Who has experiewnce with 845 use in any SET amps?

What do ppl think about what is said about the 845 at
http://www.supertnt.com/default/news20.asp ?

Where is the best supplier?

The 845M looks promising.


From what I have read, i wouldn't run 845 at more than 70w pda.

Patrick Turner.

Many people have built 845 SE amps. It is a very popular thing to do. I
personally feel 845 takes too much drive and that the 211 is in
practice a much better tube.

The 812-a lower-mu 811- or its modern variants is another good
prospect. If you are afraid to run over a grand of B+ none of these
tubes will really perform very well.

Bret Ludwig wrote:

Many people have built 845 SE amps. It is a very popular thing to do. I
personally feel 845 takes too much drive and that the 211 is in
practice a much better tube.

I plan to use a pair of parallel 6BQ5 in triode with a choke to supply the
DC,
I can easily get 200vrms of drive, since the load on the driver stage
is only mainly the bias resistors for the output tubes.
Any 2H distortion should partially cancel that produced in the output
stage.

Input will be a parallel 6CG7, so that its a 3 stage amp, not 4 stage,
and 12dB of NFB would still give 2Vrms input sensitivity.






The 812-a lower-mu 811- or its modern variants is another good
prospect. If you are afraid to run over a grand of B+ none of these
tubes will really perform very well.

I plan a 1,400V supply, with cathode biasing
so Ea = about 1,250V, with two tubes running at about 65 watts each,
and an OPT of 10K : 5 ohms, or a 2,000:1 Z ratio.

The low Ra of the 845 is a very attractive feature, since Ro of the
finished amp
is low without relying on large amounts of loop FB, because
with two output tubes Ra = about 2k, so at the output secondary on the OPT
its about
1 ohm, plus some winding resistance.

Patrick Turner.

Patrick Turner wrote:


I plan a 1,400V supply, with cathode biasing
so Ea = about 1,250V, with two tubes running at about 65 watts each,
and an OPT of 10K : 5 ohms, or a 2,000:1 Z ratio.

Why not fixed bias? You will be wasting a lot of heat with cathode bias.
The cheap 845s in my amp are the cheap chinese ones with "CHINA" spelled
"CHIAN". After a short burn in they are stable and don't drift.


The low Ra of the 845 is a very attractive feature, since Ro of the
finished amp
is low without relying on large amounts of loop FB, because
with two output tubes Ra = about 2k, so at the output secondary on the OPT
its about
1 ohm, plus some winding resistance.

Patrick Turner.

Most of the 845 schematics floating around don't use any feedback other
than un-bypassed cathode resistors.

Are you winding your own SE transformers?


Adam

Adam Stouffer wrote:

Patrick Turner wrote:


I plan a 1,400V supply, with cathode biasing
so Ea = about 1,250V, with two tubes running at about 65 watts each,
and an OPT of 10K : 5 ohms, or a 2,000:1 Z ratio.

Why not fixed bias? You will be wasting a lot of heat with cathode bias.
The cheap 845s in my amp are the cheap chinese ones with "CHINA" spelled
"CHIAN". After a short burn in they are stable and don't drift.

Where did you get your tubes from and how much?


I am thinking of using Shuguang 845M.

The fixed bias is actuallt better during warm up because the grid bias is sent
to
-200V within 3 seconds before the cathodes are hot enough to
emitt so there is little strain on cathodes.

Yeah, they should be stable, and I am considering it,
but 200V at 110mA is only 22 watts of power in am amp
with about 70 watts in each tube, and some drop across the
CLCLC chokes, and OPT.





The low Ra of the 845 is a very attractive feature, since Ro of the
finished amp
is low without relying on large amounts of loop FB, because
with two output tubes Ra = about 2k, so at the output secondary on the OPT
its about
1 ohm, plus some winding resistance.

Patrick Turner.

Most of the 845 schematics floating around don't use any feedback other
than un-bypassed cathode resistors.

I did more careful cals and Ra with 1250V and 55mA = 2.2kohms for one 845,
so for 2 it is only 1.1k, so as RL rated is going to be 10k, the DF is a healthy

9 before any NFB is added.

Well, maybe 8.2, with winding resistance losses just less than 10%.

I may try 8dB global switchable....


Are you winding your own SE transformers?

The project is for a customer. Yes, I will be winding the OPTs

100mm stack of 50 mm tongue, 3,200 P turns of 0.3mm dia wire,
5P x 4S interleaving config; two lots of windings for
no loss matching to 2.8 ohms and 5 ohms which suits most
modern speakers with a lot of their range that is well below the nominal value.

Trannies will be doubly varnished then potted.

Careful insulation is needed.

I had some chassis already made up for "normal" output tubes to make
"normal" tube amps with KT88 etc but these were not bid enough to be adapated
to the 845 project.

So I have to make up all the chassis parts.
2 mm sheet steel all welded and bolted; maybe I get someone to make it
for me, but available guys and waiting time last time I attempted
to employ someone meant a 2mth wait, so I probably will struggle to make it
myself
yet again.

Nobody wants simple straight forward amps with normal tubes........

Horrible work, but its what I have to do.


Patrick Turner.





Adam

Patrick Turner wrote:



Where did you get your tubes from and how much?

http://www.pentalaboratories.com

I think a pair was around $80-90 US.



I am thinking of using Shuguang 845M.

The fixed bias is actuallt better during warm up because the grid bias is sent
to
-200V within 3 seconds before the cathodes are hot enough to
emitt so there is little strain on cathodes.

Yeah, they should be stable, and I am considering it,
but 200V at 110mA is only 22 watts of power in am amp
with about 70 watts in each tube, and some drop across the
CLCLC chokes, and OPT.


This was my first amp using fixed bias so I added 125ma fuses in each
cathode along with a 10 ohm sense resistor. The fuse should blow in case
bias fails. Although maybe its better to have the bias keep a relay
engaged in the B+ line.


The project is for a customer. Yes, I will be winding the OPTs

100mm stack of 50 mm tongue, 3,200 P turns of 0.3mm dia wire,
5P x 4S interleaving config; two lots of windings for
no loss matching to 2.8 ohms and 5 ohms which suits most
modern speakers with a lot of their range that is well below the nominal value.

Trannies will be doubly varnished then potted.

Careful insulation is needed.


Magnet wire has better insulation than you think. At work we tested
32AWG with a HIPOT meter and it didn't arc or leak at 5kv ac/dc. Not to
say it will take that forever...

I had some chassis already made up for "normal" output tubes to make
"normal" tube amps with KT88 etc but these were not bid enough to be adapated
to the 845 project.

So I have to make up all the chassis parts.
2 mm sheet steel all welded and bolted; maybe I get someone to make it
for me, but available guys and waiting time last time I attempted
to employ someone meant a 2mth wait, so I probably will struggle to make it
myself
yet again.

A good machinist can use a mill and cut the large circular holes in no
time. See if someone will do it for a case of beer.


Nobody wants simple straight forward amps with normal tubes........


Whats the fun in that The 845 is a "normal" tube, just not normal for
home hifi use.


Adam

Adam Stouffer wrote:

Patrick Turner wrote:



Where did you get your tubes from and how much?

http://www.pentalaboratories.com

I think a pair was around $80-90 US.

These guys are quoting me usd $100 for a matched pair of 845MP.

The plain 845 is $46 each.





I am thinking of using Shuguang 845M.

The fixed bias is actuallt better during warm up because the grid bias is sent
to
-200V within 3 seconds before the cathodes are hot enough to
emitt so there is little strain on cathodes.

Yeah, they should be stable, and I am considering it,
but 200V at 110mA is only 22 watts of power in am amp
with about 70 watts in each tube, and some drop across the
CLCLC chokes, and OPT.


This was my first amp using fixed bias so I added 125ma fuses in each
cathode along with a 10 ohm sense resistor. The fuse should blow in case
bias fails. Although maybe its better to have the bias keep a relay
engaged in the B+ line.

I will be having two tubes in parallel, and after thought the fixed bias with
separate adjustment for each tube sounds the way to go.

It also makes the circuit easier to cater for 211 use, since the bias change for 211
from -200 to -75 is easier to do.



The project is for a customer. Yes, I will be winding the OPTs

100mm stack of 50 mm tongue, 3,200 P turns of 0.3mm dia wire,
5P x 4S interleaving config; two lots of windings for
no loss matching to 2.8 ohms and 5 ohms which suits most
modern speakers with a lot of their range that is well below the nominal value.

Trannies will be doubly varnished then potted.

Careful insulation is needed.


