BTW, it seems everyone always has their trimmer on the lower side of the bias spreader. I can't think of any reason it would matter, but then I've been wrong before.
I've moved mine to the top because it fits on the schematic better there. Any issues with that?
Cheers,
Jeff.
I've moved mine to the top because it fits on the schematic better there. Any issues with that?
Cheers,
Jeff.
I don't think the TL431 likes the large electrolytic either. (I'm guessing this is why some people have trouble getting the 'A' class parts to work.) It's there in the F4/BA designs to give a slow turn-on (as it's filled from the flying catch diodes).
Just leave for a little while and see what happens....
Jeff, indra makes an important point about the place to put the trim pot on the Vbe multiplier ...I once did some consultation for a company where the bias circuit failure mode was as described, can be expensive.🙂
Personally I feel you are overthinking the bias circuit requirements for a mosfet output stage. 🙂
Since bias circuits seem to be the topic, here is one option that works well.....
Jam
Jeff, indra makes an important point about the place to put the trim pot on the Vbe multiplier ...I once did some consultation for a company where the bias circuit failure mode was as described, can be expensive.🙂
Personally I feel you are overthinking the bias circuit requirements for a mosfet output stage. 🙂
Since bias circuits seem to be the topic, here is one option that works well.....
Jam
Last edited:
The TLA431 needs large caps for stability reasons.
Either large cap values or none, 1uf to 10uF is danger territory for TLA431, ie oscillation etc.
Last edited:
Cordell's book has a nice section on "bias spreader" circuits. Cross pollinate some of those, with the optoisolator type of DC-feedback bias networks found in the WHAMMY for example, and you might get an Optical Sziklai bias network with astonishingly low end-to-end impedance. Could be worth it to horse around with the idea.
At the risk of overthinking the overthinking, I had a noodle on this.
We don't really need temp compensation for MOSFETs, particularly laterals. But the BJT spreader still gives us low-noise adjustability, and with Hawksford compenstation allows us to linearise the bias for VAS current variation.
However, BuzzBomb's VAS is pretty constant-current, so Hawksford isn't really needed there. And since BuzzBomb is the most complicated of the three, there's definitely an opportunity for simplification.
SixPack's VAS is not constant-current, so Hawksford makes some sense there.
Beefcake still has that poor unloved TL431, but that's another story....
Cheers,
Jeff.
"TL431" might not be the best way to describe your idea. "Constant voltage independent of air temperature, heatsink temperature, or current" may be closer to the actual objective.
Many bias circuits rely on negative temperature coefficients (are made that way intentionally) to get the bias of the output stage to where it should be quickly in a stable manner. There are many examples of this in both first watt and pass labs products.
A biasing circuit with zero temp coefficient would be perfect for laterals.
Last edited:
I've been thinking about how to realise this collection, and have come up with the following plan:
Mount two sets of output boards on each heat-sink; one with laterals and one with verticals. Swap out the front-end boards, hooking them up to the appropriate output boards.
This means Beefcake and BuzzBomb need to agree on gate stopper values. Beefcake doesn't really seem to care, so that's not a problem.
However, while double-checking Beefcake's slew rate with a 1MHz signal the TL431 went into oscillation (this even with the 220u cap). Queue the last straw / last nail / whatever parable.
Mount two sets of output boards on each heat-sink; one with laterals and one with verticals. Swap out the front-end boards, hooking them up to the appropriate output boards.
This means Beefcake and BuzzBomb need to agree on gate stopper values. Beefcake doesn't really seem to care, so that's not a problem.
However, while double-checking Beefcake's slew rate with a 1MHz signal the TL431 went into oscillation (this even with the 220u cap). Queue the last straw / last nail / whatever parable.
Yeah bloody tl431.
Try changing the bias current in TL431, stability is very much dependent on bias of tl431 and cap values.
Or just measure what voltage you have across tl431 and replace with an appropriate zener combination.
Try changing the bias current in TL431, stability is very much dependent on bias of tl431 and cap values.
Or just measure what voltage you have across tl431 and replace with an appropriate zener combination.
I'd go with the Zener if I had bread-boarded it, but I don't trust SPICE that much.
I just reduced it to a known problem and dropped in BuzzBomb's bias spreader. I'd try out adjusting the TL431's bias current, but I've already redone the PCB and I'm not going back now. 😉
I just reduced it to a known problem and dropped in BuzzBomb's bias spreader. I'd try out adjusting the TL431's bias current, but I've already redone the PCB and I'm not going back now. 😉
Alternatively find the biggest caps laying around that will fit on the pcb, retest, then work down in cap values.
Eg Try 2200uF then 1000uF, then, 470uF etc
Or work going up putting caps in parallel one at a time.
This might be faster than playing around with bias currents.
Eg Try 2200uF then 1000uF, then, 470uF etc
Or work going up putting caps in parallel one at a time.
This might be faster than playing around with bias currents.
Last edited:
Going to a 47uF cap did fix it in SPICE, but I had lost confidence in the TL431 at that point. (God only knows if there is some other resonant frequency or output current at which it won't like the 47uF either.)
47uF is on the low side.
Depending on which TL431 device you have, you need 220uF or sometimes even greater for stability.
I once built a ring oscillator out of TL431s just to see how fast it can operate. Fun experiment, try it yourself one day when you've got a spare 30 minutes.
If you are scared to use the TL431 you can try an "amplified zener diode". There are many ways to do this but the one that is probably the least frightening, is to start with Figure 2.d. of the attachment in post #485, but then replace Q13 by a zener diode. Presto, a very very low impedance, oscillation free, bias generator whose output voltage is (Vzener + Vbe).
A bit of googling will probably reveal several other circuit ideas that all call themselves an Amplified Zener.
If you are scared to use the TL431 you can try an "amplified zener diode". There are many ways to do this but the one that is probably the least frightening, is to start with Figure 2.d. of the attachment in post #485, but then replace Q13 by a zener diode. Presto, a very very low impedance, oscillation free, bias generator whose output voltage is (Vzener + Vbe).
A bit of googling will probably reveal several other circuit ideas that all call themselves an Amplified Zener.