Then there is the issue of whether making the OT so perfect that it doesn't sound like a tube amp anymore.
Last edited:
Please put a number (approximate) on that bandwidth?
Based hopefully on a real world amplifier that has been built and tested.
It is an interesting and important claim.
Thanks.
_-_-bear
With bandwidths exceeding 20kHz then any comparison of solutions that use different wire diameters probably has to also include comparing skin depth differences.
Mark, I think ur missing the key point. This is a DIY forum, not a restorers or collectors forum, although there are people who do that here.
The output stage is more good than unique. Compare the bandwidth of the OT on the Marantz 9 with almost any McIntosh. No contest.
I think the idea is not to compare a vintage McIntosh to anything modern, as much as to incorporate the OT design into newer amplifiers. I have heard this done, with transformers wound for the purpose and the amp completely designed unlike the McIntosh input stage. The sound was decidedly excellent and not "soupy" like the stock McIntosh tends to be (for obvious causes).
I think you would do well to divorce ur thinking from what you have heard in stock McIntosh amps and think forward, that is if you care about tube amps at all. There are some solid state amps that I like and run as well.
I prefer to pick the amp depending on the task at hand, I'm not seeking the "universal" amplifier that is excellent on everything. I don't think that exists.
An IROC engine is great, but kinda useless in a garbage truck?
_-_-bear
Why?
Within current limits for the diameter wire, it makes no difference at these frequencies. Skin depth is mostly an RF phenomenon.
_-_-bear
bear: Reliable high speed transistors, FETs, LEDs, polypropylene caps. low ESR electrolytics in small packages, IC opamps, transformer materials, low noise references, high speed low distortion buffers, microcontrollers... the engineering community has been busy these past 50 years.
Elliptron? P.Millets class A with low impedance tubes, anode.grid fb and distortion cancellation between driver and output? If so, then we are in a different leage, class A transformer is much less demanding and for class A you have to giv up efficience. The beauty of cirlotron or macs approach is
that it shines especially with class B. As for the rest of P.Millets approaches they arent exactly new and they have been and can be applied to almost any other outputstage with similar advantages.
Transmissionlinetchniques, as i see it, are limited by the impedance of the universe and require also pretty good matching. I certainly look at a OPT also from a transmissionline point of view and attribute much of a transformers ill behaviour to the fact that the above mentioned conditions cannot be matched. Despite of this, out of curiousitie, i would like to hear more about the winding scheme you mention.
well, there is a japanese guys patent with at one end unconnected tertiary coil wound partly with the secondary claiming it would extend BW. No idea how this could work, propable some damped resonant effect.....
Anyway, the two windings of the mac (if bifilary wound) would have something in the order of 1-2pf/cm coupling capacitance between them wich
would give some ac current sharing at least at the high frequencies where leakage inductance matters. If you want to improve that just put some crosscoupling caps for assistance. If you measure a mac type transformer leakage inductance goes down as frequency goes up.
This all may be a bit off topics and as you said that you prefer easy and costeffective approaches may not be worth the effort. I on the other hand do this as a hobby and like to start where others would not even want to try to go, its the fun of it. May be i just want once more in my life enjoy the thrill and exitement i had at age 18 when i managed to achieve in 6 weeks something 4 fully paid scientists tried over a period of 4 years without sucess (well, it later turned out they did not try to hard, lazy b....s). Amongst other things , i manged to design and build a tubeamp with 500hz-500khz flat with very little resources and even less knowledge. I worked day and nigth on it, literally, it was like a drug, and i am still addicted, lol.
So please forgive me,should i present some day a totally silly, practically zero impedance, zero distortion amp with an insane BW and an out of any proportions OPT. I am determinded to get there just for the fun of it.
Skin and proximity is good to consider at ALL frequencies.
In a freely suspended 1mm copperwire the effect starts to show at 20khz.
In a transformers multilayer winding arrangement the number of layers and the total layer thickness dictates the maximum wire diam that can be used without running into skin-degradation.
I respectfully disagree. My amps with regulated by silicon power supplies and 0.01% of 2'nd only harmonic on 40W power still sound very tubey. But as I stated before, I don't use fancy transformers, I use made by Edcor 10K 100W transformers (with UL taps - sic! 😀 ) and resistive parallel feedback by voltage around output Gu-50 toobs.
"skin degradation"? Or self inductance?
