Mac Unity coupled

All the parts have to be put together at some point. Have fun!
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

class B involves "offness". but it is critical for success that the off time does not coincide with the opposing tubes offness... so even in a "normal" transformer meant for class B operation (pp with a gap), the "mixing action" of the transformer is part of the success story when it all goes right. if the dead zone of one tube extends into the other tubes dead zone, you get crossover distortion. the Mac unity coupled circuit clearly avoids this unless bias is lost. there are several error correction mechanisms at work: the pair of split load inverters, in push pull have as much feedback as the mu and gm allow... and the core is always behind the ball, so there is energy stored in there to fight "offness". it isn't so important that the impedances are different! the cross coupling of the screens is also a part of the error correction system. like the core, energy is stored there that can pull on the cathode when the plate can't.
my skin is thin, my skull is thick.
dave slagle
Posts: 2092
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

but my graphs / tests of output impedance are at a single winding into the load so it is representative of the output Z of the amp. In the case of less than 100% coupling we see the output Z rise with frequency within the audio band. I would't be surprised if it goes from CF outZ at low frequencies to pentode Rp out Z at high frequencies.

oddly enough things don't change much if the 6L6's are triode wired???

dave
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

yes, but this is only the output stage. i'm not sure yet about your method. i'm going to sim it low and high at see how it graphs. normally there would be another feedback loop back to the driver stage which could be optimized for the problem areas... i can't see anything i dislike so far! and that is surprising!!! i never cared much for Macs. they are plain. not fun.

but i have to say, there's a lot to recommend the output stage. and why not class A or AB? why milk everything for the last crumb?

it's interesting for higher mu triodes especially! 211s, 203's, 811As, and 805's!
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here are some images of the winding currents at 20KHz
Attachments
anode_current_20khz_144.png
anode_current_20khz_144.png (111.46 KiB) Viewed 12269 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

the anode winding current is interesting because of the effect of the screen current... above is the anode current. here is the cathode winding current
Attachments
cathode_winding_current_20khz_188.jpg
cathode_winding_current_20khz_188.jpg (91.55 KiB) Viewed 12268 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

and the screen current... i see TIM on the leading edge by the way. the phase shift of the tranny is producing increased delay and transient intermod on the attack edge. the decay edge is therefore shifted in time. the waveform is slewed. i believe the sim! but the magnitude is not entirely believable. it looks highish. it may be i have too much inductance in the windings and it needs to be scaled down for design. the Mac trannies do have a much lower inductance per winding than a "normal" output trans. this would definitely help the high frequency performance. class B is going to exaggerate the problem, however. the magnitude of the slewing is related to the amplitude of the signal and going from lean to peak current.

the sine wave is slewed in the output. and the causes are clear.... but a look at the screen current should show the most serious problems. what happens when the anode winding current is off.
Attachments
secondary_winding_current_145.jpg
secondary_winding_current_145.jpg (75.02 KiB) Viewed 12267 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

oops. that is the secondary current... showing the effects of slewing. bhere is the screen current...
Attachments
screen_current_20khz_353.jpg
screen_current_20khz_353.jpg (81.48 KiB) Viewed 12266 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

in the screen current you can clearly see transient modulation on the attack edge. this coincides with overlap of the screen current being sourced from the opposing plate before it's time. the screens and plates are no longer in sync, either. the current is increasingly delayed with level. inductors and resistors can produce TIM all by themselves so this is not necessarily a feedback related effect. actually, it looks to me that coupling between the windings is fighting the extra tug at the anode winding current from the screens... if that is right, it's yet another plus for the unity coupling idea. but there is more impedance at the plate (less current) than at the cathode in this circuit clearly, and the screen current is responsible. inductive phase shift and the timing of the overlapping offness is responsible for the distortion. i'm not yet convinced that the difference in impedance between the plate and cathode is as big of a problem for the output current as the screen current's effect on the overlapping offness. but a look at the tertiary winding arrangement would be next. it should improve things considerably. too much winding inductance is definitely NOT a good thing for designing this transformer.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

