I'm beginning to miss the smell of burning insulation. I need to build something, and the twin coupled amp is what I have in mind. See:
http://www.audiofaidate.it/articoli/Crowth-twincoupled.pdf
and
http://www.audioxpress.com/magsdirx/ax/addenda/media/1960crowhurst.pdf
I can't find any OPTs even close to what is required for the EL34s, though, so I'm thinking of using the Avel dual primary toroids from Parts Express. However, my experience with these, and knowledge of transformer design in general, is nil.
Obviously, I'd need to series connect the primaries, to get a center tap. But should secondaries be paralleled or series connected for best performance? Here, I'm wondering about capacitance & inductance. Am I right in thinking a higher VA would extend low end? Any help would be appreciated.
http://www.audiofaidate.it/articoli/Crowth-twincoupled.pdf
and
http://www.audioxpress.com/magsdirx/ax/addenda/media/1960crowhurst.pdf
I can't find any OPTs even close to what is required for the EL34s, though, so I'm thinking of using the Avel dual primary toroids from Parts Express. However, my experience with these, and knowledge of transformer design in general, is nil.
Obviously, I'd need to series connect the primaries, to get a center tap. But should secondaries be paralleled or series connected for best performance? Here, I'm wondering about capacitance & inductance. Am I right in thinking a higher VA would extend low end? Any help would be appreciated.
Stack of donuts... I like it! The smaller ones have more inductance - and as far as low end, I'd expect them to saturate at about 300V RMS at 60 Hz, regardless of size. Efficiency will be better with a larger one, of course.
Secondaries will probably end up in parallel to get the ratio you want - but I think coupling will be about the same either way.
Digi-Key has a series of AC power toroids as well - don't recall who makes them.
Secondaries will probably end up in parallel to get the ratio you want - but I think coupling will be about the same either way.
Digi-Key has a series of AC power toroids as well - don't recall who makes them.
Audioxpress published a modern update on the Crowhurst twin coupled amp using 6LU8 tubes, Hammond 125E trannies and other easily available components.
Have a look through the back issue directory. If you can't find it I have a copy at home somewhere.
Have a look through the back issue directory. If you can't find it I have a copy at home somewhere.
Thanks for the info! The Digi-Key trannys are Amveco (made in India); I used them in a build of Steve Bench's matrix amp. Not too successful, one of my first forays into solder slinging, so I can't comment on quality. Pricey little devils, though. I was serious about the "burning insulation" part!! Don't want to put no Tangos in this puppy!!
I subscribed to Audioxpress for only one year. It expired April, 2005. I'm guessing that "Crowhurst's Twin-Coupled Amp Revisited" in September, 2005 is what I want. 6LU8??? OK...did a bit of research, seems a little anemic for 98db of Santana.
CD is on the way. It'll be interesting to see what they did. EL34s are still looking good, though.
I subscribed to Audioxpress for only one year. It expired April, 2005. I'm guessing that "Crowhurst's Twin-Coupled Amp Revisited" in September, 2005 is what I want. 6LU8??? OK...did a bit of research, seems a little anemic for 98db of Santana.

CD is on the way. It'll be interesting to see what they did. EL34s are still looking good, though.
Hate to resurrect an old thread, but I too am interested in this amplifier. I can try to answer some questions, but I also have some of my own....
All the articles I've read show the output transformer secondaries connected in parallel, such that the resultant impedance matches your speakers. If you're trying to match a 4-ohm speaker, you need to parallel the 8-ohm taps.
This was published in AudioXpress August 2004 by John Stewart, which I've read. A follow-up article was published in the same magazine in September 2005, which I haven't read.
Stewart describes this in the first article. The 6LU8 is rated at 14W, but for audio can be pushed harder. He got 37W peak in a PP configuration with OPT's rated at 15W each. Stewart claims that the amp can be built for $200 per channel. At that price, it's tempting to consider paralleling outputs for more power.
Finally, here are my questions:
- I purchased but cannot find the AudioXpress CD for 2005, and therefore have not been able to read the article. Can someone post the main points of Stewart from this September 2005 article?
- Is there anything inherently flawed with the twin-coupled amp?
- I was thinking of using the Hammond 1620 rather than the 125E OPT, as the former is wider bandwidth and costs about $20 more per OPT. The 1620 appears to match the published tube parameters for this application, but I suspect that Stewart chose a different operating point for the 6LU8 and therefore the 1620 impedance won't work well. Can anyone comment?
terrydaniel said:Obviously, I'd need to series connect the primaries, to get a center tap. But should secondaries be paralleled or series connected for best performance?