Magnet wire has better insulation than you think. At work we tested
32AWG with a HIPOT meter and it didn't arc or leak at 5kv ac/dc. Not to
say it will take that forever...

I am aware of the grade 2 magnet wire, a pair of wires twisted together
can take several thousand volts before arcing.

I don't like taking chances though, and trying to get 1.3kV to not leak all over the
joint
over time means good insulation is needed.


I had some chassis already made up for "normal" output tubes to make
"normal" tube amps with KT88 etc but these were not big enough to be adapated
to the 845 project.

So I have to make up all the chassis parts.
2 mm sheet steel all welded and bolted; maybe I get someone to make it
for me, but available guys and waiting time last time I attempted
to employ someone meant a 2mth wait, so I probably will struggle to make it
myself
yet again.

A good machinist can use a mill and cut the large circular holes in no
time. See if someone will do it for a case of beer.

I have already engaged a metal worker willing to make
two chassis 490mm x 230mm x 65mm high, 2mm thick steel sheet for
powder coating, with bottom covers, but all undrilled for about usd $50 each.
This is a real good deal, with only a week's waiting time, so he must have goot
all last year's work done before Xmas.


I will probably buy a bolt operated hole press to deal with most of the holes,
to save trying to drill, since the value of the contract allows me to invest in some
additional tooling up.



Nobody wants simple straight forward amps with normal tubes........


Whats the fun in that The 845 is a "normal" tube, just not normal for
home hifi use.

845 normal? I thought perhaps extraordinary.

BTW, the 211 has Ra about 5 times higher than the 845 at 55 mA , Ea 1,250V.
But µ is 12, so they are easier to drive.

But Ro is a lot higher with DF without NFB about only 4.

I am not worried by the HV, microwave ovens have HV in them, and there is one in
every
home, although real cooks wouldn't use one.....

For best tasting Mozart, cook with 845......

Patrick Turner.




Adam

On Fri, 06 Jan 2006 11:02:04 GMT, Patrick Turner
wrote:

Who has experiewnce with 845 use in any SET amps?
What do ppl think about what is said about the 845 at
http://www.supertnt.com/default/news20.asp ?
Where is the best supplier?

The 845M looks promising.

From what I have read, i wouldn't run 845 at more than 70w pda.

Been reading your posts on this topic, but have been reluctant
to comment because my info is dated. But, for whatever it's
worth: (all comments dated 1994)

When I built my 845 amplifier, (initially thought to be just an
experiment, so made as two channels plus mil-spec power supply
on a single chassis; dumb as rocks move; I can no longer pick
it up or move it; how's this for a run-on sentence, hot shot?)
I went through some of the same issues. YMMV.

Have several RCA's, a couple strange European brands like
Halcro, a batch of United Electronics, some contemporary
Chinese production, etc. They all work; opinions on
offer.

For a single triode per channel and cathode bias I ended
up at the popular operating point of 1000 volts and 80 mA.
Probably wouldn't do much different again today; works.

In your case of multiple valves, I'd very, very definitely
go with three type 211's in parallel. The real kicker, as
you so well know, is the output transformer, which is the
same gig for two type 845's or for three type 211's.

Difference is, real GE and RCA 211/ VT-4-C's are still around;
no worrying about modern imitations/ limitations. (FWIW, I
bought Chinese 845's and 211's for testing purposes in
1994. They were fine at the time; still have 'em; don't
use 'em).

And my preference for serious monitoring are United Electronics
845's. Just opinions, no big thing,

Thanks, as always,

Chris Hornbeck
"Slugs and snails are after me; DDT keeps me happy,
Guess I'm gonna haf'ta tell 'em, that I've got no cerebellum.
Gonna get my PHD, ...
All the girls are in love with me, ..." -Ramones

Chris Hornbeck wrote:

On Fri, 06 Jan 2006 11:02:04 GMT, Patrick Turner
wrote:

Who has experiewnce with 845 use in any SET amps?
What do ppl think about what is said about the 845 at
http://www.supertnt.com/default/news20.asp ?
Where is the best supplier?

The 845M looks promising.

From what I have read, i wouldn't run 845 at more than 70w pda.

Been reading your posts on this topic, but have been reluctant
to comment because my info is dated. But, for whatever it's
worth: (all comments dated 1994)

When I built my 845 amplifier, (initially thought to be just an
experiment, so made as two channels plus mil-spec power supply
on a single chassis; dumb as rocks move; I can no longer pick
it up or move it; how's this for a run-on sentence, hot shot?)
I went through some of the same issues. YMMV.

Have several RCA's, a couple strange European brands like
Halcro, a batch of United Electronics, some contemporary
Chinese production, etc. They all work; opinions on
offer.

Er, Halcro is a famous Australian made amplifier....



For a single triode per channel and cathode bias I ended
up at the popular operating point of 1000 volts and 80 mA.
Probably wouldn't do much different again today; works.

That give about 21 watts max.



In your case of multiple valves, I'd very, very definitely
go with three type 211's in parallel. The real kicker, as
you so well know, is the output transformer, which is the
same gig for two type 845's or for three type 211's.

Slightly more HT at 1,250V and lower Ia allows for 24 watts class A1 per
tube with
845 Pda at 65 watts. RL is about 20k for one tube.

211 has a higher Ra, and the slope of the curve of Ra where
Eg = 0V limits the A1 power more than with 845, hence 211 circuits
have driver arrangements able to drive class A2, ie, work into grid
current.

The OPT for a pair of 845 or 211 can be the same with 10k primary,
and able to swing 700Vrmsd, so I calculated an E&I GOSS
core with 100mm stack. 25mm tongue, 3,200 P turns in 5 sections,
4 S sections, Lp minimum at 64H, leakage will be about 13mH.




Difference is, real GE and RCA 211/ VT-4-C's are still around;
no worrying about modern imitations/ limitations. (FWIW, I
bought Chinese 845's and 211's for testing purposes in
1994. They were fine at the time; still have 'em; don't
use 'em).

And my preference for serious monitoring are United Electronics
845's. Just opinions, no big thing,

I have emailed for a price/availability from
http://www.elexs.com/4AudioA10.htm

NOS 845 are rather expensive are they not?

Patrick Turner.



Thanks, as always,

Chris Hornbeck
"Slugs and snails are after me; DDT keeps me happy,
Guess I'm gonna haf'ta tell 'em, that I've got no cerebellum.
Gonna get my PHD, ...
All the girls are in love with me, ..." -Ramones

"Patrick Turner" a écrit dans le message de news:
...
Hi Pat,
.. . . .
The OPT for a pair of 845 or 211 can be the same with 10k primary,
and able to swing 700Vrmsd, so I calculated an E&I GOSS
core with 100mm stack. 25mm tongue, 3,200 P turns in 5 sections,
4 S sections, Lp minimum at 64H, leakage will be about 13mH.

Duh ? 25mm tongue and 100mm stack ?
Unit conversion mistake or very special bobin ?
What do you think about an ungapped 50mm stack of 42mm tonge with just 2200
P turns ?
Should have an Lp near 200H and with less turns, 4P 3S interleaving looks
good enough for less than 10mH leak L.

Yves ;)

Yves Monmagnon wrote:

"Patrick Turner" a écrit dans le message de news:
...
Hi Pat,
. . . .
The OPT for a pair of 845 or 211 can be the same with 10k primary,
and able to swing 700Vrmsd, so I calculated an E&I GOSS
core with 100mm stack. 25mm tongue, 3,200 P turns in 5 sections,
4 S sections, Lp minimum at 64H, leakage will be about 13mH.

Duh ? 25mm tongue and 100mm stack ?
Unit conversion mistake or very special bobin ?

Big Turner mistake.
Must have been tired.
core size should be 100stack x 50 tongue, not 25 tongue


What do you think about an ungapped 50mm stack of 42mm tonge with just 2200
P turns ?

Way too small a core and too few turns for SE, unless you are using parafeed.
But my website has a PP design for 50mm stack, 44 tongue, 2,320 P turns
which allows for many apps with the sec able to be
arranged to give various no waste matchings.

But for SE, start the design by making ac Bmax = 0.4Tesla at 20 Hz.

So for 50 watts into 10k, signal voltage = 707Vrms.

By educated guess after all this time and experiance, choose Afe = 100 x 50.

Np = 22.6 x V x 10,000 / ( Afe x F x B ),

where 22.6 and 10,000 are constants for all transformers,
B is in Tesla, Afe in sq.mm, F is frequency, V in Vrms.

So for the 100 x 50 core and 0.4T and 20Hz we get
Np = 3,955 turns.

