Smoking-amp, thanks for that description. Explains it all.
Anatoliy too.
Smoking amp, what freq response without FB have you gotten from the xfrmrs you are suggesting??
Mark Allen, if you want the vintage McIntosh amps to sound their best you have to do more than just play them stock. The output stage however can be discussed on its own. I disagree completely about the "sound" of the output stage's topology/design. Most of the stock unit's sonics come from elsewhere.
Imo, the old Marantz gear comes nowhere close to the performance the McIntosh output stage produces. Not even a horse race. (you might get close with some Partridge style OT designs in a classic circuit, might)
My opinions are not those of the management.
_-_-bear
Anatoliy too.
Smoking amp, what freq response without FB have you gotten from the xfrmrs you are suggesting??
Mark Allen, if you want the vintage McIntosh amps to sound their best you have to do more than just play them stock. The output stage however can be discussed on its own. I disagree completely about the "sound" of the output stage's topology/design. Most of the stock unit's sonics come from elsewhere.
Imo, the old Marantz gear comes nowhere close to the performance the McIntosh output stage produces. Not even a horse race. (you might get close with some Partridge style OT designs in a classic circuit, might)
My opinions are not those of the management.
_-_-bear
Well, that makes me all warm and fuzzy inside... LOL! The best way to destroy the value of a vintage piece of audio gear is to modify it and after you do it is no longer really vintage. I've actually had people from MAC tell me at the Amp Clinics they used to hold that the quality of sound was sacrificed for reliability. I have found that to be true. I wold say the output stage is more unique than it is good. I have yet to hear a MAC tube amp that even comes close to a stock pair Model 9's.
Mark
Mark
In this case, yeah, they are. The old Macs were brilliantly good amps, exemplars of smart engineering.
"Smoking amp, what freq response without FB have you gotten from the xfrmrs you are suggesting??"
Depends on the interleaving and primary impedance, just like for any OT. For a similar P-P design, 1/2 the turns means 1/4 the leakage L (and also 1/2 the distributed cap with half the delta V) which means 4X the frequency response. There is no contest, Circlotron blows everything away.
Depends on the interleaving and primary impedance, just like for any OT. For a similar P-P design, 1/2 the turns means 1/4 the leakage L (and also 1/2 the distributed cap with half the delta V) which means 4X the frequency response. There is no contest, Circlotron blows everything away.
Last edited:
With a circlotron you have to take into account the capacitances of the powersupplies (unless you use cascaded tubes and put the PS at the cold midpoint)
Also, in the Mac, because 2 wires in parallell gives a thinner layer and results in less voltage in the layer) the mac-approach will beat the circlotron in respect to BW, Rac losses and total capacitiv current demand from the tubes.
Also, in the Mac, because 2 wires in parallell gives a thinner layer and results in less voltage in the layer) the mac-approach will beat the circlotron in respect to BW, Rac losses and total capacitiv current demand from the tubes.
Last edited:
The capacitance of split bobbin power xfmrs is in the 50 or 60 pF range. The typical capacitance of OTs is in the 1000pF range. No contest here, Circlotron is supreme when you consider that it, at minimum, cuts the OT capacitance by half (and effectively 1/4 or 1/8, see below).
With half the turns the Circlotron has half the winding resistance or even lower if the winding window space taken by the extra turns is used for bigger wire. No contest here, Circlotron is supreme.
With half the turns the Circlotron has at least half the distributed capacitance. Even less (1/4) if the wire is kept to the same size since the wire surface area drops by half. The delta voltage across the winding is half that of a dual primary P-P xfmr as well, giving another 1/2 effective capacitance (down to 1/8). No contest here, Circlotron is supreme.
One can wind the layers any way one wants. Thats just a design choice. The best design splits the primary onto split bobbins for equal coupling to the secondary. Split bobbins will also reduce the voltage differential within the winding bobbin by 1/2 if layed up properly. Mac apparently doesn't use split bobbins. .............
Sorry, no contest whatsoever.
With half the turns the Circlotron has half the winding resistance or even lower if the winding window space taken by the extra turns is used for bigger wire. No contest here, Circlotron is supreme.