What happens in your opinion when there is a "skin degradation"??
Are you suggesting that a ribbon wound winding would be superior to a round wire then? Although I do not think they make it, a "spline" cross section wire would have more surface area, so would that perform better too?
_-_-
_-_-
Just checking now, you've discarded or sold all your tube amps because they are inferior to the things you just cited?
Btw, I am advocating in this thread the use of all that stuff but WITH the McIntosh style tube output stage... results in a well "skinned cat".
(american idiomatic expression: "many ways to skin a cat")
_-_-bear
however AWA was quite a large concern and did a lot of military, broadcast and commercial equipment for the asian region.
...not to mention authoring RDH.
...not to mention authoring RDH.
by skindegradation i ment the combined degradating effects of skineffect and proximityeffect. (diregarding the very minor inductiv effect they also have)
both represent due to a increase in Rac this is a real powerloss. And offcourse, when you want to decrease Rac at frequencies where skindepth matters "stretching" helps, what you need is to decrease surface current densitie.
"Please put a number (approximate) on that bandwidth?"
Following gorgon53's approach for analyzing the two OT's (see previous page post 140) I have concluded that the Mac and Circlotron are likely quite comparable in bandwidth. (no 4X factor, 2X of that earlier estimate was an outright mistake on my part from not including the secondary in the earlier calculation) Both OT's are probably in the 1x to 2x range better bandwidth than the best conventional OTs if wound with the very best techniques also (not including progressive wound or other advanced wound toroids, which are probably the best, of course you could do Mac or Circlotron on a toroid core as well).
Possibly a little advantage to the Circlo due to the lower winding resistance, but a lengthy detailed analysis would be required for deciding what appears to be a fairly close race. Even then, some difficulty in deciding just what winding techniques would really be of equivalent advantage, due to the 2X turns difference.
-----------------------------
The Elliptron has nothing to do with Pete Millett's amplifier or class A. It's a cross between Mac and Circlotron but using a conventional P-P xfmr and a center tapped inductor. Similar to Norman Crowhurst's twin coupled amplifier, but just one OT. Each tube connects it's plate to the usual end of the P-P primary. B+ on the OT center tap. The cathode connects thru an electrolytic (power supply cap) to the opposite 40% screen grid tap. A center tapped inductor connects to the cathodes with it's center tap connected to ground. Single power supply, but floating power supply caps.
The advantage is in using both of the closely coupled 40% winding sections as well as a 60% section for each tube. A low Z OT is used (better bandwidth), because the tapping arrangement doubles the effective primary impedance. Bandwidth further improves due to each tube using more sections of the primary windings. ( leakage inductances of the extra sections get paralleled and so reduced)
The tube operates with an effective 28% cathode feedback. (so easier to drive than Mac or Circlo) The CFB sections (the two 40% sections) have currents from both tubes thru them just like a Circlotron, thus minimizing crossover cutoff effects in the critical feedback winding.
------------------------------------------
The transmission line OTs are a group of variations based on a conventional RF transmission line xfmr (well, actually it's an unconventional version of the transmission line xfmr) with some mods to adapt it to the audio frequency environment.
Derivation for an N:1 case is as follows:
N transmission lines are used, with magnetic cores strung on them to act as common mode chokes between their ends, lines are aligned in parallel. On the secondary side the transmission lines are parallel connected. On the primary side the transmission lines are series connected.
This gives the basic scheme of its function, but several manipulations are now possible to remove redundancies and optionally add isolation. Calling the bottom line the ground level, it becomes apparent that the amount of common mode choking required along each line varies with the height of the line in the parallel array. The tube plate connects to the top line on the primary side. So anotating the amount of common mode choking by line length, we can remove a triangular section from the array. (draw a diagonal line across and shorten the lines to it, shortest lines on the bottom)
Now if one analyzes the voltages along each line it will become apparent that voltages at the end of a shorter line are recurring in the next higher line at that same length from the secondary. So we can remove another triangular section of line lengths from the array, leaving a "unit" length for each line. (draw a diagonal across the lines leaving a unit length for each line, remove the triangular top part of the array) Then reconnect the lines so that each line continues on from the lower line with equivalent voltages.