Because we will never use a full 50 watts, we can cheat a bit
and say 3,200 turns is enough, and that we could have
16 P layers with 200 turns, so 0.3mm dia copper wire will do
because its 0.35mm overall including enamel,
and 200 turns will fit the 72 mm traverse width of the bobbin.

3,200 turns of 0.3mm dia wire will have a dcr = 300 ohms on the above core.
P winding losses with ac = 100 x 300 / 10,000 = 3%, which is good for SE.
The S windings can be made to have 2% losses, so ac losses can be 5% total.
But anything up to 7% for SE is reasonable.

The layers can be divided up so that you have
5 sections of P, with layers set out :- 2L, 4L, 4L, 4L, 2L.

The secondaries will be single layers of wire futher divided for wasteless
arrangements to match impedances.

The S will be four sections each of a single layer of wire
placed between the 5 sections of P listed above.

I am guessing that the air gap required will be 0.2mm right across the
joint between abutting E&I lams.
Experimental setting of the gap for correct working at LF is 100% essential.

Perhaps the best core is 50% nickel, like Ongaku,
but I get good sound with good quality GOSS from an Oz supplier
where the µ of the material when in an ungapped and fully interleaved core
used without DC is a max of 17,000.
Eilor also make suitable C-cores,
http://www.eilor.co.il/Articles/Arti...&CategoryID=50
These are easily gapped, and a range of sizes and materials are available.

Using a core with the same 5,000 sq.mm but with 62.5mm tongue and 80 mm stack
but with window size 94mm x 31mm would be even better, but Np would still
have to be 3,200 turns, but the wire dia could be larger, and ac losses
reduced,
but going from 5% losses to 3% losses probably is not a sufficient reward for
the extra weight.



Should have an Lp near 200H and with less turns, 4P 3S interleaving looks
good enough for less than 10mH leak L.

LP can be 200H, but doesn't need to be, and would be impossible to achieve
in a gapped core even as I have suggested above.

The load with Lp shunting RL should not reduce below 0.707 x RL
when the F is 20Hz.
if RL becomes mainly inductive below 20Hz, it is OK,
as long as the L does not saturate, allowing vicious current spikes
when the field collapses.

So ZRL can be equal to RL at 20Hz, so that is 10,000 ohms in my
case for the 2 x 845.

L = 10,000 / ( 20 x 6.28 ) = 80H.

I doubt the 3,200 turns on the large gapped core would give 80H,
maybe only 60, but even so, its probably OK because
nobody is going to use full power at 20 Hz, and since
the Ra of the two 845 is so low at about 1.2k, then the response
will not sag due to Lp when less than full power is used.

I figured leakage will be about 13mH, but still ok, because
the ratio of LL to RL+ Ra is quite favourable.
Since Ra + RL = 11,200 ohms, the -3dB at HF occurrs
when the ZLL = 11,200, so if LL = 0.013H,
F for -3dB = 11,200 / ( 6.28 x 0.013 ) = 137 kHz, but you will never be able
to test it
since the lumped Cshunt will resonate with the LL and cause a series/parallel
set of
resonances.
But all the resonances will be well above the audio band, so they do not
matter,
and can be tamed if the use of a lot of global NFB is desired.

In other PP trannies of the same size as above which I have wound to match
to RL + Ra = 2k, LL needs to be about 2mH.

Its also possible to arrange the windings
so you have five S sections, each a single layer of wire, and 4 P sections,
each of 4 layers.
You still end up with only 8 P&S interfaces, so the LL will be the same.

Alternatively, the primary can be 15 layers aranged in 5 sections of 3 layers
each
with the secondary being 6 sections of one layer each to give
lower S winding losses, and a greater array of possible S matchings without
wasted sections and with even current densities in each and every S turn.
This will reduce LL about 50%, but increase C shunt, and
it would be probably impossible to fit it all into the 25 x 76 window,
and have sufficiently thick insulation between P&S layers,
important because of the 1,250V.

Patrick Turner.









Yves ;)

"Patrick Turner" a écrit dans le message de news:
...


Yves Monmagnon wrote:

"Patrick Turner" a écrit dans le message de
news:
...
Hi Pat,
. . . .
The OPT for a pair of 845 or 211 can be the same with 10k primary,
and able to swing 700Vrmsd, so I calculated an E&I GOSS
core with 100mm stack. 25mm tongue, 3,200 P turns in 5 sections,
4 S sections, Lp minimum at 64H, leakage will be about 13mH.

Duh ? 25mm tongue and 100mm stack ?
Unit conversion mistake or very special bobin ?

Big Turner mistake.
Must have been tired.
core size should be 100stack x 50 tongue, not 25 tongue


What do you think about an ungapped 50mm stack of 42mm tonge with just
2200
P turns ?

Way too small a core and too few turns for SE, unless you are using
parafeed.

Me mistake too ! I was speaking about a PP.
So you plan two paralelled 845 ? Way too small indeed ;)

But my website has a PP design for 50mm stack, 44 tongue, 2,320 P turns
which allows for many apps with the sec able to be
arranged to give various no waste matchings.

But for SE, start the design by making ac Bmax = 0.4Tesla at 20 Hz.

So for 50 watts into 10k, signal voltage = 707Vrms.

By educated guess after all this time and experiance, choose Afe = 100 x
50.

Np = 22.6 x V x 10,000 / ( Afe x F x B ),

where 22.6 and 10,000 are constants for all transformers,
B is in Tesla, Afe in sq.mm, F is frequency, V in Vrms.

So for the 100 x 50 core and 0.4T and 20Hz we get
Np = 3,955 turns.

Because we will never use a full 50 watts, we can cheat a bit
and say 3,200 turns is enough, and that we could have
16 P layers with 200 turns, so 0.3mm dia copper wire will do
because its 0.35mm overall including enamel,
and 200 turns will fit the 72 mm traverse width of the bobbin.

3,200 turns of 0.3mm dia wire will have a dcr = 300 ohms on the above
core.
P winding losses with ac = 100 x 300 / 10,000 = 3%, which is good for SE.
The S windings can be made to have 2% losses, so ac losses can be 5%
total.
But anything up to 7% for SE is reasonable.

The layers can be divided up so that you have
5 sections of P, with layers set out :- 2L, 4L, 4L, 4L, 2L.

The secondaries will be single layers of wire futher divided for wasteless
arrangements to match impedances.

The S will be four sections each of a single layer of wire
placed between the 5 sections of P listed above.

I am guessing that the air gap required will be 0.2mm right across the
joint between abutting E&I lams.
Experimental setting of the gap for correct working at LF is 100%
essential.

Perhaps the best core is 50% nickel, like Ongaku,
but I get good sound with good quality GOSS from an Oz supplier
where the µ of the material when in an ungapped and fully interleaved core
used without DC is a max of 17,000.
Eilor also make suitable C-cores,
http://www.eilor.co.il/Articles/Arti...&CategoryID=50
These are easily gapped, and a range of sizes and materials are available.

Using a core with the same 5,000 sq.mm but with 62.5mm tongue and 80 mm
stack
but with window size 94mm x 31mm would be even better, but Np would still
have to be 3,200 turns, but the wire dia could be larger, and ac losses
reduced,
but going from 5% losses to 3% losses probably is not a sufficient reward
for
the extra weight.



Should have an Lp near 200H and with less turns, 4P 3S interleaving looks
good enough for less than 10mH leak L.

LP can be 200H, but doesn't need to be, and would be impossible to achieve
in a gapped core even as I have suggested above.

The load with Lp shunting RL should not reduce below 0.707 x RL
when the F is 20Hz.
if RL becomes mainly inductive below 20Hz, it is OK,
as long as the L does not saturate, allowing vicious current spikes
when the field collapses.

So ZRL can be equal to RL at 20Hz, so that is 10,000 ohms in my
case for the 2 x 845.

L = 10,000 / ( 20 x 6.28 ) = 80H.

I doubt the 3,200 turns on the large gapped core would give 80H,
maybe only 60, but even so, its probably OK because
nobody is going to use full power at 20 Hz, and since
the Ra of the two 845 is so low at about 1.2k, then the response
will not sag due to Lp when less than full power is used.

I figured leakage will be about 13mH, but still ok, because
the ratio of LL to RL+ Ra is quite favourable.
Since Ra + RL = 11,200 ohms, the -3dB at HF occurrs
when the ZLL = 11,200, so if LL = 0.013H,
F for -3dB = 11,200 / ( 6.28 x 0.013 ) = 137 kHz, but you will never be
able
to test it
since the lumped Cshunt will resonate with the LL and cause a
series/parallel
set of
resonances.
But all the resonances will be well above the audio band, so they do not
matter,
and can be tamed if the use of a lot of global NFB is desired.