With half the turns the Circlotron has at least half the distributed capacitance. Even less (1/4) if the wire is kept to the same size since the wire surface area drops by half. The delta voltage across the winding is half that of a dual primary P-P xfmr as well, giving another 1/2 effective capacitance (down to 1/8). No contest here, Circlotron is supreme.
One can wind the layers any way one wants. Thats just a design choice. The best design splits the primary onto split bobbins for equal coupling to the secondary. Split bobbins will also reduce the voltage differential within the winding bobbin by 1/2 if layed up properly. Mac apparently doesn't use split bobbins. .............
Sorry, no contest whatsoever.
Offcourse you could wind a circlotron transformer with 2 thinner wires in parallell also in wich case the mac and the circlotron transformers would be indentical (exept for the macs hassle with the dc voltage) but in the circlotron you still would have the capacitive load of the powersupplies on your back.
With low voltages and high currents the circlotron with autotransformer givs best results. With high voltage/lowish current tubes, as in the mac, the mac approach is unbeatable.
With low voltages and high currents the circlotron with autotransformer givs best results. With high voltage/lowish current tubes, as in the mac, the mac approach is unbeatable.
You seem to be missing the fact that the Circlotron uses 1/2 the turns of the Mac for an equivalent design. The improvement of bandwidth in OTs as the number of turns is reduced has been long time standard knowledge. The power supply capacitance is nearly trivial with split bobbin power xfmrs.
The Circlotron is 4X better bandwidth than the Mac. Standard knowledge.
The Circlotron is 4X better bandwidth than the Mac. Standard knowledge.
Last edited:
you wont get anything near 50-60pf for the powersupplies unless you want to use some purposedesigned switcher
Stancor TGC175-230 (175 Watt, or Magnetek, or Triad, or Signal equivalents) split bobbin power xfmr, available off the shelf from most any distributor. Measures 64 pF common mode. True, it does take two of them for the dual supplies (but then good for 350 Watt). For something a little smaller, the TGC43-230 (43 Watt each) measures 40 pF common mode.
Granted, if one uses just a standard HV xfmr, with single bobbins, the common mode capacitance will be somewhat comparable (still not interleave multiplied like in an OT) to the amount saved in the OT.
Switcher supplies by the way have horrible common mode capacitance. There is actually a moderate size HV cap in them between the input and output to suppress common mode switching noise. (probably one of the reasons for continuing bad sound reports for switchers in audio equipment, let's all the power line noise right thru) The Lambda PF500-360 PFC units I mentioned for regulated 360V HV are switching units, but the cap there would just be between the DC output side and the AC/DC input side from the split bobbin xfmr. (so not a problem, but they do have to be mounted on insulating posts to minimize capacitance to the chassis.) The split bobbin power xfmr is superb at preventing power line noise coming thru also.
Aside from this mostly inconsequential capacitance/bandwidth detour (all depends on how much effort you want to put into it), the Mac OT is just so much more difficult to make, why bother.
Granted, if one uses just a standard HV xfmr, with single bobbins, the common mode capacitance will be somewhat comparable (still not interleave multiplied like in an OT) to the amount saved in the OT.
Switcher supplies by the way have horrible common mode capacitance. There is actually a moderate size HV cap in them between the input and output to suppress common mode switching noise. (probably one of the reasons for continuing bad sound reports for switchers in audio equipment, let's all the power line noise right thru) The Lambda PF500-360 PFC units I mentioned for regulated 360V HV are switching units, but the cap there would just be between the DC output side and the AC/DC input side from the split bobbin xfmr. (so not a problem, but they do have to be mounted on insulating posts to minimize capacitance to the chassis.) The split bobbin power xfmr is superb at preventing power line noise coming thru also.
Aside from this mostly inconsequential capacitance/bandwidth detour (all depends on how much effort you want to put into it), the Mac OT is just so much more difficult to make, why bother.
Last edited:
I have a bunch of Magnetek N-90MD 250 VA medical isolation xfmrs which I occasionally use with some HV switchers. I just checked the common mode capacitance, and it's only 130 pF. (impressive really, I thought it would be more, these are single bobbin with metal end bells)
I believe Topaz makes ultra high isolation xfmrs. Likely expensive though. The medical grade ones seem more than adequate to me. (got them new in box for around $50 each on Ebay)
You could even put two isolation xfmrs in series if you want to cut that common mode C in half again.
