Rich -- With a little more time now, let me backtrack on statement I made in my last post, regarding the original 1.8K drain resistor: scratch the value I stated in that post.
OK. Based on the EFB(tm) model as published for late model 400 receiver operation, with revised parameters including a main B+ of 400 vdc, a screen voltage of 350 vdc, and the other basic parameters of the output stage in your application (load impedance of 6500 ohms, and ~ -17.25 vdc grid bias required for the quiescent operating point you are targeting), and based on the output of the EFB grid bias regulator driving the DC bias/balance circuit I published as installed in my Fisher 800C, here are the values you will need to revise the 400's EFB installation for operation in your 500B power amp project:
1. The 1.8K 5 watt resistor become a 500 Ohm 5 watt unit.
2. The 100K Gate resistor becomes 39K @ .5 watts.
3. The Gate 270K 1 watt resistor and .47 uF 600 volt cap remain unchanged.
4. As stated before, the 22 uf Source cap becomes a 500 volt device.
5. The 1 meg coupling resistor to the EFB grid bias regulator remains unchanged, as does the .22 uf cap located at the base of the inverter transistor.
6. The collector to base feedback resistor for the inverter transistor should become 75K @ .5 watt. Alternately, you can make this resistor a 68K in series with a 10K pot to provide a trim adjustment on the output of the grid bias regulator if so desired.
7. The 22K collector load resistor for the inverter transistor remains unchanged.
8. The 4.7 uF cap at the emitter of the buffer transistor remained unchanged as well.
9. The 10K resistor at the emitter of the buffer transistor is eliminated.
The SS devices remain the same as with the 400's installation. The power mosfet will need to be heat sunk to the chassis, dissipating just under 1 watt under all operating conditions (from quiescent, to full sustained power output in both channels). The specified part is a plastic piece, allowing for easy mounting to the chassis without the need for an insulating package. Just use some grease, and your good to go.
Also, unlike the 400s installation, in your application, you will supply the inverter stage B+ dropping resistor from the EFB controlled screen grid source. The new dropping resistor to supply the inverter stage B+ will then be a 5.6K @ .5 watt resistor. You will also definitely want to install 100 ohm screen stability resistors as well (not required with the EFB modified 400).
Kevin -- You have surmised the process well.
Specifically to the 7591 family of tubes, most designs (but not all) set their quiescent operating point to mimic the ideal operating points published by tube manufacturers. With a 6500 ohm load, many such designs employ a 450 volts plate, 400 volt screen, and 38 ma cathode current level (per tube) under quiescent conditions. The problem is, it all goes to pot as increasing power levels are developed, as the plate supply drops somewhat, the screen voltage -- being supplied by a simple dropping resistor -- falls through the basement, while the fixed grid bias voltage basically remains unchanged throughout all of it. Therefore, the ideal operating relationships set up under quiescent conditions gets obliterated as power is increase, resulting in greatly increased distortion, and reduced power output capability. If quiescent current is increased to combat the drop in screen voltage, then the excess current causes the tube to greatly exceed rated dissipation levels under quiescent conditions.
In the 500B example here, things are scaled back somewhat from the example above to 400 volts plate and 350 volts screen, but otherwise, the same train wreck happens as power is elevated.
Setting the parameters for EFB operation in this case however has largely already been done, as was published here:
http://www.audiokarma.org/forums/showthread.php?t=539265
With that work, it was determined that maximum performance is in fact achieved with a quiescent current of about 38 ma. What the addition of EFB did in that case is to cause the published performance to be achieved (that is, published by tube manufacturers), that would otherwise have required a quiescent current of over 50 ma per tube to be achieved with typical conventional installations. Or, it allows the original performance to be achieve at reduced quiescent current levels, allowing you to have your cake (the original unit's performance level), and eat it too (with cooler operating temps, longer tube life, etc.).
So for this application, the EFB screen regulator is basically set to scale the screen voltage from its original stock value, and then maintain that relative scaling under all operating conditions. Ditto for the EFB control grid regulator.
In contrast, you might review the thread wherein EFB operation was added to my late model Fisher 400 receiver. Briefly, in that application, the original screen voltage was wholly inappropriate for the operating conditions of the output stage, which included a rather strange move to a 10K plate load. Without any proper readjustment of the screen grid voltage -- or any real ability to do so economically at the time -- the output tube screen grids were really getting cooked bad in that application as power was increased. In that case then, I used the EFB circuitry to not only provide EFB action, but also to readjust the screen voltage to a more appropriate value for the plate load offered in that unit. The result was that the screens then operated as cool as a cucumber under full power conditions, power output was maintained to the same level as before, but the required quiescent current for the low distortion operating point dropped like a rock (to ~ 21 ma per tube) -- as did the distortion itself, while longer tube life and cooler operation were obvious results as well.
So the EFB circuits are doing double duty in that instance. As for determining the best parameters in that case, my Heath regulated bench power supply is my friend. While the Fisher chassis provided the plate voltage, the Heath -- having a variable B+ and C- outputs -- was used to provide the screen voltage and grid bias voltage. The amplifiers were driven with a traditional 1 kHz sine wave to within 1 db of full power output. At that point, the screen and bias voltages could be adjusted for optimum operation, based on the basic drop in B+ displayed by the "plate" supply in the Fisher chassis. With the drive signal removed, the resulting quiescent conditions could then be measured (the screen and bias voltages would be unchanged since the Heath supply is regulated) and checked against tube ratings, and the EFB circuits set up to produce that scaling action from the static B+ supply in the Fisher chassis.
I have also done the same exercise plotting the events with load lines, but frankly, I've become so proficient in obtaining the same results by using the Heath PS, that I find that approach quicker, and ultimately more accurate, since real world results are obtained. For me, I would need to employ the adjustable bench supply approach to verify my on-paper results anyway, so I find that it just saves time to go straight to the bench supply approach to begin with.
I hope this helps!
Dave