What remains is a diagonal staircase pattern of "unit" line lengths. Each of these "unit" line lengths is equivalent to a bifilar common mode choke. And each "unit" line has one wire from each end cross connected together, like a center tapped inductor, except that it is bifilar. The overall array "ladder" connections are such as to parallel connect the top half of each center tapped choke to the bottom half of the next choke above it. And similarly, the bottom half of each center tapped choke parallel connects to the top half of the one below it. Net result is a staircase pattern of boxes with each box having wires horizontally across top and bottom, and inductors vertically on the sides. Bifilar coupled windings are vertically from box to box.
Since the two vertical box windings are parallel connected, they can just be combined into one winding, collapsing the boxes to a simple autoformer schematic. Except that the bifilar couplings must be maintained between each vertically adjacent coil (but no further, ie, cannot be Litz wire). Final result is to adjust each level for the current carrying requirement at that level. Or can be thought of as varying the transmission line impedances from high at the top to low at the bottom.
The final physical configuration can be realized by a pyramid structured cable cross section. Top of the pyramid has a small wire, next level down has two wires side by side (or a flat copper strip), next one down has three wires side by side (or a wider flat copper strip), on down the N levels. Since each strip has to handle the secondary voltage end to end, this pyramid cable gets wound on the xfmr core with enough turns to to form a secondary. The end of one strip connects to the begining of the next strip (just like an autoformer). Secondary output is across the bottom layer.
Obviously, the transmission line impedances cannot be correctly set this way, they will be too low. But if they could, then the pyramid cross section would further taper in cross section along the length so that the width of each layer would match the width of the next layer when connected end to end. This would form a tapered transmission line, which in theory could couple into the many Mhz region.
Two of these autoformer assemblies can be connected together at their low Z ends to form a symmetrical assembly of 2x primary Z with the low Z output at the center. This saves on copper over the same number of levels in the original assymmetric form. This general transmission line version does not support isolation primary to secondary.
-------------------
The second variant supports isolation. But I will let you digest the 1st one and comment before getting into that.
Following gorgon53's approach for analyzing the two OT's (see previous page post 140) I have concluded that the Mac and Circlotron are likely quite comparable in bandwidth. (no 4X factor, 2X of that earlier estimate was an outright mistake on my part from not including the secondary in the earlier calculation) Both OT's are probably in the 1x to 2x range better bandwidth than the best conventional OTs if wound with the very best techniques also (not including progressive wound or other advanced wound toroids, which are probably the best, of course you could do Mac or Circlotron on a toroid core as well).
Possibly a little advantage to the Circlo due to the lower winding resistance, but a lengthy detailed analysis would be required for deciding what appears to be a fairly close race. Even then, some difficulty in deciding just what winding techniques would really be of equivalent advantage, due to the 2X turns difference.
-----------------------------
The Elliptron has nothing to do with Pete Millett's amplifier or class A. It's a cross between Mac and Circlotron but using a conventional P-P xfmr and a center tapped inductor. Similar to Norman Crowhurst's twin coupled amplifier, but just one OT. Each tube connects it's plate to the usual end of the P-P primary. B+ on the OT center tap. The cathode connects thru an electrolytic (power supply cap) to the opposite 40% screen grid tap. A center tapped inductor connects to the cathodes with it's center tap connected to ground. Single power supply, but floating power supply caps.
The advantage is in using both of the closely coupled 40% winding sections as well as a 60% section for each tube. A low Z OT is used (better bandwidth), because the tapping arrangement doubles the effective primary impedance. Bandwidth further improves due to each tube using more sections of the primary windings. ( leakage inductances of the extra sections get paralleled and so reduced)
The tube operates with an effective 28% cathode feedback. (so easier to drive than Mac or Circlo) The CFB sections (the two 40% sections) have currents from both tubes thru them just like a Circlotron, thus minimizing crossover cutoff effects in the critical feedback winding.
------------------------------------------
The transmission line OTs are a group of variations based on a conventional RF transmission line xfmr (well, actually it's an unconventional version of the transmission line xfmr) with some mods to adapt it to the audio frequency environment.
Derivation for an N:1 case is as follows:
N transmission lines are used, with magnetic cores strung on them to act as common mode chokes between their ends, lines are aligned in parallel. On the secondary side the transmission lines are parallel connected. On the primary side the transmission lines are series connected.
This gives the basic scheme of its function, but several manipulations are now possible to remove redundancies and optionally add isolation. Calling the bottom line the ground level, it becomes apparent that the amount of common mode choking required along each line varies with the height of the line in the parallel array. The tube plate connects to the top line on the primary side. So anotating the amount of common mode choking by line length, we can remove a triangular section from the array. (draw a diagonal line across and shorten the lines to it, shortest lines on the bottom)
Now if one analyzes the voltages along each line it will become apparent that voltages at the end of a shorter line are recurring in the next higher line at that same length from the secondary. So we can remove another triangular section of line lengths from the array, leaving a "unit" length for each line. (draw a diagonal across the lines leaving a unit length for each line, remove the triangular top part of the array) Then reconnect the lines so that each line continues on from the lower line with equivalent voltages.
What remains is a diagonal staircase pattern of "unit" line lengths. Each of these "unit" line lengths is equivalent to a bifilar common mode choke. And each "unit" line has one wire from each end cross connected together, like a center tapped inductor, except that it is bifilar. The overall array "ladder" connections are such as to parallel connect the top half of each center tapped choke to the bottom half of the next choke above it. And similarly, the bottom half of each center tapped choke parallel connects to the top half of the one below it. Net result is a staircase pattern of boxes with each box having wires horizontally across top and bottom, and inductors vertically on the sides. Bifilar coupled windings are vertically from box to box.
Since the two vertical box windings are parallel connected, they can just be combined into one winding, collapsing the boxes to a simple autoformer schematic. Except that the bifilar couplings must be maintained between each vertically adjacent coil (but no further, ie, cannot be Litz wire). Final result is to adjust each level for the current carrying requirement at that level. Or can be thought of as varying the transmission line impedances from high at the top to low at the bottom.
The final physical configuration can be realized by a pyramid structured cable cross section. Top of the pyramid has a small wire, next level down has two wires side by side (or a flat copper strip), next one down has three wires side by side (or a wider flat copper strip), on down the N levels. Since each strip has to handle the secondary voltage end to end, this pyramid cable gets wound on the xfmr core with enough turns to to form a secondary. The end of one strip connects to the begining of the next strip (just like an autoformer). Secondary output is across the bottom layer.
Obviously, the transmission line impedances cannot be correctly set this way, they will be too low. But if they could, then the pyramid cross section would further taper in cross section along the length so that the width of each layer would match the width of the next layer when connected end to end. This would form a tapered transmission line, which in theory could couple into the many Mhz region.
Two of these autoformer assemblies can be connected together at their low Z ends to form a symmetrical assembly of 2x primary Z with the low Z output at the center. This saves on copper over the same number of levels in the original assymmetric form. This general transmission line version does not support isolation primary to secondary.
-------------------
The second variant supports isolation. But I will let you digest the 1st one and comment before getting into that.
Last edited:
Bear, in case you dont know what proximity effects are, it is a current distribution-thing dependend on the proximity (closeness) of a other conductor and the direction of the current in this conductor. Its effect is most notable when the currents are in opposit direction to each other. At frequencies where skindepth matters 2 close spaced wires with the same currentdirection will have an additional increase of losses up to approx.1.4 of the skinlosses. If the currentdirections are opposite this additional increase is 1.4 at a distance of roughly 1.6 wirediam. If the wires are closer than that the increase is very steep. At about a distance of 3times wirediam the currentdirection ceases to have any influence and they both increase skinlosses by only about 8%. Proximityeffects go to zero at a distance of roughly 10 times wirediam. Keep in mind that the above is for 2 wires in free space. In a close wound multilayercoil things are much much worse.
Last edited:
tnx smoking amp for clearing up what you ment by elliptron, never thougth it had a special name if you use a chocke. I played with that idea too, only would have put the chocke on the outer legs of the same core, just 1 more silly idea...
"Please put a number (approximate) on that bandwidth?"
no idea what you mean? please can you explain
"Please put a number (approximate) on that bandwidth?"
no idea what you mean? please can you explain
Last edited:
gorgon53 and likely a few others appreciate the significant % increase in resistance that can occur at higher frequencies due to wire and winding arrangements - this issue mostly only worries the minds of smps transformer designers, but if people are discussing differences in winding configurations, and at the same time extrapolating RL&C performance out to 100kHz, then it may be a consideration to include. There is a lot of literature in the switchmode domain on the influence of winding builds and interleaved winding contributions.
- Status
- Not open for further replies.