in terms of output impedance, the next test will be voltage changes with load changes at high frequencies... but the magnitude isn't going to be large. one can already see that in the 8 ohm load. the peak current is the same, although the duty cycle is changed by some percent. a modest rise. but i will run that later today. i think the unity coupling works to correct the entire mess in the secondary.
my skin is thin, my skull is thick.
dave slagle
Posts: 2092
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

the situation with the output impedance that I am interested in is the comparison between class A operation and class AB/B. I finally found a level dependent squirrel. Download and run the file below and explain to me what is going on. Looks like a spontaneous oscillation that happens after a set period of time determined by the input AC voltage.

you may need to insert your own tube model and I think it is related to that since I tried three different tube models and they all behaved differently. Do you have a known good 6l6 model?

It very well may be that I am chasing spice artifacts due to poor models.
Attachments
mac output.asc
(3.61 KiB) Downloaded 493 times
Screen Shot 2021-01-21 at 9.28.18 AM.png
Screen Shot 2021-01-21 at 9.28.18 AM.png (55.32 KiB) Viewed 12285 times
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

the screens are hooked up backwards.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i replaced the 6L6GC model with a plain vanilla 6L6 version. the 6L6GC oscillated and did weird stuff. the 6L6 works fine. you have the screens reversed. and the inductance is WAY WAY too high. it still works but slews like crazy.
Last edited by nanana on Thu Jan 21, 2021 9:57 pm, edited 1 time in total.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

oops. here's the corrected image. i loaded the secondary with enough to draw the same current as any other single winding. it works... but the inductance should be reduced to between 6 and 10H.
Attachments
dave_unity_cpoupled_109.jpg
dave_unity_cpoupled_109.jpg (47.43 KiB) Viewed 12270 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

actually the 2 K load draws 4 times the current! so the output Z is very low!
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

in any case, i think the mystery is solved.

the unity coupled output stage is very good. and very interesting! and i see NO reason to use it only in class B. wtf! other than economy? it's an inefficient circuit because of the 100% feedback. it is applied just like a cathode follower. it takes a lot of drive to get your 20 watts, or 50... but it's stable! the phase shift is error corrected into something quite hard to argue with, from a classical engineering standpoint! so more feedback can be added around the whole stage, if needed. but open loop, it's easy to get under 1% distortion at a useful power. that is all by itself very interesting! and it's old tech! done so as to make transformer making cheaper and less expensive. almost no interleaving and macintosh never used anything special in any of it's products. super plebeian. i am very impressed... and i never was before. a pair of 841As or even 811As is very tempting in the straight up triode bifilar arrangement.
my skin is thin, my skull is thick.
dave slagle
Posts: 2092
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

the drawing I uploaded was for the triode wired version.... the pentode with the screens cross coupled did the same thing. turns out it was the tube model again.....

dave
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here is another money shot... from your circuit! i increased the voltage to 400 volts because this 6L6 model just doesn't have enough cajones at 350... i reduced the inductance to 10 H each winding. i increased the screen current limiting resistors to 500 ohms. 2.5K gives the same current in the load as both 6L6s together. this is just before clipping 360 volts pp balanced input... look at the error correction of the unity coupled transformer and split load inverters: the screen is drawing hellacious peak current away from the plates... the cathode windings show the correction. there is very little distortion in the load!

dave, this is a good circuit!
Last edited by nanana on Thu Jan 21, 2021 9:51 pm, edited 1 time in total.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

whoops! here it is...
Attachments
power_of_feedback_348.jpg
power_of_feedback_348.jpg (95.41 KiB) Viewed 12252 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

