"In the Mac circuit you’ll still need ~1/2 B+ voltage swing to drive the output tubes, even with the bootstrapping, so I don’t think so.."
I agree that there is a large voltage swing on the driver plate(s) as you state. And this is why everyone has been fooled into thinking there is CFB here. The "real" voltage drive from the driver however is the voltage ACROSS the resistor, since one end is swinging at cathode potential (AC wise, not DC).
One can look at this as a current source level translating circuit, the driver supplies just enough current to generate the output grid to cathode drive across the load resistor regardless of where it floats. And it is floating wildly around at the cathode voltage due to the cathode winding (again AC wise).
To see that there is no CFB activity, one need only imagine that the cathode be shifted by some error voltage. Real CFB would notice the error and correct the drive signal level (grid to cathode). But here, the plate load resistor just shifts by the error voltage, and no correction voltage is generated across it to correct. Cathode swings are hence transparent to the driver, hence no CFB. The Mac CFB is an illusion.
I should clarify that this analysis was for a pentode driver, hence current drive to the load resistor. With a triode driver stage, there would be some NFBK due to cathode swing. But it is feedback to the driver stage then, not classic CFB in the output stage.
Don
😎
I agree that there is a large voltage swing on the driver plate(s) as you state. And this is why everyone has been fooled into thinking there is CFB here. The "real" voltage drive from the driver however is the voltage ACROSS the resistor, since one end is swinging at cathode potential (AC wise, not DC).
One can look at this as a current source level translating circuit, the driver supplies just enough current to generate the output grid to cathode drive across the load resistor regardless of where it floats. And it is floating wildly around at the cathode voltage due to the cathode winding (again AC wise).
To see that there is no CFB activity, one need only imagine that the cathode be shifted by some error voltage. Real CFB would notice the error and correct the drive signal level (grid to cathode). But here, the plate load resistor just shifts by the error voltage, and no correction voltage is generated across it to correct. Cathode swings are hence transparent to the driver, hence no CFB. The Mac CFB is an illusion.
I should clarify that this analysis was for a pentode driver, hence current drive to the load resistor. With a triode driver stage, there would be some NFBK due to cathode swing. But it is feedback to the driver stage then, not classic CFB in the output stage.
Don
😎
smoking-amp said:
If you do a voltage shift to the cathode it will be positive feedback, yes, but not 100% (the grid will not follow the cathode). How much positive feedback there will be depends on the driver's plate load resistors and the rp of the driver tubes. In a Mac circuit there will still be an overweight of negative feedback. 🙂
Jan E Veiset
OK, I am looking at the MC-60 schematic. The top of driver load resistor R17 does indeed follow the output cathode exactly (AC wise). If V3 were a high Zout pentode, the grid would follow the cathode exactly and there would be no CFB.
But, as you say, the drivers being triode will cause some neg. FDBK. But this is due to feedback in the driver tube itself that causes its plate to have an Rp resisting the output cathode voltage swing. Not what I would call classical CFB in the output tube. The actual resulting NFDBK here is still quite small in relation to what one would expect with a 50 % cathode winding for CFB. Most of the NFDBK in the Mac is due to the extra differential stage in front and resulting high global feedback.
The real design error here is having the positive feedback loop (which increases gain) enclose the same elements as the negative feedback loop does. Nothing is gained this way. In the Wolcott design, for example, a positive feedback loop is carefully kept around ONLY a very linear stage, while the negative FDBK loop encloses the output stage too.
To get the advantages of the 50% CFB windings, I would replace R17 and R18 with CCS's. Then the HUGE positive feedback loop is nullified, and the CFB winding actually works as NFDBK.
As the Mac is historically designed, I would liken it to designing the space shuttle, but only giving it an ounce of fuel. (so it can hop over an ant on the runway) The simple totem pole output solves the same problem as the Mac, costs half as much, and even significantly out-performs it.
