted by
some other brainy blokes and
my formula is a regurgitation of what I read in Wireless World, a
reputable magazine if ever there was one.

Anyway, it works.

Capacitance calculation is also a mystery after reading Crowhurst.

But the main C to worry about is the total C looking into
each anode connection.

In an interleaved OPT, the shunt C between a layer of P beside a layer
of S
can be approximately worked out as if they were equivalent metal plates.
With a 6S x 5P winding pattern each shunt C between P and S
occurs at fractions of the P winding, and they are transformed down in
value
by the Z ratios of the fractions of windings, then totalled.
So suppose C = 500pF between one P to S winding regarded as an earthy
winding.
say you have S-P-S-P-S-1/2P from an anode end to the CT, then
you'll have C at 1/2 way as being 1/4 x 500pF as seen at the anode end.
Typically, for the interleaving just stated, you'd get
total shunt C = ( 500 + 400 + 130 + 100 + 50 ) pF = 1,280pF.
This is 640pF a-a.
The leakage L is low enough so that the resonance between Cshunt
and LL is way above 100kHz, and of little concern, because open loop
gain will have been made low
at that F and above.

In any OPT, there can be a large number of small value LL between all
the interleaved parts,
and they all interact to form what can be a complex multio order low
pass filter.

Trying to accurately model such complex L&C arrangements has not ever
been achieved
and there's no program available to dial in all the coil details, then
click enter,
and have a model equivalent schematic displayed and with F response.
Engineers hate things they cannot model.

However, once you have wound and tested a dozen OPT, you'll learn what
works and what doesn't.

Crowhurst is fine reference material, but I hope my webisite is of
greater worth because there
is more info on how-to-do-it.

Crowhurst doesn't dwell on SET amps in the papers sent to me.
But SET are now all the rage for the DIYer brigade.

Well, ensure that ac B max is not more than 0.3T at 50Hz, and that dc B
does not exceed
0.6T because of careful air gapping, and that Lp has a reactance = RL at
20Hz, at full PO,
and , and you WILL get terrific bass.

Crowhurst makes the point that HF performance does not much depend on
the iron core.
In fact, I can say that a decent OPT can have its iron core *removed*
and the
operation above 3kHz will be unchanged.

The µ of iron used in AF trannies falls hugely as F rises, so L also
falls, but
the XL = 6.28 x L x F, so although L falls a lot by 1kHz,
the XL is still very high at 1kHz.
There is an F where the OPT coil has enough L just from being an air
cored winding
without any iron present so that the load R is a much lower resistance
than the
reactance of the primary inductance.

Patrick Turner.

[tubegarden]

Hi RATs!

I never really attempted to wind my own transformers, but, I did try
the Alan Blumlein 'garter' cathode bias current balancing circuit in
some amps and he was correct in saying that P-P output transformers
will handle lower frequencies better, if the current is balanced in
both legs of the P-P primary.

Four resistors and a pair of caps is a small price to pay for obvious
sonic improvement that even I can hear.

Pity he died thirteen years before Crowhurst's book.

Good thing some folks keep regurgitating what they read into the net,
for those of us who can't get to the library, nor manage to read a
whole book, anymore.

Happy Ears!
Al - the layabout who does amps for the joy of listening (keeping my
hands busy in my bedroom is fun, too

PS A thing to keep in mind while designing any machine is that most of
the time, it will be doing nothing. Boredom kills ideas, too.

[tubegarden]

Hi RATs!

Yes, Broskie goes off on that tangent, too

Have you tried the garter circuit?

It is low tech, but high performance, if you take time to match the
resistors. It also works bypassed, or not

Happy Ears!
Al