All the articles I've read show the output transformer secondaries connected in parallel, such that the resultant impedance matches your speakers. If you're trying to match a 4-ohm speaker, you need to parallel the 8-ohm taps.
mach1 said:Audioxpress published a modern update on the Crowhurst twin coupled amp using 6LU8 tubes, Hammond 125E trannies and other easily available components.
This was published in AudioXpress August 2004 by John Stewart, which I've read. A follow-up article was published in the same magazine in September 2005, which I haven't read.
terrydaniel said:.......6LU8??? OK...did a bit of research, seems a little anemic for 98db of Santana.
Stewart describes this in the first article. The 6LU8 is rated at 14W, but for audio can be pushed harder. He got 37W peak in a PP configuration with OPT's rated at 15W each. Stewart claims that the amp can be built for $200 per channel. At that price, it's tempting to consider paralleling outputs for more power.
Finally, here are my questions:
- I purchased but cannot find the AudioXpress CD for 2005, and therefore have not been able to read the article. Can someone post the main points of Stewart from this September 2005 article?
- Is there anything inherently flawed with the twin-coupled amp?
- I was thinking of using the Hammond 1620 rather than the 125E OPT, as the former is wider bandwidth and costs about $20 more per OPT. The 1620 appears to match the published tube parameters for this application, but I suspect that Stewart chose a different operating point for the 6LU8 and therefore the 1620 impedance won't work well. Can anyone comment?
Re: Re: Crowhurst's Twin Coupled Amplifier
and you still have 8 ohm tap .
re-read basics, or you prefer burning tubes for nothing?
weinstro said:..............
All the articles I've read show the output transformer secondaries connected in parallel, such that the resultant impedance matches your speakers. If you're trying to match a 4-ohm speaker, you need to parallel the 8-ohm taps.
...........
and you still have 8 ohm tap .
re-read basics, or you prefer burning tubes for nothing?

In fact, the Crowhurst's Twin Coupled Amplifier is a unity couplet amplifier ala McIntosh [1]. The variation is that Crowhurst is using two output transformers instead of an expensive bifilar wound transformer. As a result there will be lack of magnetic coupling and Crowhurst had to insert a capacitor between each primaries. Else, he is using most of the tricks that McIntosh did, bootstrapped drivers etc, plus that he has applied some inner feedback.
A true unity coupled output stage has some remarkable characteristics and, if proper designed, will measure and sound very good. The secret is the tight coupling between the primaries, as published by Professor Pen-Tung Sah in his paper “Quasi transients in class B audio-frequency push pull amplifiers” back in 1936. [2]
Jan E Veiset
[1] http://www.veiset.net/tech/Mac1949.pdf
[2] http://www.veiset.net/tech/Mac1936.pdf
A true unity coupled output stage has some remarkable characteristics and, if proper designed, will measure and sound very good. The secret is the tight coupling between the primaries, as published by Professor Pen-Tung Sah in his paper “Quasi transients in class B audio-frequency push pull amplifiers” back in 1936. [2]
Jan E Veiset
[1] http://www.veiset.net/tech/Mac1949.pdf
[2] http://www.veiset.net/tech/Mac1936.pdf
Jan,
Those links are VERY interesting documents. I am always amazed how often we are struggling to re-invent wheels that have been invented often more than half a century ago. Makes you humble in this so called 'hi-tech' era.
I am intrigued by that second article by Pen-Tung Sah. It seems a scan of book pages. What book would that be?
Jan Didden
Those links are VERY interesting documents. I am always amazed how often we are struggling to re-invent wheels that have been invented often more than half a century ago. Makes you humble in this so called 'hi-tech' era.
I am intrigued by that second article by Pen-Tung Sah. It seems a scan of book pages. What book would that be?
Jan Didden
janneman said:I am intrigued by that second article by Pen-Tung Sah. It seems a scan of book pages. What book would that be?
Proceedings of the IRE (Institute of Radio Engineers), Nov. 1936, Volume: 24, Issue: 11.
Jan E Veiset
Given what the Amveco transformers cost perhaps you could talk to Jack Elliano over at Electra-Print to see whether or not he could make some unity coupled transformers for you and what they would cost.
He made custom headphone output transformers for me last year for a very reasonable price and which performed extremely well.
I believe dsavitsk has had a similar experience and can corroborate..
He made custom headphone output transformers for me last year for a very reasonable price and which performed extremely well.