Don🙂
But, as you say, the drivers being triode will cause some neg. FDBK. But this is due to feedback in the driver tube itself that causes its plate to have an Rp resisting the output cathode voltage swing. Not what I would call classical CFB in the output tube. The actual resulting NFDBK here is still quite small in relation to what one would expect with a 50 % cathode winding for CFB. Most of the NFDBK in the Mac is due to the extra differential stage in front and resulting high global feedback.
The real design error here is having the positive feedback loop (which increases gain) enclose the same elements as the negative feedback loop does. Nothing is gained this way. In the Wolcott design, for example, a positive feedback loop is carefully kept around ONLY a very linear stage, while the negative FDBK loop encloses the output stage too.
To get the advantages of the 50% CFB windings, I would replace R17 and R18 with CCS's. Then the HUGE positive feedback loop is nullified, and the CFB winding actually works as NFDBK.
As the Mac is historically designed, I would liken it to designing the space shuttle, but only giving it an ounce of fuel. (so it can hop over an ant on the runway) The simple totem pole output solves the same problem as the Mac, costs half as much, and even significantly out-performs it.
Don🙂
OK, I don't want to beat this issue to death. But I noticed in Norman Crowhurst's "High Power for the Twin-Coupled Amplifier" article in Radio-Electronics Oct, 1960 he does give a circuit (figure 6) that uses pentode (EF86) drivers. This does suffer exactly from the positive feedback problem I mentioned, ie., completely wiping out the CFB effect.
For the other design's using a triode driver I am in agreement with Jan's analysis of the small effects from the positive feedback.
However, I would like to point out that the small effects are essentially due to the triode wiping out the large potential gain in the first place (due to its low Rp), so there's really nothing left to lose anyway.
Sort of like having a car salesman show you a 300 cubic inch V8 sportster with a lawn mower carburator and a model airplane air filter. The sales guy then appologizes for the small 3 dB loss of power due to the tiny air filter. Well, if you are satisfied with only 20 HP from your V8, I guess that's OK. But I would fix it.
Don🙂
For the other design's using a triode driver I am in agreement with Jan's analysis of the small effects from the positive feedback.
However, I would like to point out that the small effects are essentially due to the triode wiping out the large potential gain in the first place (due to its low Rp), so there's really nothing left to lose anyway.
Sort of like having a car salesman show you a 300 cubic inch V8 sportster with a lawn mower carburator and a model airplane air filter. The sales guy then appologizes for the small 3 dB loss of power due to the tiny air filter. Well, if you are satisfied with only 20 HP from your V8, I guess that's OK. But I would fix it.
Don🙂
When I first saw Fig6, I mistook for plate to plate Schading.
And my reaction was to think this absurd cumulative NFB!
How would such an amp have any gain?
Then I see that this is bootstrap. PFB exactly the opposite
phase of Schade plate to plate. Makes far more sense now.
They have moved low voltage gain and low plate resistance
to the output stage to drive unknown reactive and electro-
mechanical load in class AB. And moved high voltage gain to
complimentary LTP class A driver stage, which only has to
deal now with predictable resistive loadings.
I would't say it "wipes out" the CFB effect. As the CFB effect
is dealing with load variations immediately as a local feedback.
Driver stage could offer no better than a long shifty global loop.
Crowhurst has shuffled some problems around to locations best
adapted to deal with them. Seems a very well reasoned trade.
And my reaction was to think this absurd cumulative NFB!
How would such an amp have any gain?
Then I see that this is bootstrap. PFB exactly the opposite
phase of Schade plate to plate. Makes far more sense now.
They have moved low voltage gain and low plate resistance
to the output stage to drive unknown reactive and electro-
mechanical load in class AB. And moved high voltage gain to
complimentary LTP class A driver stage, which only has to
deal now with predictable resistive loadings.
I would't say it "wipes out" the CFB effect. As the CFB effect
is dealing with load variations immediately as a local feedback.
Driver stage could offer no better than a long shifty global loop.