I believe Topaz makes ultra high isolation xfmrs. Likely expensive though. The medical grade ones seem more than adequate to me. (got them new in box for around $50 each on Ebay)
You could even put two isolation xfmrs in series if you want to cut that common mode C in half again.
Last edited:
smoking amp, you miss the fact that the 2 bifilar windings of the mac are for ac IN PARALLELL and reqire EXACTLY the same turns than the circlotron (for instance 1000 turns 2x0.5mm against 1000turns 1x0.7). For ac, they have a LOWER Rac at the highest frequencies but offcourse a higher Rdc than the thicker wire would have. If you want ac current sharing also at lower frequencies you need caps approaching the size of circlotrons powersupply caps and offcourse the same amount of caps-related strays.
Anyway, in the mac, the bifilar windings possible to wind per layer are less (from an ac voltage point of view)so ac voltage between layers is also less. Less windings per layer (for example 2x50wdg/layer against 1x70wdg/layer) . We get less EFFECTIVE capacitance between layers, less stray inductance, less skin current density, in other words higher BW. Rdc losses are offcourse higher but at least we dont have to bother about the powersupply related issues of the circlotron.
Anyway, in the mac, the bifilar windings possible to wind per layer are less (from an ac voltage point of view)so ac voltage between layers is also less. Less windings per layer (for example 2x50wdg/layer against 1x70wdg/layer) . We get less EFFECTIVE capacitance between layers, less stray inductance, less skin current density, in other words higher BW. Rdc losses are offcourse higher but at least we dont have to bother about the powersupply related issues of the circlotron.
Having to put two wires side by side still takes up 2X space in the winding window. This increases leakage inductance (see below) and at least doubles surface capacitance (2X surface area per equiv. turn, and 2 times as many layers). (but then half the V differential per layer compensates as you say)
However, just using a split bobbin for the OT cuts the length of the layers in half too, so the same reduced voltage difference per layer. I'm not seeing a lot of point in arguing over this since it just depends on what level of winding design one goes to. If the Mac is really random wound as someone stated, all capacitance bets are off besides. I wouldn't want an OT made that way at all.
In any case I use an OT technique called the Elliptron that combines the best of the Circlotron with the single supply of the Mac, and it works with standard low primary Z OTs as long as they have screen taps.
If one wants to really get carried away, there is another winding scheme that even blows away the Circlotron. Uses a combination of transmission line techniques and common mode inductors on the same core. But is hard to wind. I'm happy to stick with what gives the best performance for the least effort.
Below:
Just filling up the winding space with another wire (even with the same signal) does not lower leakage L. On the contrary.
To lower leakage inductance requires a thinner winding profile so that the magnetic leakage paths are as long as the core length (so the flux stays in the common core). Filling up the winding window with 2X the wire will make the leakage L worse.
You can always use Litz wire if you want to lower Rac a little. Just another winding enhancement factor.
However, just using a split bobbin for the OT cuts the length of the layers in half too, so the same reduced voltage difference per layer. I'm not seeing a lot of point in arguing over this since it just depends on what level of winding design one goes to. If the Mac is really random wound as someone stated, all capacitance bets are off besides. I wouldn't want an OT made that way at all.
In any case I use an OT technique called the Elliptron that combines the best of the Circlotron with the single supply of the Mac, and it works with standard low primary Z OTs as long as they have screen taps.
If one wants to really get carried away, there is another winding scheme that even blows away the Circlotron. Uses a combination of transmission line techniques and common mode inductors on the same core. But is hard to wind. I'm happy to stick with what gives the best performance for the least effort.
Below:
Just filling up the winding space with another wire (even with the same signal) does not lower leakage L. On the contrary.
To lower leakage inductance requires a thinner winding profile so that the magnetic leakage paths are as long as the core length (so the flux stays in the common core). Filling up the winding window with 2X the wire will make the leakage L worse.
You can always use Litz wire if you want to lower Rac a little. Just another winding enhancement factor.