In other PP trannies of the same size as above which I have wound to match
to RL + Ra = 2k, LL needs to be about 2mH.

Its also possible to arrange the windings
so you have five S sections, each a single layer of wire, and 4 P
sections,
each of 4 layers.
You still end up with only 8 P&S interfaces, so the LL will be the same.

Alternatively, the primary can be 15 layers aranged in 5 sections of 3
layers
each
with the secondary being 6 sections of one layer each to give
lower S winding losses, and a greater array of possible S matchings
without
wasted sections and with even current densities in each and every S turn.
This will reduce LL about 50%, but increase C shunt, and
it would be probably impossible to fit it all into the 25 x 76 window,
and have sufficiently thick insulation between P&S layers,
important because of the 1,250V.

Patrick Turner.

I've only a limited experience looking to 845 SE, just one tube.
(A chineese one, unknow brand, are there so many ? )
The absolutly minimalist schemo is he
www.dissident-audio.com/SE_845/Amp.gif
I rather concentrated my efforts on the OPT design.

The first one had 2975 turns on primary in 7 equal sections interleaved with
6 secondaries, a total of 13 !
A 0.5mm teflon foil was used between each section
The measured leakage L was a decent 9mH, however, it exhibit a 12µS
irregular rise time:
(Topmost shot in www.dissident-audio.com/SE_845/3xSQ10K.gif )
I bet this is caused by too hi parasitic caps.
Primary L is 40H, impedance ratio is 1000.

The second one has less primary turns (1999 x 0,28mm dia.) and use a much
simpler interleaving scheme, but with an half first and an half last
section, that is 333 + 666 + 666 + 333 turns with 0.1mm insulation at each
layer.
3 secondary sections are interleaved and the thickess of insulation was
rised to 1mm.
The leakage L was a bit less than 8mH, but rise time went down to 3µS.
(Center shot in previus pix)
Primary L is 25H, impedance ratio is 987

The second OPT is better to my ears, more detailled, less agressive and well
balanced.

The bottom shot is from a "professional winder" unit. I'm not ashamed !

I'm not sure yet, but the next trial will probably have a bit more primary
turns, says 2100 to rise the impedance ratio up to 1125 - (while keeping the
same interleaving scheme) in order to reflect a 9000 Ohms load with a
typical 8 Ohms load.
This should reduce THD (already in the 3% range at 15 watts).
It should fit in the copper window by reducing wire dia to 0.25mm, ohmic
losses remaining lower than 5%.
Leakage L should not be affected so much.

Wait and see !

Some amazing photos of this (yet unfinished) amp.
I'm not responsible for mechanical design ;)

www.dissident-audio.com/SE_845/Lateral.jpg
www.dissident-audio.com/SE_845/Rear.jpg
www.dissident-audio.com/SE_845/Protection.jpg

Yves

Yves Monmagnon wrote:

"Patrick Turner" a écrit dans le message de news:
...


Yves Monmagnon wrote:

"Patrick Turner" a écrit dans le message de
news:
...
Hi Pat,
. . . .
The OPT for a pair of 845 or 211 can be the same with 10k primary,
and able to swing 700Vrmsd, so I calculated an E&I GOSS
core with 100mm stack. 25mm tongue, 3,200 P turns in 5 sections,
4 S sections, Lp minimum at 64H, leakage will be about 13mH.

Duh ? 25mm tongue and 100mm stack ?
Unit conversion mistake or very special bobin ?

Big Turner mistake.
Must have been tired.
core size should be 100stack x 50 tongue, not 25 tongue


What do you think about an ungapped 50mm stack of 42mm tonge with just
2200
P turns ?

Way too small a core and too few turns for SE, unless you are using
parafeed.

Me mistake too ! I was speaking about a PP.
So you plan two paralelled 845 ? Way too small indeed ;)

But my website has a PP design for 50mm stack, 44 tongue, 2,320 P turns
which allows for many apps with the sec able to be
arranged to give various no waste matchings.

But for SE, start the design by making ac Bmax = 0.4Tesla at 20 Hz.

So for 50 watts into 10k, signal voltage = 707Vrms.

By educated guess after all this time and experiance, choose Afe = 100 x
50.

Np = 22.6 x V x 10,000 / ( Afe x F x B ),

where 22.6 and 10,000 are constants for all transformers,
B is in Tesla, Afe in sq.mm, F is frequency, V in Vrms.

So for the 100 x 50 core and 0.4T and 20Hz we get
Np = 3,955 turns.

Because we will never use a full 50 watts, we can cheat a bit
and say 3,200 turns is enough, and that we could have
16 P layers with 200 turns, so 0.3mm dia copper wire will do
because its 0.35mm overall including enamel,
and 200 turns will fit the 72 mm traverse width of the bobbin.

3,200 turns of 0.3mm dia wire will have a dcr = 300 ohms on the above
core.
P winding losses with ac = 100 x 300 / 10,000 = 3%, which is good for SE.
The S windings can be made to have 2% losses, so ac losses can be 5%
total.
But anything up to 7% for SE is reasonable.

The layers can be divided up so that you have
5 sections of P, with layers set out :- 2L, 4L, 4L, 4L, 2L.

The secondaries will be single layers of wire futher divided for wasteless
arrangements to match impedances.

The S will be four sections each of a single layer of wire
placed between the 5 sections of P listed above.

I am guessing that the air gap required will be 0.2mm right across the
joint between abutting E&I lams.
Experimental setting of the gap for correct working at LF is 100%
essential.

Perhaps the best core is 50% nickel, like Ongaku,
but I get good sound with good quality GOSS from an Oz supplier
where the µ of the material when in an ungapped and fully interleaved core
used without DC is a max of 17,000.
Eilor also make suitable C-cores,
http://www.eilor.co.il/Articles/Arti...&CategoryID=50
These are easily gapped, and a range of sizes and materials are available.

Using a core with the same 5,000 sq.mm but with 62.5mm tongue and 80 mm
stack
but with window size 94mm x 31mm would be even better, but Np would still
have to be 3,200 turns, but the wire dia could be larger, and ac losses
reduced,
but going from 5% losses to 3% losses probably is not a sufficient reward
for
the extra weight.



Should have an Lp near 200H and with less turns, 4P 3S interleaving looks
good enough for less than 10mH leak L.

LP can be 200H, but doesn't need to be, and would be impossible to achieve
in a gapped core even as I have suggested above.

The load with Lp shunting RL should not reduce below 0.707 x RL
when the F is 20Hz.
if RL becomes mainly inductive below 20Hz, it is OK,
as long as the L does not saturate, allowing vicious current spikes
when the field collapses.

So ZRL can be equal to RL at 20Hz, so that is 10,000 ohms in my
case for the 2 x 845.

L = 10,000 / ( 20 x 6.28 ) = 80H.

I doubt the 3,200 turns on the large gapped core would give 80H,
maybe only 60, but even so, its probably OK because
nobody is going to use full power at 20 Hz, and since
the Ra of the two 845 is so low at about 1.2k, then the response
will not sag due to Lp when less than full power is used.

I figured leakage will be about 13mH, but still ok, because
the ratio of LL to RL+ Ra is quite favourable.
Since Ra + RL = 11,200 ohms, the -3dB at HF occurrs
when the ZLL = 11,200, so if LL = 0.013H,
F for -3dB = 11,200 / ( 6.28 x 0.013 ) = 137 kHz, but you will never be
able
to test it
since the lumped Cshunt will resonate with the LL and cause a
series/parallel
set of
resonances.
But all the resonances will be well above the audio band, so they do not
matter,
and can be tamed if the use of a lot of global NFB is desired.

In other PP trannies of the same size as above which I have wound to match
to RL + Ra = 2k, LL needs to be about 2mH.

Its also possible to arrange the windings
so you have five S sections, each a single layer of wire, and 4 P
sections,
each of 4 layers.
You still end up with only 8 P&S interfaces, so the LL will be the same.

Alternatively, the primary can be 15 layers aranged in 5 sections of 3
layers
each
with the secondary being 6 sections of one layer each to give
lower S winding losses, and a greater array of possible S matchings
without
wasted sections and with even current densities in each and every S turn.
This will reduce LL about 50%, but increase C shunt, and
it would be probably impossible to fit it all into the 25 x 76 window,
and have sufficiently thick insulation between P&S layers,
important because of the 1,250V.

Patrick Turner.