Crowhurst has shuffled some problems around to locations best
adapted to deal with them. Seems a very well reasoned trade.
Last edited:
In fig. 6, if the pentode driver had very high Rp, and with the output load changes reflected at the top of the R15/R16 driver load resistors, then, with only a current drive from the drivers, this voltage variation is immediately transferred to the output cathodes. So no CFB whatever. Then, as Jan pointed out, by making the driver have some finite Rp, the CFB effect can be restored, since that reduces the positive feedback thru the bootstraps. However, the effective forward gain of the driver also reduces at the same rate as Rp drops: Mu(driver) = Rp * gm. So whatever CFB NFDBK loop gain is generated, is also subtracted from the global loop at the same rate.
This might be fixable by replacing the R15/R16 driver load resistors by floating CCS's (these would break the positive feedback effectively). The bootstrap pathes would remain, to provide the necessary plate voltage swings (ie, keeping some voltage across the CCS's).
The drivers could remain as high Rp pentodes to maintain driver gain. So the global loop gain would be untouched, and the CFB loops would be totally functional.
Another possible weakness in the McIntosh approach is that the two cathode windings are not bifilar coupled, so there is no guarantee that the two output cathodes have equal coupling to the load. (a potential distortion source) Only a carefully constructed dual sectioned OT could guarantee this. (I don't know what Mac actually used) A conventional single bobbin concentric wind would not. (one winding has more leakage L than the other, and more shunt C)
The Twin Coupled approach actually provides the means to control this. Even if the two sides of each OT primary are not equally coupled to the load (ie, conventional single bobbin concentric winds), that can be fixed by flipping one of the twin's primaries. The capacitor couplings between the OTs (equivalent to the bifilar setup) then couples each output cathode to an OT side A and a side B, providing equalization. So I think the Twin can provide very good results on a budget.
I would probably build one if say a CXPP25-MS-2.5K (Edcor) were available. (needs 16 Ohm secondaries to get an 8 Ohm out using the Twin setup) But Schading will do the same thing with just one OT.
This might be fixable by replacing the R15/R16 driver load resistors by floating CCS's (these would break the positive feedback effectively). The bootstrap pathes would remain, to provide the necessary plate voltage swings (ie, keeping some voltage across the CCS's).
The drivers could remain as high Rp pentodes to maintain driver gain. So the global loop gain would be untouched, and the CFB loops would be totally functional.
Another possible weakness in the McIntosh approach is that the two cathode windings are not bifilar coupled, so there is no guarantee that the two output cathodes have equal coupling to the load. (a potential distortion source) Only a carefully constructed dual sectioned OT could guarantee this. (I don't know what Mac actually used) A conventional single bobbin concentric wind would not. (one winding has more leakage L than the other, and more shunt C)
The Twin Coupled approach actually provides the means to control this. Even if the two sides of each OT primary are not equally coupled to the load (ie, conventional single bobbin concentric winds), that can be fixed by flipping one of the twin's primaries. The capacitor couplings between the OTs (equivalent to the bifilar setup) then couples each output cathode to an OT side A and a side B, providing equalization. So I think the Twin can provide very good results on a budget.
I would probably build one if say a CXPP25-MS-2.5K (Edcor) were available. (needs 16 Ohm secondaries to get an 8 Ohm out using the Twin setup) But Schading will do the same thing with just one OT.
Last edited:
Smk, I see now you maths bout outplate2 = outcathode1 into drvplate1= no CFB...
Ashamed I could not grasp problem before now, even after you tried to describe.
Breaking PFB by throwing CCS in series with bootstraps could be one right answer.
Driver plate loads need a defined resistive bleed, to each other if nowhere else.
Ashamed I could not grasp problem before now, even after you tried to describe.
Breaking PFB by throwing CCS in series with bootstraps could be one right answer.
Driver plate loads need a defined resistive bleed, to each other if nowhere else.
Last edited:
- Status
- Not open for further replies.