Last edited:
Smoking amp, tnx for pointing out the probs with SMPS. I played with the idea of making a high frequency oscillator of proper design to achieve good stiff powertransfer, good shielding, low capacitance, strays a.s.o, because a circlotron with autotransformer would allow, amongst other things, a BW till now unseen off. For this i would need a 2x 250Vdc with max 5%dc regulation and calculated the static capacitance of heaters, anodevoltages a.s.o., alltogether turned out to be higher than the calculated effective capacitance of the OPT i designed so i abonded the circlotron. As i would have liked a autotransformer with symmetrically split 6 layers 1 wire in serial connected to 6layer 6wires in paralellell to give 196ohm:4ohm, or 1 wire 6 layers serial connected to 4x1.5layer, 4 wires in paralell to give 200:8ohm. 6+6layers looks about good but the wiresize turned out to introduce allready skin and proximity losses so thinner wire laid side by side would be of advantage. I then thougth why not abond ciclotron alltogether and go for a compromize between mac and circlotron, use the thinner wire double, use double powersupplies and stuff them in at the top of the secondaries to get the best of both worlds, autotransformer advantages coupled with powersupply voltages that are only a fifth or seventh above ground.
smoking amp 1000 turns of 2x05mm or 1000 turns of 1x0.7mm takes up eaxactly the same winding space. And you are wrong about the leakage inductance. Calculate the leakage of 2 paralell 0.5mm wires 100 turns = 100mm, against 140 turns 0.7mm wire. As for the layers, obviously you need more layers with the thinner wires, hence more interleaving and even more lower total leakage inductance with the thinner wires. The static capacitance between layers does not depend on wiresize as long as all factors between layers are otherwise the same. The static capacitance from winding to winding does.With the thinner parallelled wire you will have more layers wich increases static capacitance by a factor of 1.4, true, but at the same time your EFFECTIVE capacitance has decreased by up to a factor of 1.4x1.4 because your voltage between layers has decreased by a factor of 1.4!!! What has increased is the additional needed space do to more layers, but since your EFFECTIVE CAPACITANCE HAS ACTUALLY DECREASED DESPITE THE FACT THAT YOUR NUMBER OF LAYERS HAS INCREASED you can make the space between layers 30% smaller (to achieve the same effective capacitance than with thicker wire approach) and thereby get even smaller leakage inductance. Offcourse isolation between primary and secondary are not necessarely equally to what is used between layers so the above is oversimplified and actual cicumstances and ac voltages have to be considered. I try to use only equal wire in primary and secondary and 1/2layer -full layer-1/2layer stacked arrangements and if possible try to limit multiple stacked same winding layers to 2. This wont be possible at highish impedanzes therefore i always strive for lowish impedaces and ratios in the first place.
As to litzwire, no need for using litzwire in audio, has usually a horribly bad fillfactor, but i used 24x0.2mm magnetwire to some advantage only sadly, i run out of it long time ago.
As to litzwire, no need for using litzwire in audio, has usually a horribly bad fillfactor, but i used 24x0.2mm magnetwire to some advantage only sadly, i run out of it long time ago.
Last edited:
I toyed with the Circlotron also, to the point of accumulating all the parts mentioned. But in the end, I decided that the Elliptron gave most of the same advantages without the need for the two floating supplies.
Main limitation on the Circlotron approach using the split bobbin xfmrs and PFC switching modules as regulators is that you are tied to just 360V-370V B+. Otherwise custom or multiple xfmrs and regulators are required. Or some medical grade isolation xfmrs and two conventional power supplies. Getting a little bulky by then. Elliptron just worked out to be easier and more flexible.
For Elliptron, I just use a cheap Edcor 1.7K or 2.2K or 2.5K Ohm P-P OT. Tubes operate between the plate terminals (plates) and the opposite screen tap (cathode through a big cap). The DC for the cathodes comes from some surplus Variac cores I found that were unfinished (just had the winding on them, but no processing for the variable wiper.) The Variac core gets a center tap to ground, and the ends go to the cathodes to supply DC. Center tap on the OT goes to B+. ( I add in a surplus sourced dual bobbin 240 V:240V C core to the Variac core to get enough voltage capability, the combo has to handle the AC voltage between the screen taps) The usual 40% screen taps work out to 28% CFB windings this way, so the tube driver stage is simplified too. The C core and toroid core draw miniscule inductive current since they are undervoltaged and grain aligned.
