Intact Audio

(split this off from an unrelated thread since it is gold. I queried JC about the mac circuit and it lead to this.....dave)

the Mac trannie is only weird because the screen winding? am i right? it's not weird from an operating point... it's a nested feedback scheme using a coupling transformer. 100% feedback. and saves the class B operation from crapping out during crossover. a mix of positive current and negative voltage feedback is also possible through the cathode connections. that can also help with saving a pp stage from "offness". it turns cut off tetrodes into linear triodes.

now i am remembering things better... the mac scheme is a splitload inverter. it's equal windings for plate and cathode. there's no voltage gain - 100% feedback. the trick is the screen... it's wired cross coupled with the other side... that keeps the screen "ON" because it swings up when the plate swings down. this is what saves the class B arrangement from crapping out from dead zone operation. one can get very close to cutoff before the feedback fails. the pp output stage works just like a cathode follower.

imagine a split load inverter with equal windings in the plate and cathode... if it's a triode, it's simple. but add the screen, and there's a problem: if the screen tracks the plate, it's triode operation. if it's fixed, it works against the cathode voltage in a really shitty fashion, because the cathode moves. also, in class B, a dead zone (where both tubes are off for a short time) will produce horrific distortion. it is nearly impossible to match tubes for cutoff, by the way. especially triodes! turning "on" is not an even process. no tube is the same and they are mainly strange. sharp cutoff tubes are actually better at turning "on", no surprises there. however, reverse the phase of the screen winding ensures the tubes operate as tetrodes/pentodes. the screen voltage and current is high, when the plate swings low... that is the clever part of the the Mac arrangement. it really takes advantage of multigrid tubes. but the entire idea is ALL about the driver stage. what you are really listening to in a Mac amp is the driver stage. there is NO voltage gain in the output stage. only current gain.

if the secondary is allowed to "float", the 4 ohm center tap can be used to return positive current feedback to the output stage. which adds gain which is then lumped in with the negative feedback. this lowers distortion more and slightly relieves the driver stage from having to do all the work. i believe Mac used this trick on the big triode amps? something like it, anyway... that idea was invented by Herb Keroes. not Mac.

I initially looked at it as a split load inverter and the other trick beyond the cross screen reference is the bifilar pair is the plate of tube 1 and the cathode of tube 2 (which are in phase with each other.) My conceptual problem comes to what happens with a split load when the loads are unmatched which is what I would expect class B would represent.

spice doesn't seem to care much if it is class A / AB pentode or triode but what concerns me is the differential output Z should depend on the loading and class of operation.

the one big advantage you are not taking into account is the core. it is a split load, but the core is ALSO a feedback mechanism.

for a 300B

ah i fucked up... here is the right one.

and then the pp for triode

i just kept reducing the load on the secondary until i got the same AC current in the load. note that at that load, it is unity gain.

When R1 is 600 ohms with pp 300Bs the voltage gain is unity. each winding is identical in this experiment.

and to go to tetrodes, and a tertiary winding... two more identical windings, cross coupled.

the next next thing is to increase the bias so that the tubes become insanely lean and look at the fft. the sim above is really really good. almost scary low distortion. but that's class A. next next is class B

300 B's are not so efficient in class B because the mu is so low... but still, there's a lot of swing when the bias leans things out... with -100V bias in this circuit (11 mA per tube) it's easy to drive 600 Vpp in without clipping... the third harmonic is high, but this is what Macintosh is all about... i made the winding resistance 100 ohms per coil. the feedback really keeps things together even when the tubes are almost off. with high mu triodes, and tetrodes you need less swing into the grids... this explains the entire design strategy.

now the secondary can be scaled to a practical speaker...

here's a sweep of the current in the 300Bs

and the output current into 600 ohms

sorry about the crazy images... i don't know the best way to show them here... i switched to 811As with the higher mu. and directly coupled the grids, which is more likely. the unity coupled idea really does work to keep shit together in class B. you can go down to 4mA per output tube without them creating so much dead zone that the notch appears. not at least in spice. i have worked on Mac 275s with 10mA per 6550... and no notch. so we know it works.

playing around with the bias and 10H windings, i found a sweet spot at +8 volts and 24 mA per 811A cold. peak current with 800 volts on the plate is 190 mA. third harmonic is -72dB. there is no doubt it's an interesting technique.

190mA per plate.

this is amazing... 100 watts and okay distortion. just 2 tubes. but needs 700 Vpp drive to do it. a hi mu class B triode or a tetrode will always be easier... with unity coupled circuits.

Hey JC... thanks for the stephie-esque teardown of the circuit. I split it off to its own thread and sized the screen grabs down.... hope I didn't eff this up.

It is the concertina type behavior of this that has me interested. A long time ago i did a sim and built the IT's for the "chokertina" which is kinda a unity couple splitter. My thinking was that the close coupling of the anode and cathode will assure identical loading of each making the differential output Z always ~ 1/gm or that of a cathode follower.

Looking at the mac circuit it still bothers me that for class B operation one tube is shut off which I see totally messing with the balance. I have no doubt the circuit works but I am still having a tough time getting my head around what is really going on.

it's feedback that makes it work. error correction 101. really.

and the coupled windings.... that's why it really helps to start with the split load. a split load with a 1:1 transformer. tightly coupled and 100% feedback.

with chokes it isn't going to work right.

just as a test, i tried a 6922 with an 80 volt B+. 4 henry windings... and with as much signal as i could before clipping. 31 volts! the amazing thing is you can see the feedback keeping it together just before it craps.

the complete harmonics flat across the sweep is an indication of feedback holding them down! this should be clipping but there's enough fb to correct it at this point.

sorry it's 60 volts! not 80 volts of supply. brutal conditions. it still works amazing. it's a fabulous use of a transformer. but this is not the exact Mac tranny... this is what they used for the Mac class triode amps in the output stage. the splitload phase splitter is not a Mac invention. it's in the Radiotron... someone else.

the parallel i was trying to draw was between the mac concept and the split load inverter. Both are unity stages and if loaded equally the differential Z-out should be that of a cathode follower. There have been several long drawn out discussions about this and if you examine the plate or the cathode to ground independently you get the traditional expected Z-out of the respective plate / cathode follower. However... when used as a splitter into the grids of A1 tubes the loads can be considered matched and each grid is driven from the Zout of a cathode follower. The diff output z is 2*(1/gm) and each to ground is 1/gm. I have verified this multiple times in spice.

It took me a while to get my head around all of the phase relationships and it is starting to make sense to me.

Now the thing that is kinda interesting me is the output Z when triode wired vs. pentode wired ala mac.

For class A the pentode connection I get a Z out of 470Ω or just about Triode Rp of the 6L6 and for triode wiring I get 420Ω. The minute I hit class B my method of determining output Z by driving the output with a current "breaks" as does the more traditional way of measuring output Z.


dave

it's not weird as you think. in a splitload the Z out is the same until the bottom clips. that's what you get for the feedback. but the unity coupled splitload is better! the 1:1 coupling lowers Z more, and increases the input Z. if you make a splitload with 50 volts across it with resistors, AND compare the same to the example above... it's no wonder Macintosh has held out all these years. this would make a good circuit, by the way! i would do it.