I have to take issue with the idea of no crossover distortion occuring in the Mac. The bifilar windings prevent inductive flyback spikes in the OT (from leakage L, so not absorbed by the load) when a tube cuts off by ensuring the other tube/winding can absorb that (high mutual inductance between the bifilars). Class B operation makes this an issue because the other tube is barely conducting at crossover, so has minimal conduction to control the current. Class AB operation largely avoids this issue since the other side tube has significant gm at the cutoff of the 1st tube.
The bifilar scheme simply fixes the switching transients so as to ALLOW class B operation. It does not fix the usual crossover distortion caused by the sum of the tube gm's not staying constant. Operating in/near class B is what aggravates the usual gm crossover distortion, so you COULD conclude that Mac's innovation simply allowed them to make a bad crossover distortion amplifier. Sort of the ultimate fix to the wrong problem. What it mainly does for them is get better tube power efficiency in class B.
Not that there is anything wrong with fixing the switching spikes, but it still needs to operate in class AB to avoid crossover distortion, or else massive feedback is needed to fix that. Or an elegant crossover fix like I mentioned is needed (which is NOT just local feedback, see below).
By the way, the circlotron scheme fixes the switching transients even better than bifilar. And its even a lot easier to find an OT suitable for a circlotron. (Hammond 1650T, operated with an 8 Ohm load on the 16 Ohm sec winding, giving a 1000 Ohm P-P primary, and moving the low freq. spec down to 15 Hz instead of 30 Hz due to the extra turns)
My issue with the AR scheme is that its just 5% or so CFB, not really enough to fix much crossover dist. If you try to move the CFB back to the driver stage, the CFB effect becomes much larger, but then it's really just a form of global FDBK, with the usual OT phase issues.
Citation II gets closer to doing it "right" by Schading back to the driver inputs. But then loses it with the UL outputs, and the -crossed- Schade feedbacks ("local" feedback has to then go thru the OT).
RCA comes the closest to the ultimate design I've seen with their handbook amplifier (RC-30 page 696). I would just replace its output stage with a circlotron stage to reach ultimate status. Why this circuit? It has hidden in it the magic crossover fix scheme Broskie mentioned. The Schade feedbacks to the driver cathodes. Current sense resistors are implicit in the OT primary winding resistances. Some improvements could still be made such as a CCS tail for the driver. Input stage and splitter can be anything you prefer.
I had often wondered why the RCA circuit used two types of Schading (to the driver cathodes, and to the driver plates). I had surmised that something was different between the two, but what? Well, the cathode feedbacks are differential, while the plate to plate feedbacks are SE for each output tube. Broskie's "magic fix" article (but in totem pole form), revealed the secret.
Looking at the SS totem pole circuit form, the feedback pickoff would transition linearly for a class A output stage. But notice what happens with class AB, the pickoff transitions smoothly until the crossover region is reached, then it speeds up, and then returns to smooth tracking once the crossover region is departed. This is NOT LINEAR feedback!
This is ERROR ENHANCED feedback. It is similar to Hawksford Error Correction in that it has an enhanced error component, but here it is mixed with the linear tracking component. (HEC computes just the error component alone).
This means that the Schade loop using this feedback only needs a FINITE amount of feedback to completely fix the crossover error. So if one were to adjust the "magic" feedback level, one would find a "Null or minimum" error point (crossover dist. null).
This is why the RCA amp has the two Schade types. You only adjust the cathode feedback to the crossover null point. Any more Schading (lower output Z) has to be done by plate to plate Schading.
The question remaining in my mind is whether the RCA designer actually knew what was going on here. Is there some obscure article in the RCA Review journals describing this? Seems that no one else ever caught on to it. The SS totem pole equivalent versions of this were patented later at least 3 times.
In any case, this RCA scheme is WAY WAY more sophisticated than the Mac design. No one seems to have ever figured it out till this day. There you have it.
Don
European Triode Festival and Crossover Notch Distortion and New OTL Design
(2nd diagram up from bottom)
The bifilar scheme simply fixes the switching transients so as to ALLOW class B operation. It does not fix the usual crossover distortion caused by the sum of the tube gm's not staying constant. Operating in/near class B is what aggravates the usual gm crossover distortion, so you COULD conclude that Mac's innovation simply allowed them to make a bad crossover distortion amplifier. Sort of the ultimate fix to the wrong problem. What it mainly does for them is get better tube power efficiency in class B.
