Interpreting datasheets for class AB2

[cdemike]

Between what I'm learning here and hearing the results of an AB2 amp made by one of the members of the Telecaster forum (https://www.tdpri.com/threads/class-ab2 ... l.1108744/), I'm considering converting a test mule amp to experiment with AB2 myself. Something I've been thinking about in the process is whether it'd be possible to simplify the driver arrangement by directly coupling the phase inverter to the output grids. It seems to me that a fixed-bias LTPI's voltage references could be shifted in reference to ground so as to provide appropriate bias to the power tube grids via the phase inverter's plates. While doing so would involve a slightly busier power supply, that's already somewhat complicated when designing for AB2 anyway and it might save on real estate within the chassis, headache sourcing an appropriate interstage transformer, and need for additional triode drivers.

I drew up this draft schematic to explore the idea further. The amp I have in mind is a Hiwatt clone that I'm thinking would be a natural candidate since the phase inverter topology is already somewhat similar, so that was this draft schematic's starting place. As it stands now, the phase inverter drops about 80V across its plate load resistors, which should provide for bias voltage of around -40 to -50v, depending on how hot it gets biased, with a phase inverter node voltage of 33V. That voltage choice was based just on what higher-wattage zeners are readily available from AES, so not a really deep thought there. Does this design make any sense to y'all?

An alternate approach that seemed even weirder to my mind would be to just ground where the plate load resistors would usually connect to the B+ rail while keeping the negative voltage source instead of grounds on the cathodes. That was far enough away from anything I have seen before that I figured I must be missing something, though. Additionally, it may make more sense to just use equal plate load resistor values, but it seemed like that would likely mean a pretty trivial difference in bias voltage that could be compensated for by switching the positions of the output tubes if necessary.

Edit: I realized the draft schematic doesn't clarify the output tubes, but in light of what I've learned here, this would likely mean swapping 6L6s or KT66s for the EL34s in there now.

[Helmholtz]

Just a general remark:
The plates of your driver are high impedance sources, which means that the drive voltages will collapse and distort when grid current starts to flow.
That's why the driver typically is wired as cathode follower having low source impedance and allowing for a few mAs of grid/load current.

[cdemike]

Just a general remark:
The plates of your driver are high impedance sources, which means that the drive voltages will collapse and distort when grid current starts to flow.
That's why the driver typically is wired as cathode follower having low source impedance and allowing for a few mAs of grid/load current.

Thanks! I wasn't sure how low the threshold was for source impedance, but after reading your post it's making more sense why I'm mostly seeing 3 ways of approaching grid current-capable circuits: interstage transformers, cathode followers, and source followers. I did some more reading after thinking over your post and am seeing that some have said that even 12AX7 is a poor choice of driver given its comparatively high source impedance. So I'm gathering that in terms of more common dual triodes, 12AT7 and 12AU7 would be good while 12BH7 and ECC99 would be better. I'll take another look at my power transformer specs and see if I can think of something akin to Martin's bipolar SSS power supply that would work with what I have. I doubt my PT is stout enough to tolerate a power supply as radical as the Fenders mentioned above. Since my PT doesn't have a bias winding, I'm thinking a full-wave rectifier with a Zener diode could do the job with enough filtering to eliminate the Zener's noise. Seems I'll likely need to also think through a separate node for the drivers' anodes, which could also be used for the screens node. Another approach could be a MOSFET-regulated supply. Is noise the main reason I'm not often seeing Zeners in power supplies outside of efforts to lower B+?

I do have an extra ECC99 lying around, so I'll likely try to design a driver around it if a power supply seems feasible.

[martin manning]

...I'll take another look at my power transformer specs and see if I can think of something akin to Martin's bipolar SSS power supply that would work with what I have.

The bipolar supply running off the HV secondary bias tap works because the current demand from the 12AX7 driver is low. A bipolar supply would be a good choice, but if you want to increase current sinking capability I would go with a small auxiliary transformer with CT, and a FWB rectifier for the driver. Like this: https://ampgarage.com/forum/viewtopic.p ... 72#p361472

[Helmholtz]

Is noise the main reason I'm not often seeing Zeners in power supplies outside of efforts to lower B+?

The main reason to avoid zener (shunt) regulation typically is the rel. large additional zener current.

[pdf64]

Regulating the bias supply voltage seems counterproductive, unless the screen grid supply is too.
Allowing them both to track the mains supply reduces the impact of mains voltage variation on idle anode current and dissipation.

[cdemike]

I really appreciate y'all's patience with me while I'm learning more about this. After digging through some datasheets and drawing up some load lines, I drew up a more conventional driver arrangement centered around an ECC99. The amp in question has a Mojotone PT meant for a DR504 which is rated at 350-0-350 @ 300mA and 6.3-0-6.3 @ 7.5A. I suspect the HT primary rating is optimistic going off of my B+ being about 470V at idle, but it seems like it should be strong enough to support the beefier AB2 power supply. The heater supply should definitely be adequate (3x 12AX7 [0.3A each], 1x 12AT7 [0.3A], 1x ECC99 [0.8A], 2x KT66 [1.3A each] for a total of ~3.6A).

I'm coming to understand that the power supply to the driver needs to be able to supply a lot of current, so it shouldn't sag appreciably, hence the big filter values here, but I'm hoping to get input on whether this is still too anemic. If so, a second full wave rectifier using a resistive rather than capacitive load would provide ~320V. Maybe that would be a better approach? The resistor values I picked out here came from what seemed to me to be a reasonable load line (attached), which showed a quiescent current of 5.83mA per side with 12k cathode resistors. The AC load line (green) is borrowed from the 72V peak-to-peak figure from the Marconi KT66 datasheet.

Edit: the first draft didn't have grid leak resistors on the driver's inputs, which has been corrected.