I believe dsavitsk has had a similar experience and can corroborate..
I see the 1949 McIntosh article I scanned five years ago has made its way around the web... 🙂 I got it from a friend who was inspired to build his own versions almost 40 years ago. His included a pair with push-pull parallel 6550s and a 6AR6 stereo version for his car... making his own transformers, of course.
I can't find any OPTs even close to what is required for the EL34s
Maybe I'm just dumb but I always have trouble understanding Crowhurst's articles, including these two (both the original one for EL84 and the second one for EL34).

Could somebody who does understand please explain to me what OP transformers would be required for a twin-coupled EL34 amp, to match an 8 ohm speaker? (My guess from what I think it means, based on 3.5k P-P for normal pentode operation, would be two OP transformers, each giving 1.75k P-P at 16 ohm loading, but that sounds wrong somehow).
TIA
For a nominal 3.5K P-P OT, one would want two xfmrs that sum to the 3.5K figure for primaries, and would still use their 8 Ohm secondaries paralleled if 8 Ohm is the target output Z. Ie, total plate to cathode loading Z per tube stays the same.
Paralleling the 8 Ohm secondaries is actually a bit non-ideal, since nominally two xfmrs will allow twice the power, which would require 1.414 times as much voltage and current on the secondaries. This obviously stresses the voltage rating/saturation of the xfmrs with 1.414 x voltage. Connecting the two 4 Ohm secondaries in series will give the correct voltage and current ratings for 2X power, but then part of the secondaries are typically left unused, and the CFB winding will only be sampling part of the total output, which is not ideal either.
A more interesting design (since there is nothing bifilar wound here) would be to use a smaller primary Z xfmr for the cathode windings than for the plate windings (their sum still remaining the same however) This would allow less % cathode feedback, so less grid drive voltage required. 20 % CFB (by turns, not Z) seems to be cited often.
A much more sensible approach than the two xfmrs idea, is to simply ask the xfmr winder to take a standard 40% UL design and separate out each of the two 40% sections as separate windings with a center tap on each. (one of which is insulated sufficiently from the rest for zero volt CFB operation) Then you can build anything. And the xfmr winders should really make it their standard design.
Don
Paralleling the 8 Ohm secondaries is actually a bit non-ideal, since nominally two xfmrs will allow twice the power, which would require 1.414 times as much voltage and current on the secondaries. This obviously stresses the voltage rating/saturation of the xfmrs with 1.414 x voltage. Connecting the two 4 Ohm secondaries in series will give the correct voltage and current ratings for 2X power, but then part of the secondaries are typically left unused, and the CFB winding will only be sampling part of the total output, which is not ideal either.
A more interesting design (since there is nothing bifilar wound here) would be to use a smaller primary Z xfmr for the cathode windings than for the plate windings (their sum still remaining the same however) This would allow less % cathode feedback, so less grid drive voltage required. 20 % CFB (by turns, not Z) seems to be cited often.
A much more sensible approach than the two xfmrs idea, is to simply ask the xfmr winder to take a standard 40% UL design and separate out each of the two 40% sections as separate windings with a center tap on each. (one of which is insulated sufficiently from the rest for zero volt CFB operation) Then you can build anything. And the xfmr winders should really make it their standard design.
Don
OOPs, I forgot the 16 Ohm secondaries case mentioned:
Connecting the 16 Ohm secondaries in parallel will give the correct voltage and current capabilities for 2X power into 8 Ohms.
Primary Z is preserved since the 8 Ohm load can be pictured as two 16 Ohm loads in parallel, one for each 16 Ohm secondary.
So looks like this IS the correct way to do it.
(NOT the 8 Ohms secondaries in parallel I mentioned above. Let's see, 8 Ohm load on the paralleled 8 Ohm secondaries would look like two 16 Ohm loads in parallel, or 16 Ohms on each 8 Ohm secondary. So would double the primary Z seen)
Hmmm..., I guess if one goes to less than the 50% CFB case, by using dissimilar xfmrs, one can not use capacitors to couple the two xfmr primaries for class B service. (ie, no way to capture the inductive kick-back spike of the full winding when one tube turns off)
Don
Connecting the 16 Ohm secondaries in parallel will give the correct voltage and current capabilities for 2X power into 8 Ohms.
Primary Z is preserved since the 8 Ohm load can be pictured as two 16 Ohm loads in parallel, one for each 16 Ohm secondary.
So looks like this IS the correct way to do it.