I've only a limited experience looking to 845 SE, just one tube.
(A chineese one, unknow brand, are there so many ? )
The absolutly minimalist schemo is he
www.dissident-audio.com/SE_845/Amp.gif

That design is very minimalist for 845.
The gain of the EL802 driver/input pentode must be very high since the
peak voltage swing would be maybe +/-180V pk.


I prefer to use an EL34 in triode with 6SN7 paralleled input,
and If I wanted more gain a 6SL7 in µ follower will do.






I rather concentrated my efforts on the OPT design.

The first one had 2975 turns on primary in 7 equal sections interleaved with
6 secondaries, a total of 13 !
A 0.5mm teflon foil was used between each section
The measured leakage L was a decent 9mH, however, it exhibit a 12µS
irregular rise time:
(Topmost shot in www.dissident-audio.com/SE_845/3xSQ10K.gif )
I bet this is caused by too hi parasitic caps.
Primary L is 40H, impedance ratio is 1000.

But what size was the core cross section and window?


Getting a decent square wave with any tube amp isn't always easy
because of the combinations of the many amounts of leakage inductances between
the many sectiuons and ditto for the capacitances.
Its impossible to model the behaviour of the LCLCLCLC filter at HF, and usually
some
ring F will be present.
However, if your square wave was at 10kHz at full power its not too bad
since it looks like it will be -3dB at well over 30kHz.

I am surprised that your measured leakage L was 9mH.
I'd have guessed all that interleaving and close coupling
of insulation = 0.5mm only would give less LL.

NFB makes it harder to get a good square wave, especially if the load
is a capacitor.





The second one has less primary turns (1999 x 0,28mm dia.) and use a much
simpler interleaving scheme, but with an half first and an half last
section, that is 333 + 666 + 666 + 333 turns with 0.1mm insulation at each
layer.
3 secondary sections are interleaved and the thickess of insulation was
rised to 1mm.
The leakage L was a bit less than 8mH, but rise time went down to 3µS.
(Center shot in previus pix)
Primary L is 25H, impedance ratio is 987

Yes, but isn't the Lp a bit low relative to the load value of 8k?

The HF looks OK.



The second OPT is better to my ears, more detailled, less agressive and well
balanced.

The bottom shot is from a "professional winder" unit. I'm not ashamed !

Its not unusual to be able to do better than pro winders.

Many ppl try to have thin insulation between P&S sections to help reduce
LL, but C rises too much, so using 1mm of insulation is quite OK if you have
enough sections, so C is kept low.



I'm not sure yet, but the next trial will probably have a bit more primary
turns, says 2100 to rise the impedance ratio up to 1125 - (while keeping the
same interleaving scheme) in order to reflect a 9000 Ohms load with a
typical 8 Ohms load.
This should reduce THD (already in the 3% range at 15 watts).
It should fit in the copper window by reducing wire dia to 0.25mm, ohmic
losses remaining lower than 5%.
Leakage L should not be affected so much.

Any slight increase in Np won't change the thd very much.

The only way to lower thd to a minimum is to settle for
less max po and use say 16k for the load; maybe you make thd 1.5%

But the driver tube is important with thd.

Your pentode may or may not produce much 2H.

Some pentodes have a lot of 2H when RL is small, then as RL is raised in value
2H declines to zero at some RL value then rises again when RL is increased
beyond this RL value for minimum 2H.

But the 2H when RL goes high is the opposite phase of that when RL is low.

The 2H of the pentode driver tube when RL is high adds to ther 2H of the output
tube,
and when RL is low it tends to cancel.

The odd numbered thd products don't get cancelled and just get amplified,
so I would never use a pentode driver to make such a large drive voltage because

you would have maybe several % of driver thd at max po.

A driver triode tube always has very similar and mainly 2H in its thd,
and it will always cancel the 2H in the output stage, with a slight penalty
that some small amount of 4H is produced.

Since the 845 requires such a large drive voltage, it souldn't be too hard to
select
a triode driver such as an EL43 in triode whose 2H increases from zero watts at
the output
in much the same % level as po is increased for most loads.
So it should be possible to set up the EL34 so that it is nicely loaded, ie,
has a load that is a high ohm load to ensure its thd isn't unpleasant,
but which does cancel 2H in the output stage so that maximum cancelation occurs
with a 6 ohm load.

Its possible to get thd at 15 watts down to less than 1%, and at 1 watt
it will be negligible, and you hear tubes behaving themsleves with Motzart.


Wait and see !

Some amazing photos of this (yet unfinished) amp.
I'm not responsible for mechanical design ;)

www.dissident-audio.com/SE_845/Lateral.jpg
www.dissident-audio.com/SE_845/Rear.jpg
www.dissident-audio.com/SE_845/Protection.jpg

What you are doing looks quite impressive.

My approach is much more traditionally styled
with a normal steel chassis with potted trannies and tubes on top and cap under
with all hard wiring.

Your OPT look like it has 44tongue, 50 stack, wasteless E&I GOSS material,
and that explains the inductance figures you have got.

I used about 2,000 turns on 44T x 62S to get about 15H for a load of 1.2K to
suit 4 x 6CA7 to make 35 watts; Ia is 240mA, so gap has to be about 0.28mm
across all the E&I.

But now I am making an amp for 50 watts with 2 x 845 in parallel,
the Ia = 125mA approx, and since the Va = 700Vrms, the core must be huge.

A colleague in Sydney is making an SE amp with 6 x GM70,
and I think his OPT will be quite large for the 120 watts.

Patrick Turner.



Yves

"Patrick Turner" a écrit dans le message de news:
...
The absolutly minimalist schemo is he
www.dissident-audio.com/SE_845/Amp.gif

That design is very minimalist for 845.
The gain of the EL802 driver/input pentode must be very high since the
peak voltage swing would be maybe +/-180V pk.

I prefer to use an EL34 in triode with 6SN7 paralleled input,
and If I wanted more gain a 6SL7 in µ follower will do.


I rather concentrated my efforts on the OPT design.

The first one had 2975 turns on primary in 7 equal sections interleaved
with
6 secondaries, a total of 13 !
A 0.5mm teflon foil was used between each section
The measured leakage L was a decent 9mH, however, it exhibit a 12µS
irregular rise time:
(Topmost shot in www.dissident-audio.com/SE_845/3xSQ10K.gif )
I bet this is caused by too hi parasitic caps.
Primary L is 40H, impedance ratio is 1000.

But what size was the core cross section and window?

Seen below you reached the target closely !



Getting a decent square wave with any tube amp isn't always easy
because of the combinations of the many amounts of leakage inductances
between
the many sectiuons and ditto for the capacitances.
Its impossible to model the behaviour of the LCLCLCLC filter at HF, and
usually
some
ring F will be present.
However, if your square wave was at 10kHz at full power its not too bad
since it looks like it will be -3dB at well over 30kHz.

Exactly ! There is a notch just above.


I am surprised that your measured leakage L was 9mH.
I'd have guessed all that interleaving and close coupling
of insulation = 0.5mm only would give less LL.

My guess is that increasing interleaving above says, 10 sections has very
limited effect.


NFB makes it harder to get a good square wave, especially if the load
is a capacitor.





The second one has less primary turns (1999 x 0,28mm dia.) and use a much
simpler interleaving scheme, but with an half first and an half last
section, that is 333 + 666 + 666 + 333 turns with 0.1mm insulation at
each
layer.
3 secondary sections are interleaved and the thickess of insulation was
rised to 1mm.
The leakage L was a bit less than 8mH, but rise time went down to 3µS.
(Center shot in previus pix)
Primary L is 25H, impedance ratio is 987

Yes, but isn't the Lp a bit low relative to the load value of 8k?

The HF looks OK.



The second OPT is better to my ears, more detailled, less agressive and
well
balanced.

The bottom shot is from a "professional winder" unit. I'm not ashamed !

Its not unusual to be able to do better than pro winders.

Many ppl try to have thin insulation between P&S sections to help reduce
LL, but C rises too much, so using 1mm of insulation is quite OK if you
have
enough sections, so C is kept low.

Curiously, this OPT shows a relatively hi Leak L (around 30mH) and a small
shunt cap (less than 300Pf) !
Low interleaving, large insulation ?
The core has 36mm tongue and probably at least 50mm stack. I can't measure,
it's potted !
Looks a bit too small.



I'm not sure yet, but the next trial will probably have a bit more
primary
turns, says 2100 to rise the impedance ratio up to 1125 - (while keeping
the
same interleaving scheme) in order to reflect a 9000 Ohms load with a
typical 8 Ohms load.
This should reduce THD (already in the 3% range at 15 watts).
It should fit in the copper window by reducing wire dia to 0.25mm, ohmic
losses remaining lower than 5%.
Leakage L should not be affected so much.