The Edcors (and Hammonds that I've checked) have good coupling between the secondary and the center 40% windings. The outer 60% winding on one side is coupled fairly well, and the other 60% outer winding is poorly coupled. This is OK with the Elliptron configuration since the cathodes are using the center 40% windings for the CFB (each tube uses both 40% windings), and the plates are using the outer 60% (+ opposite 40%). Being a current source from the plate, the poor leakage L coupling does not matter much out there.
Using both 40% windings for each tube gives a big improvement in coupling performance (BW) over a conventional P-P OT setup. The overlap in conduction of both tubes thru the central 40% CFB windings gives the cross coupling advantages of either the Circlotron or Mac. And while I wouldn't say it equals the Mac or Circlotron, it's maybe getting 80% of the performance for 20% of the cost and effort.
--------------
With twice as many turns you need twice as many layers and there is twice the voltage across the total winding too. That makes the effective capacitance and leakage L go up. There is simply twice as much dielectric and air core in the winding with unwanted E and B fields in it.
Or if considering the two bifilar wires as the same as the single wire in the Circlo, the expansion of the winding to twice volume increases the leakage L, and the doubling of wire surface area doubles the capacitance. (if one doubled the size of the Circlo wire to fill the same volume, then one could argue that the leakage L would be the same, but with the original same size wire, the winding profile is half as thick, which is well known to reduce leakage L)
I guess that should be 2/3 as thick, not 1/2, considering the secondary is still a factor.
Main limitation on the Circlotron approach using the split bobbin xfmrs and PFC switching modules as regulators is that you are tied to just 360V-370V B+. Otherwise custom or multiple xfmrs and regulators are required. Or some medical grade isolation xfmrs and two conventional power supplies. Getting a little bulky by then. Elliptron just worked out to be easier and more flexible.
For Elliptron, I just use a cheap Edcor 1.7K or 2.2K or 2.5K Ohm P-P OT. Tubes operate between the plate terminals (plates) and the opposite screen tap (cathode through a big cap). The DC for the cathodes comes from some surplus Variac cores I found that were unfinished (just had the winding on them, but no processing for the variable wiper.) The Variac core gets a center tap to ground, and the ends go to the cathodes to supply DC. Center tap on the OT goes to B+. ( I add in a surplus sourced dual bobbin 240 V:240V C core to the Variac core to get enough voltage capability, the combo has to handle the AC voltage between the screen taps) The usual 40% screen taps work out to 28% CFB windings this way, so the tube driver stage is simplified too. The C core and toroid core draw miniscule inductive current since they are undervoltaged and grain aligned.
The Edcors (and Hammonds that I've checked) have good coupling between the secondary and the center 40% windings. The outer 60% winding on one side is coupled fairly well, and the other 60% outer winding is poorly coupled. This is OK with the Elliptron configuration since the cathodes are using the center 40% windings for the CFB (each tube uses both 40% windings), and the plates are using the outer 60% (+ opposite 40%). Being a current source from the plate, the poor leakage L coupling does not matter much out there.
Using both 40% windings for each tube gives a big improvement in coupling performance (BW) over a conventional P-P OT setup. The overlap in conduction of both tubes thru the central 40% CFB windings gives the cross coupling advantages of either the Circlotron or Mac. And while I wouldn't say it equals the Mac or Circlotron, it's maybe getting 80% of the performance for 20% of the cost and effort.
--------------
With twice as many turns you need twice as many layers and there is twice the voltage across the total winding too. That makes the effective capacitance and leakage L go up. There is simply twice as much dielectric and air core in the winding with unwanted E and B fields in it.
Or if considering the two bifilar wires as the same as the single wire in the Circlo, the expansion of the winding to twice volume increases the leakage L, and the doubling of wire surface area doubles the capacitance. (if one doubled the size of the Circlo wire to fill the same volume, then one could argue that the leakage L would be the same, but with the original same size wire, the winding profile is half as thick, which is well known to reduce leakage L)
I guess that should be 2/3 as thick, not 1/2, considering the secondary is still a factor.
Last edited:
smoking amp, since after all you are the first person i talked here to, that understands transformer related stuff in depth, i would like to get back to you later, at the moment i am just to tired. I realize we think much alike and different views need more explanation but are worth discussing further.
Getting sleepy here too.
I suspect there are just too many form factors to take into account, winding details like split bobbin etc, as well as some SQRT(2) instead of 2X hand waving factors etc to come to an accurate comparison.