(NOT the 8 Ohms secondaries in parallel I mentioned above. Let's see, 8 Ohm load on the paralleled 8 Ohm secondaries would look like two 16 Ohm loads in parallel, or 16 Ohms on each 8 Ohm secondary. So would double the primary Z seen)
Hmmm..., I guess if one goes to less than the 50% CFB case, by using dissimilar xfmrs, one can not use capacitors to couple the two xfmr primaries for class B service. (ie, no way to capture the inductive kick-back spike of the full winding when one tube turns off)
Don
smoking-amp said:A much more sensible approach than the two xfmrs idea, is to simply ask the xfmr winder to take a standard 40% UL design and separate out each of the two 40% sections as separate windings with a center tap on each.
Better yet, ask for two pairs of bifilar windings a la' McIntosh. Couple the windings inductively instead of tacking on a big ugly capacitor as a band aid. McIntosh did it the right way. Crowhurst's method is a kludge. Come to think of it, I seem to recall another paper where the author suggested a non-bifilar wound transformer with 'bridging caps.' That was a kludge. Separate transformers is worse than that.
Bifilar windings are not hard. The only potential problem is the DC voltage between them, but it should not be a problem if they are careful not to put any kinks or tight bends in the wire that would compromise the insulation. It's really not much harder than putting down two separate windings with the same number of turns.
p.s. Ray Moth: I agree that Crowhurst's papers are hard to understand. It does not make you stupid.
"Better yet, ask for two pairs of bifilar windings a la' McIntosh. "
Well, there are some other ways too. One can avoid the class B inductive spike problem if both ends of the center tapped primary are driven by class B totem poles (ie, two tubes per end, one always in control) . But then why bother with a center tapped winding, just use one totem pole driver and a single primary. This needs doubled up B+ however.
The other being the usual Circlotron. This however requires two floating B+ supplies. Even worse! Still has the hard to drive 50% CFB as the Mac too. But a couple of double bobbin industrial xfmrs or isolation xfmrs will do for B+, sometimes used with a couple of Ebay derived PFC controllers (Lambda, Vicor or cannibalized Kepco) added for voltage regulation.
These two designs do however modestly improve on the McIntosh in some respects. For the same winding window in the OT, a center tapped primary will have a certain leakage inductance to the secondary for each primary side, and only one is active at a time.
For a single side primary, like the totem pole or circlotron, twice as much wire can be active at all times (since both tubes use the same winding), and so L leakage can be roughly halved. With half the primary turns, distributed capacitance is also roughly halved. (the McIntosh bifilar does roughly halve dist. cap. too) This will improve bandwidth by 2 = SQRT(2*2). Winding resistances can be dropped some too. And no DC breakdown problem between adjacent wires.
Don
Well, there are some other ways too. One can avoid the class B inductive spike problem if both ends of the center tapped primary are driven by class B totem poles (ie, two tubes per end, one always in control) . But then why bother with a center tapped winding, just use one totem pole driver and a single primary. This needs doubled up B+ however.
The other being the usual Circlotron. This however requires two floating B+ supplies. Even worse! Still has the hard to drive 50% CFB as the Mac too. But a couple of double bobbin industrial xfmrs or isolation xfmrs will do for B+, sometimes used with a couple of Ebay derived PFC controllers (Lambda, Vicor or cannibalized Kepco) added for voltage regulation.
These two designs do however modestly improve on the McIntosh in some respects. For the same winding window in the OT, a center tapped primary will have a certain leakage inductance to the secondary for each primary side, and only one is active at a time.
For a single side primary, like the totem pole or circlotron, twice as much wire can be active at all times (since both tubes use the same winding), and so L leakage can be roughly halved. With half the primary turns, distributed capacitance is also roughly halved. (the McIntosh bifilar does roughly halve dist. cap. too) This will improve bandwidth by 2 = SQRT(2*2). Winding resistances can be dropped some too. And no DC breakdown problem between adjacent wires.
Don
Variable CFB
Just occured to me that the totem pole approach allows variable CFB % easily. Can't do that with twin coupled or Mac!
But even better, I think, is to just use feedback to the driver stage by connecting C to D and Cfdbk to A. (and delete CFB pot.) Puts more gain in the loop.
I guess if one leaves the CFB pot in at 50% for either scheme, one gets the best PSRR.
Don
Just occured to me that the totem pole approach allows variable CFB % easily. Can't do that with twin coupled or Mac!