Any slight increase in Np won't change the thd very much.

The only way to lower thd to a minimum is to settle for
less max po and use say 16k for the load; maybe you make thd 1.5%

But the driver tube is important with thd.

Your pentode may or may not produce much 2H.

Some pentodes have a lot of 2H when RL is small, then as RL is raised in
value
2H declines to zero at some RL value then rises again when RL is increased
beyond this RL value for minimum 2H.


Yeap, I admit it's even a bit provocative !
Look at this:
www.dissident-audio.com/Yves/EL802LdLine.gif

At first, this penthode is obviously a tetrode with a very sharp knee and
distorstion is almost H2.
I've used a 12K load (blue line) and, with a 500V supply, the voltage swing
is impressive !
Since the cathode is not bypassed, the linearity is not so bad and the
sensitivity (about 1,7V rms) remains sufficient for use with "modern"
sources.
I've obtained some global H2 cancelling by tweaking screen voltage and bias.
Indeed, I was paid for many years to count beans. Old habits are hard to
kill ;)


But the 2H when RL goes high is the opposite phase of that when RL is low.

The 2H of the pentode driver tube when RL is high adds to ther 2H of the
output
tube,
and when RL is low it tends to cancel.

The odd numbered thd products don't get cancelled and just get amplified,
so I would never use a pentode driver to make such a large drive voltage
because

you would have maybe several % of driver thd at max po.

A driver triode tube always has very similar and mainly 2H in its thd,
and it will always cancel the 2H in the output stage, with a slight
penalty
that some small amount of 4H is produced.

Since the 845 requires such a large drive voltage, it souldn't be too hard
to
select
a triode driver such as an EL43 in triode whose 2H increases from zero
watts at
the output
in much the same % level as po is increased for most loads.
So it should be possible to set up the EL34 so that it is nicely loaded,
ie,
has a load that is a high ohm load to ensure its thd isn't unpleasant,
but which does cancel 2H in the output stage so that maximum cancelation
occurs
with a 6 ohm load.

Its possible to get thd at 15 watts down to less than 1%, and at 1 watt
it will be negligible, and you hear tubes behaving themsleves with
Motzart.


Wait and see !

Some amazing photos of this (yet unfinished) amp.
I'm not responsible for mechanical design ;)

www.dissident-audio.com/SE_845/Lateral.jpg
www.dissident-audio.com/SE_845/Rear.jpg
www.dissident-audio.com/SE_845/Protection.jpg

What you are doing looks quite impressive.

My approach is much more traditionally styled
with a normal steel chassis with potted trannies and tubes on top and cap
under
with all hard wiring.

Your OPT look like it has 44tongue, 50 stack, wasteless E&I GOSS material,
and that explains the inductance figures you have got.

Exactly, a 42mm tongue !


I used about 2,000 turns on 44T x 62S to get about 15H for a load of 1.2K
to
suit 4 x 6CA7 to make 35 watts; Ia is 240mA, so gap has to be about 0.28mm
across all the E&I.

But now I am making an amp for 50 watts with 2 x 845 in parallel,
the Ia = 125mA approx, and since the Va = 700Vrms, the core must be huge.

A colleague in Sydney is making an SE amp with 6 x GM70,
and I think his OPT will be quite large for the 120 watts.

Mamma mia ! How much DC current ?

Yves.


Patrick Turner.

Yves Monmagnon wrote:

"Patrick Turner" a écrit dans le message de news:
...
The absolutly minimalist schemo is he
www.dissident-audio.com/SE_845/Amp.gif

That design is very minimalist for 845.
The gain of the EL802 driver/input pentode must be very high since the
peak voltage swing would be maybe +/-180V pk.

I prefer to use an EL34 in triode with 6SN7 paralleled input,
and If I wanted more gain a 6SL7 in µ follower will do.


I rather concentrated my efforts on the OPT design.

The first one had 2975 turns on primary in 7 equal sections interleaved
with
6 secondaries, a total of 13 !
A 0.5mm teflon foil was used between each section
The measured leakage L was a decent 9mH, however, it exhibit a 12µS
irregular rise time:
(Topmost shot in www.dissident-audio.com/SE_845/3xSQ10K.gif )
I bet this is caused by too hi parasitic caps.
Primary L is 40H, impedance ratio is 1000.

But what size was the core cross section and window?

Seen below you reached the target closely !



Getting a decent square wave with any tube amp isn't always easy
because of the combinations of the many amounts of leakage inductances
between
the many sectiuons and ditto for the capacitances.
Its impossible to model the behaviour of the LCLCLCLC filter at HF, and
usually
some
ring F will be present.
However, if your square wave was at 10kHz at full power its not too bad
since it looks like it will be -3dB at well over 30kHz.

Exactly ! There is a notch just above.

Then the C component is too high.

there us an optimum number Ni, of interleavings for any audio tranny.

If Ni is too high, then the response will suffer from the C being too high
even though LL is very low.
And also if Ni is too high, too much bobbin space is used for insulation
so the fill factor for the window becomes poor and the copper losses
too high.

The ideal core shape for SE amps would be to have a window that is larger than
what you get with a wasteless core.

Some of the Eilor C-cores achieve this idea very well.

You can still get enough Lp, but with a larger window/Afe ratio,
you can interleave without the penalty of too much C.

The C becomes series resonant with LL, and hence the notch;
it is a common feature in many OPTs, so reduce C and LL to make this notch
appear well above the audio band.





I am surprised that your measured leakage L was 9mH.
I'd have guessed all that interleaving and close coupling
of insulation = 0.5mm only would give less LL.

My guess is that increasing interleaving above says, 10 sections has very
limited effect.

Here is the formula for calculating LL :-

0.417 x Np squared x TL x [ ( 2 x n x c ) + a ]
LL = ----------------------------------------------------------------
1,000,000,000 x n squared x b

Where LL = leakage inductance, in Henrys,
0.417 is a constant for all equations to work,
Np = primary turns,
TL = average turn length around bobbin,
2 is a constant, since there is an area at each end of a layer where
leakage occurs,
n = number of dielectrics, ie, the junctions between layers of P and S
windings,
c = the dielectric gap, ie, the distance between the copper wire surfaces
in P and S windings,
a = height of the finished winding in the bobbin,
b = the traverse width of the winding across the bobbin.

Distances are all in mm!!!

Notice that as you increase the number of P-S interfaces the LL is reduced at a
square law,
but this is offset by the n, c, and a in the top line.

Also notice that LL varies with the Np squared, so hence to get LL low,
it is best to use a big core and keep Np low.


For an iron cored inductor, with no net DC flow in only one direction, as in an
SE amp,
or with balanced DC, as in the case of a PP amp,

1.26 x Np squared x Afe x µ
Lp = ----------------------------------------
1,000,000,000 x ML

Where Lp = primary inductance in Henrys,
1.26 is a constant,
Np = primary turns,
Afe is the cross sectional area of the core in square mm,
U is the mu of the iron at low voltage levels,
1,000,000,000 is a constant,
ML is the magnetic path length around the core in mm.

Notice that Lp varies with Np squared.

In the case of a PP tranny, µ might be 17,000 for good GOSS in E&I lams,
but because you have an air gap it might be as low as only 250.

I have no formulas for working out Cshiunt because its too difficult,
and not even RDH4 is any help, since its all way to
confusing and incomprehensible what they do have on capacitance
calculations in OPTs

So for C, its better to use conventional methods to
keep C low; use 0.05mm mylar between P-P layers, self capacitance will be kept
low enough this way for all OPTs, and then use the largest P-S insulation
possible for the window, or at least more than 0.5mm.

Avoinding C-shunt is important where RL is a high value.

The load seen by the tubes becomes -3dB down and partially reactive
when ZC = RL, so if RL = 10k, so if Cin looking into the end of the primary
is 600pF, then load will be down 3dB and partially reactive at 27kHz only.

In a large PP amp with the same sized tranny but for 1k RLa-a,
this amount of C shunt wouldn't matter because the -3dB load point would
be at 270kHz.

The actual response due to C shunt will not be down -3dB at
27kHz because the tube Ra is about 2.2 kohms; in fact the response
will be very good with an 845 triode at levels below full power.

Say you amp produces clipping power at between 50Hz and 20kHz,
then as you increase F towards where Cshunt makes the load = 0.7 times the
"mid band" load, the output will become distorted due to the load
becoming too low to achieve undistorted midband po.
The same happens at LF, and because of Lp, and saturation shouldn't be confused
with
the amp not being able generate full po because of the Lp shunting
current.



