Yeah, the 10K/200K improved many schemes (but not enough) until I lowered the VAS degeneration -- then 2K4/47K works better.
500R gate stopper on the VAS still oscillates, although if I go back to the BJT CCS then it's the stuttering oscillation instead of the constant fuzz or motorboating.
500R gate stopper on the VAS still oscillates, although if I go back to the BJT CCS then it's the stuttering oscillation instead of the constant fuzz or motorboating.
Slower cascodes (2N5087) make the gain curve nice and smooth (particularly with 10K/200K dividers), but the &*$*@ still oscillates.
No cascodes at all oscillates much later (making it through 4ms of nice clean signal).
Is there anything wrong with the LTP phase comp? It does seem to work....
No cascodes at all oscillates much later (making it through 4ms of nice clean signal).
Is there anything wrong with the LTP phase comp? It does seem to work....
Interestingly, with the LTP phase comp the slower cascodes make no difference -- which leads me to believe we're on to something there....
I think this is the crux of it:
Most of the schemes we've tried lowered, smoothed, or even flattened the gain spike. But they all leave the big inversion in the phase curve (even though it still stays between +40 and -120º).
The LTP phase comp kills the big inversion.
Most of the schemes we've tried lowered, smoothed, or even flattened the gain spike. But they all leave the big inversion in the phase curve (even though it still stays between +40 and -120º).
The LTP phase comp kills the big inversion.
Attachment to #788 looks like a closed loop gain plot. Its low frequency gain suspiciously smells like 26 dB = 20x = the feedback ratio. Stability analysis by Bode plots is conventionally performed using the open loop gain plot ... the one with >> 50dB of gain at 1 Hertz.
None of the suggestions between #772 and #788 made you happy before
... but ...
maybe you are thinking, perhaps I ought try them all again using the new analysis procedure?
My engineering tech used to say "even a BLIND pig finds an acorn, from time to time" and perhaps he meant, a random shot in the dark occasionally hits something worthwhile, if you make enough attempts.
... but ...
maybe you are thinking, perhaps I ought try them all again using the new analysis procedure?
My engineering tech used to say "even a BLIND pig finds an acorn, from time to time" and perhaps he meant, a random shot in the dark occasionally hits something worthwhile, if you make enough attempts.
He he... oddly it was a couple of those suggestions backwards (increasing the VAS degeneration slightly to 15R and decreasing the gate stopper to 47R) that helped.
Compensation is indeed a mystery to me, but I've got lots of patience with SPICE. 😎
Compensation is indeed a mystery to me, but I've got lots of patience with SPICE. 😎
Armed with my new tools I've massaged the compensation on all three front ends. Margins are up across the board, as are slew rates, and frequency response in the CFA case.
The Hitachi front end is really starting to strut its stuff driving laterals:
-3dB FR DC to 250KHz
160 V/usec slew rate
0.008% 20kTHD @ 100 watts
Remember when it was called BuzzBomb? It now has 70º phase margin and 15dB gain margin.
😀
The Hitachi front end is really starting to strut its stuff driving laterals:
-3dB FR DC to 250KHz
160 V/usec slew rate
0.008% 20kTHD @ 100 watts
Remember when it was called BuzzBomb? It now has 70º phase margin and 15dB gain margin.
😀
Jeff,
How good is the Hitachi front end driving a single pair of laterals with no drivers? Just curious,
Thanks.
Jam
How good is the Hitachi front end driving a single pair of laterals with no drivers? Just curious,
Thanks.
Jam
You'd need another current limiting strategy; even with 30V rails the flying catch diodes either clip too aggressively or don't keep the FETs from smoking with a short.
Ignoring that, I set the rails at 36V, dropped the drivers, flying catch diodes, and source resistors, reduced the currents in both differentials and set the bias at 110mA for 50W class A/B. It seems to perform admirably:
Ignoring that, I set the rails at 36V, dropped the drivers, flying catch diodes, and source resistors, reduced the currents in both differentials and set the bias at 110mA for 50W class A/B. It seems to perform admirably:
Thanks Jeff.
Just thinking about a pre-amp/ low powered amp, topology to use.
When will you be ordering boards to build the first unit?
Please let me know if there is anything I can help you with.🙂
Jam
Just thinking about a pre-amp/ low powered amp, topology to use.
When will you be ordering boards to build the first unit?
Please let me know if there is anything I can help you with.🙂
Jam
Last edited:
I made a change yesterday to better-correct some square-wave overshoot, so they're "settling" again. (I like to have them make it through a week or so with no changes before I declare them "done".)
After that the plan is to order a pair of the KISS boards first to test the regulators since they're shared between all the front-end boards.
After that the plan is to order a pair of the KISS boards first to test the regulators since they're shared between all the front-end boards.
While I let things settle, I've got a couple of questions about another topology. This is Greg Ball's SKA GB150:
Question 1:
What's the purpose of the bootstrap from the VAS output back to the LTP leg?
Question 2:
Why the bypass caps (C10, C11) in the LTP CCSs?
Question 1:
What's the purpose of the bootstrap from the VAS output back to the LTP leg?
Question 2:
Why the bypass caps (C10, C11) in the LTP CCSs?