But even better, I think, is to just use feedback to the driver stage by connecting C to D and Cfdbk to A. (and delete CFB pot.) Puts more gain in the loop.
I guess if one leaves the CFB pot in at 50% for either scheme, one gets the best PSRR.
Don
Attachments
A Disturbing Realization about the Mac Output
I was just thinking this morning about whether the variable % CFB (cathode feed-back) scheme shown above for a totem pole output, using bootstrapping, could be applied to the McIntosh circuit (or Twin Coupled circuit).
Well, the answer is a resounding YES, and in fact the Mac driver (and Twin Coupled) uses it to the MAX already. The 50% CFB that is usually associated with the Mac circuit is totally NULLED OUT. There is NO CFB in the Mac circuit or the Twin Coupled.
The bootstrapping from outputs to driver plate resistors causes the output grid voltage drive developed on them to be with respect to the output cathodes, NOT ground. This is the SAME scheme used in the totem pole design above to eliminate CFB on the top tube. All these years the Mac CFB output has been a big HOAX. Shocking!
(Now one CAN restore some % CFB in the outputs if one attenuates the bootstraps some. So CFB enthusiasts can relax, no need to hurriedly put your Mac up on Ebay yet.)
But I have to say, that the simple totem pole output accomplishes the same class B fixup for half the OT (output transformer) cost and 1.4X the OT BW (bandwidth) performance of the Mac and without the floating power supplies of the Circlotron, and can easily be adjusted for some % CFB.
This is the SECOND sacred cow that has ingloriously sunk to the bottom. (I already dispatched UL earlier.) Have a nice day!
Don😀
I was just thinking this morning about whether the variable % CFB (cathode feed-back) scheme shown above for a totem pole output, using bootstrapping, could be applied to the McIntosh circuit (or Twin Coupled circuit).
Well, the answer is a resounding YES, and in fact the Mac driver (and Twin Coupled) uses it to the MAX already. The 50% CFB that is usually associated with the Mac circuit is totally NULLED OUT. There is NO CFB in the Mac circuit or the Twin Coupled.
The bootstrapping from outputs to driver plate resistors causes the output grid voltage drive developed on them to be with respect to the output cathodes, NOT ground. This is the SAME scheme used in the totem pole design above to eliminate CFB on the top tube. All these years the Mac CFB output has been a big HOAX. Shocking!
(Now one CAN restore some % CFB in the outputs if one attenuates the bootstraps some. So CFB enthusiasts can relax, no need to hurriedly put your Mac up on Ebay yet.)
But I have to say, that the simple totem pole output accomplishes the same class B fixup for half the OT (output transformer) cost and 1.4X the OT BW (bandwidth) performance of the Mac and without the floating power supplies of the Circlotron, and can easily be adjusted for some % CFB.
This is the SECOND sacred cow that has ingloriously sunk to the bottom. (I already dispatched UL earlier.) Have a nice day!
Don😀
Re: A Disturbing Realization about the Mac Output
In the Mac circuit you’ll still need ~1/2 B+ voltage swing to drive the output tubes, even with the bootstrapping, so I don’t think so..
Let’s look at the Mac’s driver circuit: a 12BH7 (mu=16.5, rp=~5.3) with a 12k plate resistor. That has a gain of approx mu*Ra/(Ra+rp)=11.5x, with 100% bootstrapping the gain would have been mu or 16.5x. This means that the bootstrapping will give a maximum of (16.5/11.5) or ~3dB positive feedback. And this is far less than the 50% CFB in the output stage.
Just my 0.2dB.
Jan E Veiset
smoking-amp said:in fact the Mac driver (and Twin Coupled) uses it to the MAX already. The 50% CFB that is usually associated with the Mac circuit is totally NULLED OUT. There is NO CFB in the Mac circuit or the Twin Coupled.
In the Mac circuit you’ll still need ~1/2 B+ voltage swing to drive the output tubes, even with the bootstrapping, so I don’t think so..
Let’s look at the Mac’s driver circuit: a 12BH7 (mu=16.5, rp=~5.3) with a 12k plate resistor. That has a gain of approx mu*Ra/(Ra+rp)=11.5x, with 100% bootstrapping the gain would have been mu or 16.5x. This means that the bootstrapping will give a maximum of (16.5/11.5) or ~3dB positive feedback. And this is far less than the 50% CFB in the output stage.
Just my 0.2dB.
Jan E Veiset
- Status
- Not open for further replies.
- Home
- Amplifiers
- Tubes / Valves
- Crowhurst's Twin Coupled Amplifier