Not that there is anything wrong with fixing the switching spikes, but it still needs to operate in class AB to avoid crossover distortion, or else massive feedback is needed to fix that. Or an elegant crossover fix like I mentioned is needed (which is NOT just local feedback, see below).
By the way, the circlotron scheme fixes the switching transients even better than bifilar. And its even a lot easier to find an OT suitable for a circlotron. (Hammond 1650T, operated with an 8 Ohm load on the 16 Ohm sec winding, giving a 1000 Ohm P-P primary, and moving the low freq. spec down to 15 Hz instead of 30 Hz due to the extra turns)
My issue with the AR scheme is that its just 5% or so CFB, not really enough to fix much crossover dist. If you try to move the CFB back to the driver stage, the CFB effect becomes much larger, but then it's really just a form of global FDBK, with the usual OT phase issues.
Citation II gets closer to doing it "right" by Schading back to the driver inputs. But then loses it with the UL outputs, and the -crossed- Schade feedbacks ("local" feedback has to then go thru the OT).
RCA comes the closest to the ultimate design I've seen with their handbook amplifier (RC-30 page 696). I would just replace its output stage with a circlotron stage to reach ultimate status. Why this circuit? It has hidden in it the magic crossover fix scheme Broskie mentioned. The Schade feedbacks to the driver cathodes. Current sense resistors are implicit in the OT primary winding resistances. Some improvements could still be made such as a CCS tail for the driver. Input stage and splitter can be anything you prefer.
I had often wondered why the RCA circuit used two types of Schading (to the driver cathodes, and to the driver plates). I had surmised that something was different between the two, but what? Well, the cathode feedbacks are differential, while the plate to plate feedbacks are SE for each output tube. Broskie's "magic fix" article (but in totem pole form), revealed the secret.
Looking at the SS totem pole circuit form, the feedback pickoff would transition linearly for a class A output stage. But notice what happens with class AB, the pickoff transitions smoothly until the crossover region is reached, then it speeds up, and then returns to smooth tracking once the crossover region is departed. This is NOT LINEAR feedback!
This is ERROR ENHANCED feedback. It is similar to Hawksford Error Correction in that it has an enhanced error component, but here it is mixed with the linear tracking component. (HEC computes just the error component alone).
This means that the Schade loop using this feedback only needs a FINITE amount of feedback to completely fix the crossover error. So if one were to adjust the "magic" feedback level, one would find a "Null or minimum" error point (crossover dist. null).
This is why the RCA amp has the two Schade types. You only adjust the cathode feedback to the crossover null point. Any more Schading (lower output Z) has to be done by plate to plate Schading.
The question remaining in my mind is whether the RCA designer actually knew what was going on here. Is there some obscure article in the RCA Review journals describing this? Seems that no one else ever caught on to it. The SS totem pole equivalent versions of this were patented later at least 3 times.
In any case, this RCA scheme is WAY WAY more sophisticated than the Mac design. No one seems to have ever figured it out till this day. There you have it.
Don
European Triode Festival and Crossover Notch Distortion and New OTL Design
(2nd diagram up from bottom)
Last edited:
Here is a scan out of my RC-30 handbook. I think you can find better (larger) scans on the WWW. There were some similar amps from RCA earlier too. RC-19 had the 6V6 amp, and the SP-20 amp was commercially offered. Although the SP-20 has only the cathode Schades.
"I wish I had some idea what you guys are saying to each other."
Sorry to be confusing here. This admittedly takes a good grasp of both SS (solid state) design and hollow state design to follow. So is no doubt confusing to any electronic newbie. Maybe some clarifiers:
a) In the Broskie SS totem pole schematic, the magic Schade feedback pickoff would be one of the 100 Ohm to 100 Ohm tie points. The equivalent of the tube driver stage (one side) would be the transistor with its emitter connected to the 100 -100 Ohm tie point. The tube version just has its cathode connected there, equivalently.
b) The tube version uses a center tapped OT for tube loading, while the SS version is totem poled.