Either a very lengthy analysis or a comparative build would be required. I'm not too confident in the usual winding rules to be able to account for the fine detail of some differences. Nor what would be accurately comparable as far as a build, like wire insulation thickness or layer separation thickness when different voltage differentials are present and different resistances are present.
I just see the Mac OT as too much trouble to bother with for maybe a 2X bandwidth improvement over a conventional OT. Circlotron is either 2X or 4X improvement over conventional I think. You can get that much just by using paralleled tubes. If I were going to go to that much trouble to wind a special OT I would make the transmission/common mode laddered OT, more like 1000X improvement (100 MHz BW). Or a partial variant that can be done with wire instead of coax.
I suspect there are just too many form factors to take into account, winding details like split bobbin etc, as well as some SQRT(2) instead of 2X hand waving factors etc to come to an accurate comparison.
Either a very lengthy analysis or a comparative build would be required. I'm not too confident in the usual winding rules to be able to account for the fine detail of some differences. Nor what would be accurately comparable as far as a build, like wire insulation thickness or layer separation thickness when different voltage differentials are present and different resistances are present.
I just see the Mac OT as too much trouble to bother with for maybe a 2X bandwidth improvement over a conventional OT. Circlotron is either 2X or 4X improvement over conventional I think. You can get that much just by using paralleled tubes. If I were going to go to that much trouble to wind a special OT I would make the transmission/common mode laddered OT, more like 1000X improvement (100 MHz BW). Or a partial variant that can be done with wire instead of coax.
Last edited:
Still a little groggy this morning from being up so late posting last night.
I should comment that it is possible to build switching supplies without the common mode suppression cap if they use a balanced full bridge switch arrangement and use a controlled slewrate (V & I) controller like the Linear Tech parts. Plus some common mode glitch chokes. Was just saying that the typical off the shelf commercial switcher unit would not be suitable for a low noise floating supply. Possibly could use a top quality commercial class D amp module as a floating power supply.
Some of the confusion of comparing the Mac with the Circlo is from just what constitutes comparable windups. I guess the easiest starting point would be to consider the two bifilar wires to be the same wire as in the single Circlo wire, as Gorgon53 mentioned. Then we have near identical physical windups. The dual wire will have a little more capacitance due to 2X surface area versus Sqrt(2) = 1.4 for the single wire, but the adjacent wires being in contact will reduce the 2X factor some. A similar factor (<2 versus Sqrt(2)) will show up in the voltage differential between layers but will also increase the # of layers slightly. So maybe equivalent overall effects.
If the HV between the dual wires is taken into account, then heavier enamel insulation will be required. Which would increase leakage L but decrease distributed capacitance. Maybe an equivalent LC product there? However that is pushing the winding space equivalence off equality. Could maybe balance that against the common mode capacitance in the power supplies. These would need some detailed calculations to nail the numbers down.
Possibly not any dramatic difference between the two winding schemes in that form. But the Circlo has 1/2 the winding resistance in that form since the two Mac wires get series connected. So now the main issue is how to translate winding resistance tradeoff to BW tradeoff. If one shrinks the Circlo wire by .707 diameter to get equal resistance, the windup window shrinks some, but not the full .707:1 due to the secondary winding and insulation factors. This will lower leakage L some, but not dramatically. BW coming from Sqrt (L C) makes it even smaller. On the other hand, if we were to keep the wire diameter the same but use a 2X longer winding window to get the same resistance, leakage would probably drop by 2X. But that ignores the secondary doubling in Ohms, so maybe a 1.4X longer winding.
So I'll clearly have to retract my comment about 4X better bandwidth for the Circlo based on the simple minded 2X turns squared formula (which wasn't taking into account the same secondary turns either). I still think there will be some modest advantage for the Circlo, but it's too small to estimate without detailed calculations.
Looking at actual specs, I think the Mac gets typically around 90 to 100 KHz bandwidth? (correct me if I'm wrong) A top spec conventional OT will get between 50 and 100KHz. (Edcor or Hammond at 25KHz-30KHz are not one of them) ( some of the Plitron toroidals are even higher than 100 KHz) One should be able to do an apples to apples comparison of a Circlotron to a conventional P-P OT I think, using the simple turns squared rule. However one needs to include the unchanged secondary in the calc. So the primary side goes up 4X L and 4X C, but the secondary side stays the same. Transforming to the same side and averaging both, gives 2x L and 2X C, with BW = Sqrt (L C) giving just 2X.
