R9 does set the bias but its better if you use the cathode resistor. The reason is the role D1 plays. In class AB it has very little effect. In class B however it clips the peak of the voltage. This reduces the bias drift which would otherwise occur.
I can send you the KiCad stuff for the bias board. That way if you want a PCB you can get one made by say JLCPCB. You could also build it on some stripboard.
I, in turn, apologize if I've expected too much. Just got back from the "SubFest" (*) at Klipsch and have gotten used to a camaraderie so rare these days.
OPT modelling is a 3D (at least) world in what we all pretend to be a 2D world. Some times, maybe enough times, it's close enough. But it's always a rough model needing rule-of-thumb nudging. No biggie, so's everything else that really matters. Everything we do is run in subroutines - just try talking by moving your vocal muscles (or whatever), can't work.
Practical OPT use comes from designing for some imagined "worst-case", and not from an imagined "usual" case. An audio transformer can never be perfect enough to ignore, but its parasitic imperfections can be wrapped into a model. Everything modulated by the iron core permeability is a third dimension, so we must treat it in our 2D world iteratively. The classical (practical design) solution is of course brute force and this fortunately coincides with an easily definable HP pole (somewhere else) significantly above the OPT/output valve source resistance paralleling OPT primary reflected.
As this RL product pole changes with signal level (because iron permeability changes, and both non linearly and non monotonically, and inductance varies linearly with permeability), so does stability margin.
So, even real world hardware measurements taken in the best of intentions are still subject to scrutiny. We can never see reality, but we can fight our damnedest to catch a peek
.
(*) This weekend at the Klipsch R&D rooms at their Hope, Arkansas plant I got to hear Roy Delgado's newest variant prototype subwoofers and the new Jubilees. Gotta say, my classical modeling of horn loading went out the window. Mind blown by amazing performance. It will be famous soon, but I was only reluctant to accept what i was hearing for less than a minute. I'm a Veteran but not a flag-waver, hear for yourself.
All good fortune,
Chris
OPT modelling is a 3D (at least) world in what we all pretend to be a 2D world. Some times, maybe enough times, it's close enough. But it's always a rough model needing rule-of-thumb nudging. No biggie, so's everything else that really matters. Everything we do is run in subroutines - just try talking by moving your vocal muscles (or whatever), can't work.
Practical OPT use comes from designing for some imagined "worst-case", and not from an imagined "usual" case. An audio transformer can never be perfect enough to ignore, but its parasitic imperfections can be wrapped into a model. Everything modulated by the iron core permeability is a third dimension, so we must treat it in our 2D world iteratively. The classical (practical design) solution is of course brute force and this fortunately coincides with an easily definable HP pole (somewhere else) significantly above the OPT/output valve source resistance paralleling OPT primary reflected.
As this RL product pole changes with signal level (because iron permeability changes, and both non linearly and non monotonically, and inductance varies linearly with permeability), so does stability margin.
So, even real world hardware measurements taken in the best of intentions are still subject to scrutiny. We can never see reality, but we can fight our damnedest to catch a peek
.
(*) This weekend at the Klipsch R&D rooms at their Hope, Arkansas plant I got to hear Roy Delgado's newest variant prototype subwoofers and the new Jubilees. Gotta say, my classical modeling of horn loading went out the window. Mind blown by amazing performance. It will be famous soon, but I was only reluctant to accept what i was hearing for less than a minute. I'm a Veteran but not a flag-waver, hear for yourself.
All good fortune,
Chris
I gave this a try this morning, a negative supply for the 6922, hoping to gain a little bit more output and reduced distortion. After trying different voltages and current settings it didn't prove worth it as I gained about 1/2 Watt and the distortion was worse in all cases.
Why not dc-couple the 1st stage and the LTP? One cap less on the signal path? It should be duable with the 6922 and EF86.
Well if you couple the first stage and the LPT you need a cap to ground on the other side of the LPT with a leak resistor. Actually that's still a HPF and yes one less cap in the signal path but still the same from a LF stability point of view.
I don't think you improve on that. I've started splitting R16 into 220+15 to ground adding a 22nF with the 15R and making a hum buck input. R18 C21 and the braid of the input coax return to the junction. Good for valve amps when you want to tie the chassis to ground but not the inputs. Try on ltspice with a common mode signal.
Think I would add grid stopper 1K on pin 7 U4 and a small resistor into the base of Q1. Both can go off at VHF. If your looking for more open loop gain R29 R35 47R or a 100R trim pot to get the DC balance.
Last edited:
Thanks, that's worth exploring indeed, as is your other suggestion in the above comments.
Agreed, but I personally prefer DC coupling between stages when possible, and it also removes the need for the LTP negative supply.
How would you approach that here? I’d be happy to try it out and measure the results on the prototype.
Give it ago you will need the plate voltage on U5 between 80-100V. When you do this if you keep the CCS watch the dissipation in Q1 or add a collector resistor to dissipate some power or don't bother with a CCS as the match will be better.
Something like this, sorry, I got a circuit that I already had, 1st stage triode instead of pentode, but the idea is the same. One of the drawbacks is that DC coupling restricts the freedom of choosing the LTP tube due to biasing needs, but the EF86+6922 should be fine.
In Schematic of Post # 164:
The 6922 cathodes each have 100 Ohm cathode resistors.
That raises the plate impedance by u x 100.
That means the plate impedance, rp, is raised by 33 x 100 = 3300 Ohms plate impedance.
The rp of 2600 Ohms, now becomes 2600 + 3300 Ohms = 5900 Ohms
Perhaps I am more worried about raising the plate impedance, than I am about a small amount of un-balanced DC current in the plate loads (select 6922 for good DC balance).
Try the simulation with the two 100 Ohm cathode resistors taken out (direct cathode to cathode connection), and then be sure to put 1k Ohm grid stoppers directly on the 6922 socket pins (6922 triodes will oscillate into the VHF regions if you give it the least of an excuse to do so).
Ask me how I know.
For the same reason, Post # 178 and Post # 179, I would not build it without two 1k Ohm grid stoppers directly at the 6922 socket pins.
The 6922 cathodes each have 100 Ohm cathode resistors.
That raises the plate impedance by u x 100.
That means the plate impedance, rp, is raised by 33 x 100 = 3300 Ohms plate impedance.
The rp of 2600 Ohms, now becomes 2600 + 3300 Ohms = 5900 Ohms
Perhaps I am more worried about raising the plate impedance, than I am about a small amount of un-balanced DC current in the plate loads (select 6922 for good DC balance).
Try the simulation with the two 100 Ohm cathode resistors taken out (direct cathode to cathode connection), and then be sure to put 1k Ohm grid stoppers directly on the 6922 socket pins (6922 triodes will oscillate into the VHF regions if you give it the least of an excuse to do so).
Ask me how I know.
For the same reason, Post # 178 and Post # 179, I would not build it without two 1k Ohm grid stoppers directly at the 6922 socket pins.
Last edited: