In my homebrew Unity-Coupled amp, I just used an 841 with a CCS load and 800V B+ to drive the output stage, no bootstrapping. I used the Plitron transformers which have a higher turns ratio than the Mac transformers so AB2 wasn't required to reach saturation.
re: SY
Indeed, there was another thread where it was claimed the Mac OT eliminated crossover distortion. Clearing that up was not well received.
One could even make the case that the Mac OT simply allows one to make a worse amplifier, since class B is allowed.
------
Instead of driver bootstrapping (with attendant positive Fdbk implications), one could just use the bootstrapping with a floating CCS from that for each driver plate. (the CCS V compliance need only cover the eliminated positive Fdbk distortion)
-
Indeed, there was another thread where it was claimed the Mac OT eliminated crossover distortion. Clearing that up was not well received.
One could even make the case that the Mac OT simply allows one to make a worse amplifier, since class B is allowed.
------
Instead of driver bootstrapping (with attendant positive Fdbk implications), one could just use the bootstrapping with a floating CCS from that for each driver plate. (the CCS V compliance need only cover the eliminated positive Fdbk distortion)
-
Last edited:
I feel kind of dumb, but I just realized that I have a socket extender so I could very easily measure input voltage vs Va-k without flipping my amp again (and risking a back injury). I have a HV differential probe so can easily measure input voltage swing and make a differential measurement of Va-k swing. I know what the result will be, but I could make the measurement anyway.
I am in 100% agreement with this drawing and the measurement numbers that you took....
The original MC75 Outputs were wound on a single C-core using only one bobbin.. No Double C-core and no twin bobbins.. I cant see how your leakage inductance would match up to an original one....what leakage did you measure ??
I remember when you were winding those transformers, I felt bad because you were being misguided by someone who never seen the inside of a real MC output transformer....
Last edited:
The reason why cathode followers have never been employed is the high voltage gain required for high output power. If the output stage of the MAC were a unity gain it would not go very far past its initial introduction either. Not worth a patent.....
Instead the MAC circuit does have a voltage gain and doesn't require special devices for voltage amplification or complicated power supplies. This is far more important than the possibility to run it class B.
To be exactly the same yes. To see just how things work, no! The cathode follower gain won't change too much and the result can be easily corrected, if you want, as it is like a conventional push-pull.
Last edited:
I used a single c-core configuration (in this case two cores stacked to get enough Afe) with twin bobbins.
This configuration is best for push-pull transformers as equal DC resistances in the primary halves is guaranteed. The McIntosh output transformer is a push pull one and is no exception to this rule. There are more ways to wind good unity coupling transformers using the various core geometries.
Leakage induction is not worse than with a single bobbin transformer.
Who would have me misguided? I myself took the shorted real original output transformer apart so I exactly know how it was wound. This version used an EI core by the way.
This configuration is best for push-pull transformers as equal DC resistances in the primary halves is guaranteed. The McIntosh output transformer is a push pull one and is no exception to this rule. There are more ways to wind good unity coupling transformers using the various core geometries.
Leakage induction is not worse than with a single bobbin transformer.
Who would have me misguided? I myself took the shorted real original output transformer apart so I exactly know how it was wound. This version used an EI core by the way.
Last edited by a moderator:
From what some Mac OT reverse engineer-ers have said, the later ones were random wound. (No wonder they potted them, they HAD to hide that.) Even an Edcor OT would outperform those.
The Lockhart article gives some hints on a professional layered wind-up, but does not seem to include sufficient interleaves.
http://www.tubebooks.org/Books/lockhart.pdf
There was another winding scheme around with the bifilar wires flipped over within a layer, every other turn, to minimise distributed capacitance, but that would be disastrous for reliability. Plus increasing leakage L from the bobbin space used up for the crossovers.
The two remaining techniques would be (1) simply placing the bifilar wires in an alternating sequence in the layer as a continuous wind, or (2) to put the bifilar wires on alternating layers. Either should work quite well (some more leakage with the insulation). The latter giving the option for additional DC layer insulation between the bifilars, a major weakness in the design.
Polyimid HV insulation on the wire is the preferred reliability fix for technique (1), and it also increases the leakage L similarly by taking up more winding space. So not much difference between the two techniques.
Then there are split bobbins which allow both bifilar groups to be wound symmetrically. Equal leakage L and equal winding resistance. Some cross interleaves between the bobbin sections would be needed to couple the two bifilar groups as well. This would clearly surpass the original OT.
Not often appreciated, but the two bifilar groups ALSO need to be well coupled to each other. The Mac OT does NOTHING to fix this! The split bobbin with cross interleaves does however.
Its quite clear that one can wind a better OT than Mac did. Sounds like Pieter did.
-
The Lockhart article gives some hints on a professional layered wind-up, but does not seem to include sufficient interleaves.
http://www.tubebooks.org/Books/lockhart.pdf
There was another winding scheme around with the bifilar wires flipped over within a layer, every other turn, to minimise distributed capacitance, but that would be disastrous for reliability. Plus increasing leakage L from the bobbin space used up for the crossovers.
The two remaining techniques would be (1) simply placing the bifilar wires in an alternating sequence in the layer as a continuous wind, or (2) to put the bifilar wires on alternating layers. Either should work quite well (some more leakage with the insulation). The latter giving the option for additional DC layer insulation between the bifilars, a major weakness in the design.
Polyimid HV insulation on the wire is the preferred reliability fix for technique (1), and it also increases the leakage L similarly by taking up more winding space. So not much difference between the two techniques.
Then there are split bobbins which allow both bifilar groups to be wound symmetrically. Equal leakage L and equal winding resistance. Some cross interleaves between the bobbin sections would be needed to couple the two bifilar groups as well. This would clearly surpass the original OT.
Not often appreciated, but the two bifilar groups ALSO need to be well coupled to each other. The Mac OT does NOTHING to fix this! The split bobbin with cross interleaves does however.
Its quite clear that one can wind a better OT than Mac did. Sounds like Pieter did.
-
Last edited:
I don't agree. Gain will either be the voltage output of both tubes divided by input to both tubes (which is how we compute it in conventional output stages) or it will be the voltage input of one half over the output swing of one half. Both will give the same answer.
So in this case it would be (11.7V +11.7V)/15.5V for one tube. 23.4V is the change in Va-k for one tube and 15.5V is the input to one tube.
Alternatively, since the tubes are working into the coils in series, we can take the differential input to the entire stage and divide it by the total of the Va-k swings of the two tubes together (just as we do to compute gain in a conventional output stage). Then the computation becomes (23.4V + 23.4V)/31V for the entire output stage.
We can't draw a load line that makes sense for gain < 1. Imagine drawing a load line from a 450V quiescent operating point that only swings down 200V for an input signal of greater than 200V. What caused clipping? Why can't we drive the tube harder and get more swing?
The answer is of course that the plate of each tube swings more than the voltage at the input of each tube, because it has gain.
But your in slovakia. That will cost more to send it than if I just have Doc do it for 500 provided I unpot it myself..If he unpots it,it's an extra 100 dollars.
OK..my bad... I had you confused with another forum member who was winding one of these...
I spoke with Sidney Corderman about the MC75 OT's and MacIntosh hired him back as a consultant for the Anniversary MC75, they kind of cornered him into making a EI lamination design with bobbin to keep cost down... I don't believe he was 100% on board with the idea...
The C-core and the E-I lamination versions do not have the same behavior with their inductance curves... Also they had to make the new MC75 transformers comply with IEC_60065 ....
With a twin bobbin some sections do not get the same coupling if there in opposite end bobbins..
Last edited:
"With a twin bobbin some sections do not get the same coupling if there in opposite end bobbins.."
Exactly the opposite, the twin or split bobbins with interleave crossovers between bobbins fixes this problem. Mac's OT design is the one that's NOT optimized well.
And if one is going to DIY using E-I lams for cost savings, then LONG E-Is are the way to go. ( they are even lower cost lams than the standard scrapless ones, and with higher % grain orientation in the correct direction, since constant V xfmr lams are a hard sell these days; one manufacturer even sent me a free 60 lb box of them!)
Exactly the opposite, the twin or split bobbins with interleave crossovers between bobbins fixes this problem. Mac's OT design is the one that's NOT optimized well.
And if one is going to DIY using E-I lams for cost savings, then LONG E-Is are the way to go. ( they are even lower cost lams than the standard scrapless ones, and with higher % grain orientation in the correct direction, since constant V xfmr lams are a hard sell these days; one manufacturer even sent me a free 60 lb box of them!)
Last edited:
I am still waiting for you to define the exact nodes to measure to get this (23.4V+23.4V) .. OK so you say these AC voltage are in SERIES, if that is true, then it should be measureable...
So I have diff probes to make this measurement...
Once again...please define the test points on the output stage schematic where I measure this 46.8V AC RMS...
So we have two scenarios:
1. 46.8 / 31 = 1.5 and 16 : 1 output transformer gives 2.6 V at the 9 ohm output terminal;
2. 23.4 / 31 = 0.75 and 8 : 1 output transformer gives the same 2.6 V at the 9 ohm terminal.
That's exactly the point! A KT88 cathode follower in class B achieving the same output power could not be driven using the bootstrapping as in the MAC because the voltage supply and small signal tubes would not be enough. Hence the unity coupling has voltage gain of about 2.
The way to end this drama..is simply prove the 46.8V exist...by simply measuring it.... If the windings are truly in Phase and in SERIES, then this 46.8V can simply be measured with a diff probe...
So where do you propose we measure at ??? What nodes to measure at ??
re: SpreadSpectrum
"We can't draw a load line that makes sense for gain < 1. Imagine drawing a load line from a 450V quiescent operating point that only swings down 200V for an input signal of greater than 200V. What caused clipping? Why can't we drive the tube harder and get more swing?
The answer is of course that the plate of each tube swings more than the voltage at the input of each tube, because it has gain."
I don't see where the gain (using CFB) has anything to do with plotting the plate load curve. Just plot the plate curve as normal (grounded cathode style) for the given primary load Z.
Then add the cathode V swing to the plotted grid swing to get the apparent (or drive signal) grid swing. (its just a ground reference issue)
Distortion, derived from the normal load line plot, can be easily converted to the CFB case by multiplying by:
(plotted grid swing)/(apparent grid swing)
or multiply by:
(plotted grid swing)/(plotted grid swing + cathode winding V swing)
same thing
-
"We can't draw a load line that makes sense for gain < 1. Imagine drawing a load line from a 450V quiescent operating point that only swings down 200V for an input signal of greater than 200V. What caused clipping? Why can't we drive the tube harder and get more swing?
The answer is of course that the plate of each tube swings more than the voltage at the input of each tube, because it has gain."
I don't see where the gain (using CFB) has anything to do with plotting the plate load curve. Just plot the plate curve as normal (grounded cathode style) for the given primary load Z.
Then add the cathode V swing to the plotted grid swing to get the apparent (or drive signal) grid swing. (its just a ground reference issue)
Distortion, derived from the normal load line plot, can be easily converted to the CFB case by multiplying by:
(plotted grid swing)/(apparent grid swing)
or multiply by:
(plotted grid swing)/(plotted grid swing + cathode winding V swing)
same thing
-
Last edited:
- Status
- Not open for further replies.