Converting over is not too hard here fortunately, since the SS version is drawn with collector outputs just like a tube stage. The 0.1 Ohm current sense resistors are in the SS collector to load path. Similarly, the current sense resistors for the tube version would be inserted between the plate and OT winding. Since the typical OT has significant winding resistance already, it is inherently present without adding a real resistor. Hence no plate resistors show up in the RCA schematic.
"I wish I had some idea what you guys are saying to each other."
Sorry to be confusing here. This admittedly takes a good grasp of both SS (solid state) design and hollow state design to follow. So is no doubt confusing to any electronic newbie. Maybe some clarifiers:
a) In the Broskie SS totem pole schematic, the magic Schade feedback pickoff would be one of the 100 Ohm to 100 Ohm tie points. The equivalent of the tube driver stage (one side) would be the transistor with its emitter connected to the 100 -100 Ohm tie point. The tube version just has its cathode connected there, equivalently.
b) The tube version uses a center tapped OT for tube loading, while the SS version is totem poled.
Converting over is not too hard here fortunately, since the SS version is drawn with collector outputs just like a tube stage. The 0.1 Ohm current sense resistors are in the SS collector to load path. Similarly, the current sense resistors for the tube version would be inserted between the plate and OT winding. Since the typical OT has significant winding resistance already, it is inherently present without adding a real resistor. Hence no plate resistors show up in the RCA schematic.
Attachments
Last edited:
c) Crossover distortion occurs when the two P-P (push-pull) devices controlling the output do not sum to constant gm (control voltage to output current figure of merit). Most device's gm also vary with current. So you can see the problem at crossover (one device slowly turning off while the other is slowly turning on). The gm's are varying with current to some non-liner rule usually. And for class AB (and especially class B for efficiency) the device currents are very small in the crossover region. So a large drop in combined gm's occurs. This makes for distortion when driving a load, since the mover of the load is getting weak during the crossover.
d) This calls for some kind of control correction to fix. The usual kind is negative feedback. Unfortunately linear neg. FDBK has diminishing returns when more is used. An infinite amount is required to ideally fix a problem.
Enhanced error feedback can do better in theory. Only a finite amount of feedback is required to fix the problem. Hawksford Error Correction is the ultimate form of this, using feedback of only the error alone. It requires just unity feedback gain. Here we have enhanced error combined with some linear signal, so finite feedback loop gain is required. In reality, these schemes are limited by the quality of this enhanced error signal, so some null point can be found where best but not total error cancellation occurs. The HEC scheme is interesting to play with, because you can see the output distortion actually invert when too much correction is applied. But the null point in between is still not a true zero due to 2nd order and above effects polluting the correction.
The big advantage here with the "magic fix" is that it only requires some resistors to be added to an existing design. (well, not even that for the tube version, some of the resistors are implicit already in the OT) Resistors are generally quite linear. Broskie mentioned 20 dB of distortion improvement when simulated (I assume this was for the SS version). That's 100X, not bad. The tube crossover is generally more linear than the SS crossover to begin with, so we should be flying high here in hollow state land with this fix implemented.
d) This calls for some kind of control correction to fix. The usual kind is negative feedback. Unfortunately linear neg. FDBK has diminishing returns when more is used. An infinite amount is required to ideally fix a problem.
Enhanced error feedback can do better in theory. Only a finite amount of feedback is required to fix the problem. Hawksford Error Correction is the ultimate form of this, using feedback of only the error alone. It requires just unity feedback gain. Here we have enhanced error combined with some linear signal, so finite feedback loop gain is required. In reality, these schemes are limited by the quality of this enhanced error signal, so some null point can be found where best but not total error cancellation occurs. The HEC scheme is interesting to play with, because you can see the output distortion actually invert when too much correction is applied. But the null point in between is still not a true zero due to 2nd order and above effects polluting the correction.
The big advantage here with the "magic fix" is that it only requires some resistors to be added to an existing design. (well, not even that for the tube version, some of the resistors are implicit already in the OT) Resistors are generally quite linear. Broskie mentioned 20 dB of distortion improvement when simulated (I assume this was for the SS version). That's 100X, not bad. The tube crossover is generally more linear than the SS crossover to begin with, so we should be flying high here in hollow state land with this fix implemented.
Last edited:
OOPS,
I forgot to mention that the tube version requires subtracting the two Schade feedbacks to get the crossover error term, hence the need for a differential driver stage. The SS totem pole setup has this easier, since the two pickups are already inverted with respect to each other, just requiring resistor summing.
I forgot to mention that the tube version requires subtracting the two Schade feedbacks to get the crossover error term, hence the need for a differential driver stage. The SS totem pole setup has this easier, since the two pickups are already inverted with respect to each other, just requiring resistor summing.
I wish I had some idea what you guys are saying to each other.
you'll catch on.
just watch.
Last edited:
I wish I had some idea what you guys are saying to each other.
Buy a copy of Morgan Jones's "Valve Amplifiers", study it thoroughly, and all will be clear.
No, he won't. I believe his idea of recreating a magical moment/jaw dropping experience from the past is bound to be a fiasco. Who would be the tech to build this amp anyway? 3000$? Is it worthy? Probably not. I smell disappointment. The best thing he can do is to forget the old lover and find a modern one. I have read the Leben CS-600 sounds like the old MC-275. A bit expensive though but he will probably find an used unit that will fit his budget. But you got to listen to modern tube amps, there's no other way to find out. Don't know where you live but it shouldn't be that difficult.
Obsessive and obstinate are not synonyms. I am listening and learning. You might lose that bet. I'm wondering if a good tube preamp paired with a very clean SS power amp might achieve my goal. I can remember being stunned by the crispness of the old Phase Linear 700B, but at the same time not wanting to be in the same room with it. I'm looking for the warmth and voice and breath and soundstage of a high quality tube amp. So maybe.....all I need is a good tube preamp. Still learning.......
"I'm looking for the warmth and voice and breath and soundstage of a high quality tube amp. So maybe.....all I need is a good tube preamp. Still learning....... "
A tube preamp could be a good place to start. Next step up would be something like Pete's Red Board amplifier, inexpensive, certain to work and work well. Lots of info on it here and related experiments on the forum, and if more power is needed, go with Tubelab's upgrades, maybe 6HJ5 tubes. These will get your feet wet in the hobby without breaking the wallet. From there you can join the rest of the obsesives here at building the next best greatest amp. But I sure would not recommend cloning the Mac as a first project. The most experienced here would likely and rightly be quite wary of this effort, not to mention the cost or possible dissapointing outcome.
Don
A tube preamp could be a good place to start. Next step up would be something like Pete's Red Board amplifier, inexpensive, certain to work and work well. Lots of info on it here and related experiments on the forum, and if more power is needed, go with Tubelab's upgrades, maybe 6HJ5 tubes. These will get your feet wet in the hobby without breaking the wallet. From there you can join the rest of the obsesives here at building the next best greatest amp. But I sure would not recommend cloning the Mac as a first project. The most experienced here would likely and rightly be quite wary of this effort, not to mention the cost or possible dissapointing outcome.
Don
Unless you choose to have a bifilar McIntosh OPT custom wound at expense,
Or purchase two regular transformers to twin couple per Crowhearst, or give
up in frustration and abandon the entire project completely...
I still have to submit the final output stage of ARC's design as the very most
reasonable compromise. You get some McIntosh style cathode feedback. You
get feedback of the true output, and not just the primary. And need only the
single output transformer of common manufacture. Whats not to love?
Whole tangent about class B crossing distortion is neither here nor there...
Use common 6L6 family output tubes in AB. They are purposely designed to
cross nicely when you run them at reasonable bias.
Make an amp that measures well into a resistor without any global feedback.
Then add GNF to tighten the damping only if your loudspeaker is booming too
much with its own natural resonances.
Or purchase two regular transformers to twin couple per Crowhearst, or give
up in frustration and abandon the entire project completely...
I still have to submit the final output stage of ARC's design as the very most
reasonable compromise. You get some McIntosh style cathode feedback. You
get feedback of the true output, and not just the primary. And need only the
single output transformer of common manufacture. Whats not to love?
Whole tangent about class B crossing distortion is neither here nor there...
Use common 6L6 family output tubes in AB. They are purposely designed to
cross nicely when you run them at reasonable bias.
Make an amp that measures well into a resistor without any global feedback.
Then add GNF to tighten the damping only if your loudspeaker is booming too
much with its own natural resonances.
The new MC275 retails for $4500. That would buy a lot of parts and chassis, if you could get the right transformers.
Correction/Amendment on the earlier described "magic" feedback scheme.
Sorry about going a bit off topic again, but a couple of clarifications in the "magic" feedback concept need fixing up for those reading earlier. After thinking about the current modified feedback some more, it obviously does not have a sharp null to the crossover distortion like Hawksford Error Correction does. Although it is non-linear feedback in the sense of voltage sensed feedback, it IS linear feedback in the sense of differential current sensed feedback. So what happens is the two output device characteristics get modified so that they fit together progressively more perfectly with more of this feedback.
Net result is it approaches a smooth current versus output voltage curve, ie, constant impedance. Since the uncorrected magic feedback signal is quite distorted with respect to the voltage output under load, a finite amount of this feedback will still make a big improvement in crossover dist. Hence the 100X figure noted.
But possibly more importantly, constant output Z versus signal is one of the holy grails for removing "unmeasurable" amplifier sound artifacts when connected to a variable load Z like a speaker. This scheme may even improve a class A amplifier somewhat.
Also, it is possible some of the Mac designs DO effectively use this same current correction too. If the output stage CFB feedbacks are also taken back to the previous class A driver stage, the resistance of the cathode winding could provide this effect. I recall seeing at least one Mac schematic that has this setup. Still, this would work much better with class AB biasing though, since the "magic" FDBK is fitting the two output device curves together, some overlap is needed to be effective.
....Sorry for the interruption. Back to your regularly scheduled programming.....
Oh..., and I agree with Ken on the AR type design as another practical amp to build as long as the ultra Mac sound duplication spec is relaxed. Class AB biased for sure. Maybe even some kit PC boards available?
(edits for removing distortions #2871 and #5201107
Sorry about going a bit off topic again, but a couple of clarifications in the "magic" feedback concept need fixing up for those reading earlier. After thinking about the current modified feedback some more, it obviously does not have a sharp null to the crossover distortion like Hawksford Error Correction does. Although it is non-linear feedback in the sense of voltage sensed feedback, it IS linear feedback in the sense of differential current sensed feedback. So what happens is the two output device characteristics get modified so that they fit together progressively more perfectly with more of this feedback.
Net result is it approaches a smooth current versus output voltage curve, ie, constant impedance. Since the uncorrected magic feedback signal is quite distorted with respect to the voltage output under load, a finite amount of this feedback will still make a big improvement in crossover dist. Hence the 100X figure noted.
But possibly more importantly, constant output Z versus signal is one of the holy grails for removing "unmeasurable" amplifier sound artifacts when connected to a variable load Z like a speaker. This scheme may even improve a class A amplifier somewhat.
Also, it is possible some of the Mac designs DO effectively use this same current correction too. If the output stage CFB feedbacks are also taken back to the previous class A driver stage, the resistance of the cathode winding could provide this effect. I recall seeing at least one Mac schematic that has this setup. Still, this would work much better with class AB biasing though, since the "magic" FDBK is fitting the two output device curves together, some overlap is needed to be effective.
....Sorry for the interruption. Back to your regularly scheduled programming.....
Oh..., and I agree with Ken on the AR type design as another practical amp to build as long as the ultra Mac sound duplication spec is relaxed. Class AB biased for sure. Maybe even some kit PC boards available?
(edits for removing distortions #2871 and #5201107
Last edited:
FWIW, the sounds I want has this compressed feature. The Yaqin I am using sounds somewhat flabby. In the meantime, I will continue to seek enlightenment, from you guys and from the literature.
I'd like someone to explain the term "compressed". It's used a lot on the boards and I wonder if it means one thing to some people and something else to another.
If I had to guess, I think it's a reference to frequency response. A non-flat, roll-off of highs and lows. An amp that doesn't have a wide FR.
My understanding of a compressor for signal processing is that it is an automatic volume control of sorts in that it limits strong signals to a maximum level as desired, keeping unwanted high inputs at the same highest level of the primary material.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.