the green is the plate winding, you can see the clipping caused by the peaking screen on the opposing side. and yet in the cathode (the blue trace), the subtracted third from the sine... (tall and narrow). this is the power of feedback. the sum of these is a near perfect sine of current in the 2.5K resistor.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here is the circuit with 350V supply and 60mA per tube. class A - AB old school. with 240 pp input and into 8 ohms, about 12 watts, no additional error correction... you can see the onset of clipping just coming in the current waveforms of the windings... it looks horrible! output distortion just over 1%! all 3rd. compared with pp 2A3s in the ordinary manner w/o fb, 10 watts at 5% (mostly 3rd).
Attachments
class_a_60ma_per_tube_b4_clipping_195.jpg
class_a_60ma_per_tube_b4_clipping_195.jpg (123 KiB) Viewed 12251 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here's the output
Attachments
output_8_ohms_144 (1).jpg
output_8_ohms_144 (1).jpg (134.73 KiB) Viewed 12250 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

here's the output sweep
Attachments
fft_205.png
fft_205.png (89.44 KiB) Viewed 12249 times
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

just to summarize a few important things, the bifilar unity coupled circuit is definitely based on the split load inverter. in push pull, 2 inverters can work well in class B... the coupling between the windings is also a feedback system, besides the degenerative fb inherent in the split load. the system has no voltage gain... only current gain. it works just as well in class A as it does in class B... the internal fb keeps things together until clipping, or to the point where the screen grids no longer sync with the plates from the transformer phase shift (at some high frequency).

can you think of anything to add?
my skin is thin, my skull is thick.
dave slagle
Posts: 2092
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

the only other thing I can offer is that it also appears to work the same with triodes as it does with pentodes which is odd since in the texts they say how a key element is the plate connection to the opposite screen.

might be a good place to use an elektromakano M7 as a driver. DHT with a gain of 100 and an rP of ~50K

dave
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

the mu in tetrode operation is much higher, so there is more internal feedback. it isn't the same! transconductance does matter for a follower, but most important for insertion loss is mu. this is why 5687's are such lousy cathode followers all by themselves, and 6922s or 6C45Ps are such excellent cathode followers! low mu tubes need a little xtra to be great. a mu of 16 vs. a mu of 36 or more. for a 6L6, the loss in gain is nearly 5dB in triode. in tetrode operation, it's 0.5 dB. with a lossy stage like this, every dB counts!. the tetrode is lower distortion too, although not by much. until clipping this is an inherently low distortion design.
Last edited by nanana on Fri Jan 22, 2021 3:40 pm, edited 2 times in total.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

i have a pair of M7s! it does sound perfect for this...
my skin is thin, my skull is thick.
dave slagle
Posts: 2092
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

you are correct about the gain... for tetrode I actually see precisely unity gain so it isn't quite the CF I thought it was since they max out at ~0.9. For triode I see a gain of 0.8 so it doesn't quite fit the CF either.

I'm still really interested about the output Z as you dynamically change class of operation. For example in a class AB PP circuit operating in class A each tube sees 1/2 the A-A impedance and in theory no AC signal current traverses the CT. As you shift to class B, the CT starts seeing signal current and suddenly the load for that current follows the turns squared so the load becomes 1/4 A-A.

dave
Get Your Fix
www.hifiheroin.com
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

output Z is inversely proportional to transconductance in tetrodes/pentodes, and a ratio in triodes. high gm is low Z. in the transition from A to B, it's going to be Z/2, to a bit more than Z, as only one tube is on at a time... so you already know the answer to your question. but there's a wrinkle in the transition through the AB. because both tubes go from a peak gm to not fully on... it's an "S" curve, and it's two tubes that won't be perfectly matched through turning off/on. so it is not a smooth transition. you can plot this easily with a step level sweep. there will be ripple. not a lot, but not smooth.
my skin is thin, my skull is thick.
nanana
Posts: 143
Joined: Mon Jun 14, 2010 7:34 pm
Location: malmö sweden

Post by nanana »

also, i know the formula for triode follower gain ends up at around "0.9". it is NEVER that high in reality. it's always less, unless there's gain somewhere else. low mu in a triode with the normal loss of gm with aging, in particular, really bleeds gain. most CFs are 0.8, or even a bit less, unless it's a high mu triode. tetrodes are 0.9. and less sensitive to gm.
my skin is thin, my skull is thick.
Post Reply