Whoops......
I love a good mystery, and this has become one. So despite what I should be doing with my time today, I've dug into this a little more, and found out some very interesting things -- and something I should have remembered early on....... Ah, the price of aging.........
First, in an effort to ensure that all last minute requests for clemency were covered for the Floating Paraphase inverter, I re-configured my outboard experimental inverter board into a Cathodyne inverter, to see if that would then force equal clipping. After all, Hafler's highly successful SCA/ST-35 amplifiers use that configuration with the very transformers of this project, and I've never recalled those designs displaying uneven clipping when matched tubes are installed. If the Cathodyne configuration produces even clipping, where the Floating Paraphase does not, then the Floating Paraphase inverter's fate is sealed.
The Cathodyne inverter was rigged up (with no NFB like before) and the amplifier driven to clipping, with the following results.....
Well, that theory got blown to Smitherines. So the Floating Paraphase design gets a last minute reprieve. And yet, it was still complicit in the crime, because it had previously been proven that when the drive impedance levels were made equal -- be they very low, or many thousands of Ohms, the clipping evens out. That left only one last test to make: Take the 9300 out of rotation, configure one of its
6BQ5/EFB/Z-565 output stages for UL operation, and drive it with the same Cathodyne inverter rigged up on the experimental board to see how it would perform at the onset of clipping. That produced the following results:
Technically not perfectly equal, as the bottom crest shows a fattening of the wave at this point before flattening out with further drive, whereas the top crest tends to flatten right away when entering clipping. But the important point is that both of these events begin at the same time -- and -- it's one heck of an improvement, and close enough that if
that had been the initial result, its doubtful that any of this exercise ever would have been pursued. So what the heck is going on?
If anything, this last test proved that there was more to what was going on than just the imbalance of drive impedance presented to a UL configured output stage, as originally thought. After all, the Cathodyne inverter itself (using 22K plate and cathode loads) produces an imbalanced drive impedance from its two outputs as well -- albeit certainly less of a difference than that of the Floating Paraphase design. But that reduced inequity seemed to make no difference in the clipping results produced when driving the 8800's output stage in place of the Floating Paraphase design. And yet, it produced by comparison excellent results when driving the 9300's output stage. With the significant difference shown between the two clipping scenarios then, something else clearly had to be going on -- and it was.
In my post #103, I muse that the issue may even be tube specific, given the significant difference in Gm produced between the two family of tubes involved. And that may in fact be involved. But of more significance is a peculiarity of the 6BQ5 family of tubes that the 6V6 family does not enjoy. I had recognized this long ago, but it got lost in the cobwebs of time with this project until this last test jogged my memory: The 6BQ5 family of tubes -- upon which virtually every small UL amplifier is based on -- will actually reach saturation current levels slightly
before Ecg = 0 is reached (and BTW: the EL34 acts the very same way). That means, that the 6BQ5 can reach clipping
without drawing any grid current. That then is the missing piece, because even with an unequal drive impedance presented, a UL output stage using this tube will have far less disparity in clipping between the two wave crests because no grid current is being drawn at the onset of clipping.
On the other hand, the 6V6 family of tubes enjoys no such advantage, meaning that in that tube, the onset of clipping coincides precisely with the onset of grid current, such that when configured for UL operation, and driven with an unequal drive impedance at the control grids, then unequal clipping will be the result.
So what has been learned from all of this? The take away points -- in no particular order -- would seem to include:
1. Push-pull Class AB1 operation of a UL output stage causes some tube families to be very sensitive to the equity of drive impedance presented to each output tube control grid as the onset of clipping is reached. This is because the inequity of control grid impedance in an otherwise balanced stage causes the screen grid Gm of each tube to differ, causing each tube to react differently to the UL signal presented there. All else being equal, the tube with the higher control grid drive impedance will clip first before the other tube in a UL push-pull pair.
2. In push-pull Class AB1 pentode mode, the screen grids of the output tubes are grounded with respect to AC, which doesn't make the control grids completely immune to drive impedance, but does act to balance the Gm of each tube, causing them to largely be insensitive to drive impedance at the control grids as the onset of clipping is reached.
3. The 6BQ5/EL84 (and 6CA7/EL34) family of tubes enjoy a unique advantage when operated in UL mode, because these tube families reach saturation current slightly before Ecg = 0 is reached. Therefore, in UL mode, they are rather insensitive to the drive impedance presented to each control grid as the onset of clipping is reached.
4. The 6V6/6L6/6550 family of tubes enjoy no such advantage, so that when operating in UL mode, precautions must be taken to ensure that the drive impedance presented to each control grid is nearly equal, thus ensuring that even clipping is achieved as the onset of that event is reached. With the larger tubes, a driver stage is often needed (or used) which will also fulfill the requirement of providing an equal drive impedance. If a phase inverter is to directly drive a UL output stage with these tube types, then a cathode-coupled type is recommended to minimize any drive impedance inequity.
The conclusions suggest then that the 9300 series Magnavox amplifiers are a good candidate for conversion to UL operation if so desired, but that the 175, 185, and 8800 series 6V6 based amplifiers are questionable in this regard. It also suggests that all of those who (for example) casually rotate 6L6 family tube types into Dynaco tapped screen operation amplifiers designed for use with EL34 tubes (think MK II, MK IV, Stereo 70, etc) are not only hurting performance due to the reflected impedance mismatch and improperly placed screen tap this combination creates, but are also further hurting it due to the compromised performance produced by these tubes as the onset of full power output is approached in such designs.
Finally, these findings further support the decision to finish out the project of this thread as a fixed bias pentode based amplifier.
As always, comments are welcomed.
Dave