NFB makes it harder to get a good square wave, especially if the load
is a capacitor.





The second one has less primary turns (1999 x 0,28mm dia.) and use a much
simpler interleaving scheme, but with an half first and an half last
section, that is 333 + 666 + 666 + 333 turns with 0.1mm insulation at
each
layer.
3 secondary sections are interleaved and the thickess of insulation was
rised to 1mm.
The leakage L was a bit less than 8mH, but rise time went down to 3µS.
(Center shot in previus pix)
Primary L is 25H, impedance ratio is 987

Yes, but isn't the Lp a bit low relative to the load value of 8k?

The HF looks OK.



The second OPT is better to my ears, more detailled, less agressive and
well
balanced.

The bottom shot is from a "professional winder" unit. I'm not ashamed !

Its not unusual to be able to do better than pro winders.

Many ppl try to have thin insulation between P&S sections to help reduce
LL, but C rises too much, so using 1mm of insulation is quite OK if you
have
enough sections, so C is kept low.

Curiously, this OPT shows a relatively hi Leak L (around 30mH) and a small
shunt cap (less than 300Pf) !
Low interleaving, large insulation ?

Just low interleaving, and this reduces C; probably it may not be well
insulated...


The core has 36mm tongue and probably at least 50mm stack. I can't measure,
it's potted !
Looks a bit too small.

Core is probably 38T x 52T, a common size for OPTs.





I'm not sure yet, but the next trial will probably have a bit more
primary
turns, says 2100 to rise the impedance ratio up to 1125 - (while keeping
the
same interleaving scheme) in order to reflect a 9000 Ohms load with a
typical 8 Ohms load.
This should reduce THD (already in the 3% range at 15 watts).
It should fit in the copper window by reducing wire dia to 0.25mm, ohmic
losses remaining lower than 5%.
Leakage L should not be affected so much.

Any slight increase in Np won't change the thd very much.

The only way to lower thd to a minimum is to settle for
less max po and use say 16k for the load; maybe you make thd 1.5%

But the driver tube is important with thd.

Your pentode may or may not produce much 2H.

Some pentodes have a lot of 2H when RL is small, then as RL is raised in
value
2H declines to zero at some RL value then rises again when RL is increased
beyond this RL value for minimum 2H.

Yeap, I admit it's even a bit provocative !
Look at this:
www.dissident-audio.com/Yves/EL802LdLine.gif

From what you have with 8k at Q point of 200V, ( green line ),
the thd at anode swing of -175V and + 140?V
is about 5% at least, but you are lucky because it is in cancelling
phase, so if the 845 had 8%, you'd see 3% 2H because of the cancellations.
But the pentode curves show there would be a lot of other crapo harmonics,
so i would never use one like this, and prefer a two stage triode
input /driver amp.

You only need an input of about 0.75V to give 175 anode change, so your pentode
gain is high
(and wild) at 233 times, so hence the sensitivity is OK for your amp.



At first, this penthode is obviously a tetrode with a very sharp knee and
distorstion is almost H2.
I've used a 12K load (blue line) and, with a 500V supply, the voltage swing
is impressive !
Since the cathode is not bypassed, the linearity is not so bad and the
sensitivity (about 1,7V rms) remains sufficient for use with "modern"
sources.

Open loop gain as I said would be about 230,
so with Rk unbypassed, gain would be still high enough, but a
6db gain reduction will only reduce thd of the pentode about 4dB, and
introduce some increased 3H perhaps.


I've obtained some global H2 cancelling by tweaking screen voltage and bias.
Indeed, I was paid for many years to count beans. Old habits are hard to
kill ;)

Yes, counting beans is excellent training for counting electrons.
After counting a few ships full of beans, you become ready
for electronics :-)






But the 2H when RL goes high is the opposite phase of that when RL is low.

The 2H of the pentode driver tube when RL is high adds to ther 2H of the
output
tube,
and when RL is low it tends to cancel.

The odd numbered thd products don't get cancelled and just get amplified,
so I would never use a pentode driver to make such a large drive voltage
because

you would have maybe several % of driver thd at max po.

A driver triode tube always has very similar and mainly 2H in its thd,
and it will always cancel the 2H in the output stage, with a slight
penalty
that some small amount of 4H is produced.

Since the 845 requires such a large drive voltage, it souldn't be too hard
to
select
a triode driver such as an EL43 in triode whose 2H increases from zero
watts at
the output
in much the same % level as po is increased for most loads.
So it should be possible to set up the EL34 so that it is nicely loaded,
ie,
has a load that is a high ohm load to ensure its thd isn't unpleasant,
but which does cancel 2H in the output stage so that maximum cancelation
occurs
with a 6 ohm load.

Its possible to get thd at 15 watts down to less than 1%, and at 1 watt
it will be negligible, and you hear tubes behaving themsleves with
Motzart.


Wait and see !

Some amazing photos of this (yet unfinished) amp.
I'm not responsible for mechanical design ;)

www.dissident-audio.com/SE_845/Lateral.jpg
www.dissident-audio.com/SE_845/Rear.jpg
www.dissident-audio.com/SE_845/Protection.jpg

What you are doing looks quite impressive.

My approach is much more traditionally styled
with a normal steel chassis with potted trannies and tubes on top and cap
under
with all hard wiring.

Your OPT look like it has 44tongue, 50 stack, wasteless E&I GOSS material,
and that explains the inductance figures you have got.

Exactly, a 42mm tongue !


I used about 2,000 turns on 44T x 62S to get about 15H for a load of 1.2K
to
suit 4 x 6CA7 to make 35 watts; Ia is 240mA, so gap has to be about 0.28mm
across all the E&I.

But now I am making an amp for 50 watts with 2 x 845 in parallel,
the Ia = 125mA approx, and since the Va = 700Vrms, the core must be huge.

A colleague in Sydney is making an SE amp with 6 x GM70,
and I think his OPT will be quite large for the 120 watts.

Mamma mia ! How much DC current ?

It must be around 360mA+, since he is using Ea = 1,000V,
and like to run the tubes maybe at least 60W Pda.

The core for the 845 I am attempting has Afe = 5,000 sq.mm;
If I had a core = 44T x 57S then Afe = 2,508 sq.mm, so turns for the
smaller core would have to be approximately 1.4 times those used for the larger
core to achieve the same
Lp, but then the winding resistance will be much higher.

Using the larger 50T material with a larger window allows much better
SE tranformers, and in fact for lower powered SE amps a stack that is less than
the
tongue is better than having S greater than T.

With the larger core I would get twice the L with the same turns for the
smaller 44T tranny, but since I can easily fit 3,200 turns on instead of
the approx 2,200 you are using with 44T material, I also
can get another doubling of Lp.

So Lp for the 50T x 100S can possibly be 4 times that of the 44T x 57S.

The down side of such a large core and all these turns
is that the LL and Csh may be high, but LL
with 5P + 4S is OK with respect to the 10k load, since number of interfaces is
8,
and 1mm insulation should be OK.

I might use polyethylene insulation because it has a lower
dielectric constant than polyester or teflon, so effective capacitance
will be even better.
Is polyethylene ( or otherwise known as lowly polythene ) better sounding?
I think so.

The power dissipated in the OPT primary of 300 ohms dcr with
Ia = 120mA is only 4.3 watts, so a large tranny will barely ever get warm,
and there will be fuses and active protection to
guard against tubes saturating and over heating the OPT.

Patrick Turner.








Yves.


Patrick Turner.

On Fri, 13 Jan 2006 09:31:26 GMT, Patrick Turner
wrote:

Er, Halcro is a famous Australian made amplifier....

Very cool. Do I have an Australian valve here? It
certainly looks unique in my experience. Prewar?
During?

That give about 21 watts max.

If yer generous about what ya call "max". But yeah,
that's what most folks call it. I do too, if cornered.

Slightly more HT at 1,250V and lower Ia allows for 24 watts
class A1 per tube with
845 Pda at 65 watts. RL is about 20k for one tube.

A beautiful operating point.

211 has a higher Ra, and the slope of the curve of Ra where
Eg = 0V limits the A1 power more than with 845, hence 211 circuits
have driver arrangements able to drive class A2, ie, work into grid
current.

Sure, but it's not carved in stone. A1 with three bottles
at 1250 volts is also about 50 watts out, even better
linearity, close-enough-for-hand-grenades source
resistance, and availability of NOS bottles.

NOS 845 are rather expensive are they not?

Pretty much all gone now, I'd guess. But type 211/
VT-4-C are still around. Shoot, I've got several
lifetimes worth. Want some?

