i have a grand statement to make...

the ampeg svt, the mighty of mighties does NOT run in class ab2. in fact, i maintain that very little if ANY grid current flows in this design. it's very simple if you look at the schematic:

the grid stoppers on the output tubes, for both the 6550 and 6146, are on the order of 47k.

that's a mighty high value if you expect to pull current. for example, say you expect to run 5mA Ig1, that means the voltage drop across Rg1 will be .005 * 47000, or 235V. clearly, you are no longer pulling 5mA Ig1. you probably aren't pulling any Ig1 at all.

now, from a first-hand experience i know the importance of isolating the respective g1 circuits in a parallel output stage. yesterday i rigged up one channel of the BAGA with small 80uH 2R DCR chokes in parallel with the existing 3K grid stoppers. the hope was to preserve isolation at RF, but lower the DCR which was counter productive to grid current flow. (that means MO' POWA! it's a mantra of mine.)

so after tweaking away, i fired it up and it worked for a bit, but then the amp broke in to fierce oscillation, and i dove for the off switch. upon opening the chassis (today) i realized that the small chokes had shorted out, no doubt from exceeding their max current rating. they had essentially turned into dead shorts, of 0.1R DCR and ~100nH of L. there was a small subtle bubble on the side of each burnt choke, and three of them were toasted. that's what lead to the oscillations, essentially because when they shorted i then had two g1's directly connected together. bad news. luckily tubes are tough, especially kt90's, and i was fast on the off switch.

so out came the small chokes. i think the idea is sound, but the current rating of the little buggers is too low (it's something like 20-30mA, iirc) and the consequences of their shorting out is very high.

the current value of grid stoppers, which were simply paralleled by the chokes, is 3k. i did some quick calculations... assuming a (not entirely unreasonable) grid current of 5mA, that STILL drops 15V, which is enough to knock it out of grid current entirely. (really, it never ENTERS grid current, since as soon as it starts to flow the positive excursion is limited by the series impedance of 3k.) if you want to get into Ig1 with that kind of series R, then you've got to increase input Vac, but more on that later.

next, i went back and perused "an approach to audio frequency amplifier design" published by GEC and distributed by audio amateur press. at the back of the book is a description of a PPP kt88 amp good for about 450W. to make a long story short, i was looking to see what value of stopper they used to isolate g1's. GEC used 15k--that's high enough to stop Ig1 too, just like the SVT. they were driving their outputs with a interstage tranny coupled cathode follower, so low drive Z was not an issue in terms of stabilizing the output stage; they still relied on the high (at least by my standards) value of grid stopper.

soooo, i will have to be content with the 3k stoppers, which seem a good compromise. the amp was stable before, so i know they work at isolating the tubes from one another. maybe at some later date i will pursue the small chokes--or maybe not.

what i then decided to do was what i had been wanting to do for a while: pentode-wiring the cathode followers. clearly, if the CF's are triodes, then as the grid swings high the CF will run out of steam, since Vak decreases. (in fact, i actually pushed the CF's into grid current THEMSELVES, by attempting to pull the cathodes too high. i could tell this was happening by the all-too-familiar blocking distortion that ensued.) pentode CF's ain't gonna do this, since the screens will soar just as high as the cathodes do, maintaining a fixed Vs. the actual plate voltage, as pentodes, doesn't matter much at all. they've also got a higher gain (lower loss) of signal amplitude when compared to triode CF's. so that's what i did.

pentode connecting a CF is a bit of tricky business, since you've got to bypass the cathode to the screen. during power up and power down you can have some strange voltages, so a non-polar cap is best. the screen of a pentode is akin to the plate of a triode, and as such it has a low "plate" impedance. the quick'n'dirty method of determining the screen dynamic impedance is to take the screen mu (which in my case using sv83's is approximately 25) and divide it by the gm of the tube (in this case 15,000uS, or .015S). the result is 1.6K. the cap from cathode to screen must be sized with this dynamic impedance in mind... in PARALLEL with the "feed" resistor which supplies the screen's current. in my case, a 2uF 400VDC plastic cap did nicely, going down to about 50Hz. as the frequency drops below that the tube will gradually revert to triode operation.

while in there, i also paralleled each cap with a 200V zener diode, limiting the max Vs to 200V. this was done with an eye to the data sheets of the sv83. i suppose i could have gotten away with a lower voltage cap, but that's what i had on hand in a quantity of four.

when testing the performance of the new driver stage, i monitored the output of the CF on the scope while gradually increasing signal. peaks and valleys started to flatten at around 250VACpk. that means for a nominal bias voltage of ~-80VDC, i can expect to pull the CATHODES about 45V positive. that's enough to dump about 15mA of Ig1.

oh yeah, i guess the moral of the story is that if you really want to run grid current, do it WITHOUT paralleled output tubes. that way one could safely reduce/eliminate the stoppers and go crazy. i know there is SOME benefit to the DCR of a grid stopper (namely limiting grid dissipation, being easier on the tube), but i say crank the shit out of 'em.

one thing i'd still like to do is reroute some grounds, and move towards more of a star scheme. (it can never be quiet enough--dynamic range is indeed a battle fought on both ends of the spectrum.)

after that, i gotta button it up, lug the bastard back out to the garage, and shake some shit up.

kg