So looks to me that the Cyclo starts out equivalent to the Mac and could possibly exceed it by 2X if all stops pulled winding technique were used. But then it doesn't sound like the Macs were wound with all stops pulled winding technique either. So I would have to call it a horse race.
After all these BW scribblings, I would also comment that just using more local feedback from the plates and lesser feedback thru the OT gets one to almost the same place, for the cost of a couple of resistors. Probably more important OT factors (as far as sound goes) are the P-P balance in the OT from split bobbin configuration and better permeability from correctly grain oriented steel (C core, toroid, maybe permalloy striped EI).
I should comment that it is possible to build switching supplies without the common mode suppression cap if they use a balanced full bridge switch arrangement and use a controlled slewrate (V & I) controller like the Linear Tech parts. Plus some common mode glitch chokes. Was just saying that the typical off the shelf commercial switcher unit would not be suitable for a low noise floating supply. Possibly could use a top quality commercial class D amp module as a floating power supply.
Some of the confusion of comparing the Mac with the Circlo is from just what constitutes comparable windups. I guess the easiest starting point would be to consider the two bifilar wires to be the same wire as in the single Circlo wire, as Gorgon53 mentioned. Then we have near identical physical windups. The dual wire will have a little more capacitance due to 2X surface area versus Sqrt(2) = 1.4 for the single wire, but the adjacent wires being in contact will reduce the 2X factor some. A similar factor (<2 versus Sqrt(2)) will show up in the voltage differential between layers but will also increase the # of layers slightly. So maybe equivalent overall effects.
If the HV between the dual wires is taken into account, then heavier enamel insulation will be required. Which would increase leakage L but decrease distributed capacitance. Maybe an equivalent LC product there? However that is pushing the winding space equivalence off equality. Could maybe balance that against the common mode capacitance in the power supplies. These would need some detailed calculations to nail the numbers down.
Possibly not any dramatic difference between the two winding schemes in that form. But the Circlo has 1/2 the winding resistance in that form since the two Mac wires get series connected. So now the main issue is how to translate winding resistance tradeoff to BW tradeoff. If one shrinks the Circlo wire by .707 diameter to get equal resistance, the windup window shrinks some, but not the full .707:1 due to the secondary winding and insulation factors. This will lower leakage L some, but not dramatically. BW coming from Sqrt (L C) makes it even smaller. On the other hand, if we were to keep the wire diameter the same but use a 2X longer winding window to get the same resistance, leakage would probably drop by 2X. But that ignores the secondary doubling in Ohms, so maybe a 1.4X longer winding.
So I'll clearly have to retract my comment about 4X better bandwidth for the Circlo based on the simple minded 2X turns squared formula (which wasn't taking into account the same secondary turns either). I still think there will be some modest advantage for the Circlo, but it's too small to estimate without detailed calculations.
Looking at actual specs, I think the Mac gets typically around 90 to 100 KHz bandwidth? (correct me if I'm wrong) A top spec conventional OT will get between 50 and 100KHz. (Edcor or Hammond at 25KHz-30KHz are not one of them) ( some of the Plitron toroidals are even higher than 100 KHz) One should be able to do an apples to apples comparison of a Circlotron to a conventional P-P OT I think, using the simple turns squared rule. However one needs to include the unchanged secondary in the calc. So the primary side goes up 4X L and 4X C, but the secondary side stays the same. Transforming to the same side and averaging both, gives 2x L and 2X C, with BW = Sqrt (L C) giving just 2X.
So looks to me that the Cyclo starts out equivalent to the Mac and could possibly exceed it by 2X if all stops pulled winding technique were used. But then it doesn't sound like the Macs were wound with all stops pulled winding technique either. So I would have to call it a horse race.
After all these BW scribblings, I would also comment that just using more local feedback from the plates and lesser feedback thru the OT gets one to almost the same place, for the cost of a couple of resistors. Probably more important OT factors (as far as sound goes) are the P-P balance in the OT from split bobbin configuration and better permeability from correctly grain oriented steel (C core, toroid, maybe permalloy striped EI).
Last edited:
- Status
- Not open for further replies.