Thanks, as always,

Chris Hornbeck

Chris Hornbeck wrote:

On Fri, 13 Jan 2006 09:31:26 GMT, Patrick Turner
wrote:

Er, Halcro is a famous Australian made amplifier....

Very cool. Do I have an Australian valve here? It
certainly looks unique in my experience. Prewar?
During?

Halcro is an exclusive brand of state of the art SS class AB mosfet amp
with 0.0001% thd at 200 watts.

The makers were not even born before WW2.

But Oz factories churned out some nice tubes
from well before ww2 to about 1970 perhaps;
I dunno what exact date they were forced to close.



That give about 21 watts max.

If yer generous about what ya call "max". But yeah,
that's what most folks call it. I do too, if cornered.

Slightly more HT at 1,250V and lower Ia allows for 24 watts
class A1 per tube with
845 Pda at 65 watts. RL is about 20k for one tube.

A beautiful operating point.

211 has a higher Ra, and the slope of the curve of Ra where
Eg = 0V limits the A1 power more than with 845, hence 211 circuits
have driver arrangements able to drive class A2, ie, work into grid
current.

Sure, but it's not carved in stone. A1 with three bottles
at 1250 volts is also about 50 watts out, even better
linearity, close-enough-for-hand-grenades source
resistance, and availability of NOS bottles.

Well yes. But Ra that explodes might offer a 100dB excessively
torrid listening experience.




NOS 845 are rather expensive are they not?

Pretty much all gone now, I'd guess. But type 211/
VT-4-C are still around. Shoot, I've got several
lifetimes worth. Want some?

I will have to see.

I'd rather me build an 845, since that's what i have told my customer he
will get
after he pointed to the state of art other 845 amps around.
but he won't pay the same high price,
and he will get far better engineering.


Patrick Turner.



Thanks, as always,

Chris Hornbeck

On Sun, 15 Jan 2006 05:49:44 GMT, Patrick Turner
wrote:

Very cool. Do I have an Australian valve here? It
certainly looks unique in my experience. Prewar?
During?

Halcro is an exclusive brand of state of the art SS class AB mosfet amp
with 0.0001% thd at 200 watts.

The makers were not even born before WW2.

Ah-hem. On closer inspection (should'a been done
before posting; sorry) it says "Haltron" and
"Made in England". Dunno where I imagined the
Halcro name; a mystery. Sorry.


A1 with three bottles
at 1250 volts is also about 50 watts out, even better
linearity, close-enough-for-hand-grenades source
resistance, and availability of NOS bottles.

Well yes. But Ra that explodes might offer a 100dB excessively
torrid listening experience.

My line of thought is towards your strength. If you can
make your own custom OPT and have whatever loading you want,
you can include a cathode (filament) winding providing
whatever amount of relative cathode loading you want.

Type 845's and type 211's are identical except for grid pitch.
Ignoring, temporarily, G1 current, a pretty compelling
argument could be made for the higher mu varient combined
with "cathode" loading/ feedback. Lower distortion,
especially.

It's an option that probably nobody here but you actually has,
though.



And getting back to G1 current: this is a tough one after
we've built kilojolt-plus power supplies and amplifiers
too heavy to even put back up on the bench. I'll just
suggest that this is purely and simply a mental hurdle,
seen from either direction.

Seen as a drive issue, you can solve it, and trivially.

Seen as a power output issue, just use more bottles!
This is HiFi, right? An extra bottle doesn't make a
single ended triode amplifier built with valves older
than we are any stranger than it already is. And,
makes the OPT compromises slightly more bearable.


I'd rather me build an 845, since that's what i have told my customer he
will get
after he pointed to the state of art other 845 amps around.
but he won't pay the same high price,
and he will get far better engineering.

Doubtless. And my comments are only for the general case of
multiple valves, type 211's or type 845's, where decisions
get difficult at two type 845's vs. three type 211's, A1.

Here in the Northern Hemisphere winter, it's no contest.


Good fortune,

Chris Hornbeck

Chris Hornbeck wrote:

On Sun, 15 Jan 2006 05:49:44 GMT, Patrick Turner
wrote:

Very cool. Do I have an Australian valve here? It
certainly looks unique in my experience. Prewar?
During?

Halcro is an exclusive brand of state of the art SS class AB mosfet amp
with 0.0001% thd at 200 watts.

The makers were not even born before WW2.

Ah-hem. On closer inspection (should'a been done
before posting; sorry) it says "Haltron" and
"Made in England". Dunno where I imagined the
Halcro name; a mystery. Sorry.

Well so am I sorry.
Halcro use SS devices; imagine what they'd achieve with tubes!




A1 with three bottles
at 1250 volts is also about 50 watts out, even better
linearity, close-enough-for-hand-grenades source
resistance, and availability of NOS bottles.

Well yes. But Ra that explodes might offer a 100dB excessively
torrid listening experience.

My line of thought is towards your strength. If you can
make your own custom OPT and have whatever loading you want,
you can include a cathode (filament) winding providing
whatever amount of relative cathode loading you want.

There is absolutely no need for any cathode loading winding, ie, a
cathode feedback winding with 211 or 845; they are already
equipped with a large amount of internal triode NFB,
and there is no need to add more in the output stage.
But a CFB winding is OK with higher gain 805, or 833 perhaps,
except that the filament AC supply must float on the CFB winding....



Type 845's and type 211's are identical except for grid pitch.
Ignoring, temporarily, G1 current, a pretty compelling
argument could be made for the higher mu varient combined
with "cathode" loading/ feedback. Lower distortion,
especially.

the sensitivity of the 845 is abysmal;
it needs about 150Vrms to drive, and 211 about 90Vrms,
and loop FB would be far easier and more effective.



It's an option that probably nobody here but you actually has,
though.

I routinely use CFB windings in nearly all the amps i build
and it does miracles with multi grid tubes such as KT66/88/90
or EL34 etc.

I am not inclined to do an 805/833 amp.



And getting back to G1 current: this is a tough one after
we've built kilojolt-plus power supplies and amplifiers
too heavy to even put back up on the bench. I'll just
suggest that this is purely and simply a mental hurdle,
seen from either direction.

The mono 845 will be movable when i am done.

But yeah, as my customers age the weight of the amps
also increases, but some go bike riding, and they say the hills are steeper
and the winds blow harder.




Seen as a drive issue, you can solve it, and trivially.

Seen as a power output issue, just use more bottles!
This is HiFi, right? An extra bottle doesn't make a
single ended triode amplifier built with valves older
than we are any stranger than it already is. And,
makes the OPT compromises slightly more bearable.

That's why I want to use two 845.

The OPT is only 10k : 5 ohms, not 20k : 5 ohms.
I'd prefer 1.2k : 5 ohms.

But regardless of the impedances and LP and LL and C concerns,
winding the tranny isn't all that difficult; once designed it soon becomes
plain old work.





I'd rather me build an 845, since that's what i have told my customer he
will get
after he pointed to the state of art other 845 amps around.
but he won't pay the same high price,
and he will get far better engineering.

Doubtless. And my comments are only for the general case of
multiple valves, type 211's or type 845's, where decisions
get difficult at two type 845's vs. three type 211's, A1.

Here in the Northern Hemisphere winter, it's no contest.

We get about 82 frosts a year; it used to average 100, but
the weather is warming a bit.

You are right though, tube amp heat isn't wasted in winter.

I'd say most ppl here use about 15Kgs of coal a day,
maybe 8gs in summer, and 25 in winter, lots anyway,
and the tube amps don't make much difference.
( teenage girls seem to use about 20Kgs just having a shower for 2
hours.....)

I run my system for short periods, other uses of energy contribute to the
overall
1KWHr per hr everyday.
Some rich folks in bigger houses consume far more than I do.

Western life is not sustainable as it is of course, so someone will have to
find an answer.
If cheap alternative energy becomes available then the money we spend on
energy will be spent
on other consumerables, and the death of the planet will simply be hastened.

All these do gooders running around telling us to save power and turn unused
lights off
are all wasting their time, because even if we in the west abandoned tube
amps and
cut our way of life in half, ( pigs will fly before that happens ), then
all the have nots in developing countries will raise energy and increase
demand
so much that the environment and weather will be ruined anyway.

Some ppl I know have reacted to the warmer weather by simply spending
$4,000 on an air con unit, so even more power is used.

But one dude is getting a roof mounted evaporation cooler
with ducting, about $1,500, and that should do the trick on real hot days
that now
occur more often.
it should work well in our dry heat summer.

Patrick Turner.







Good fortune,

Chris Hornbeck