Bricolo said:
No particular cct from my side - just the principal that the best place for high quality caps is to have them after the regulators rather than before.
The guys building gain clones discovered that if they use regulators then caps of 100uF or less are best after the regulators - and if they are good quality, low impedance then bypassing is not necessary. But this may vary with different circuits.
mike
Bricolo said:If only I knew the source of oscillation
Bricolo,
Your circuit has many places that need work.
The first thing is you should make the thing on a solid copper
ground plane. Uninterupted if possible. You are running pretty
high currents so BW will be very high.
OK secondly, the lower CCS needs a low Z electrolytic bypassing
the 3 x LEDs to gnd and then a 100R (maybe more) base resistor
to the lower BJT.
Upper CCS gets the same treatment but electro goes to V+
not gnd.
Next, the 2999R should not really be connected between the
2 super pairs, this will encourage some interaction between
the two. I would use separate R's to + and Gnd for biasing the
emitters of super pair BJT's.
Vref should be bypassed to gnd and OP load 999R should be
bypassed to gnd.
If you do all this it should be stable, especially with an
uninterupted gnd plane.
If it STILL oscillates there is one more option for last call.
You need to insert a damping resistor in the *collector*
of the super pair aux BJT, ie the one that runs low current and
feeds the base error current back to the emitter of main BJT.
This can be a bit of a problem in that there will be some voltage
drop across this R but even 50R to 75R may be enough. A ferrite
bead on the collector may also work.
Obviously + supply needs proper decoupling to gnd.
I've built quite a few of these type of circuits and always
get them stable.
Good luck.
Cheers,
Terry
Bricolo said:
OK I just checked out schematic on first post.
Similar tteatment applies. There is possibility for QS11 & 12 to
interact through miller C coupling. QS11 & 12 need 100R base
stoppers and green led needs a low Z electro bypass cap across
it.
R26 10K needs to be split into 2 separate R's to + and - supply
rails.
It may help to bypass 2 x red LEDs for ref to gnd with low Z
electro. Bypass green LED (D1) with same and 100R base
stopper for Q1.
The most important thing is good ground plane so all the caps
have a low Z current return at HF's.
As per my last post, the last option is small damper resistor in
collectors of Q2P and Q3N. Looking at the currents you are
running these at (1.5mA approx?) 100R will only drop 0.15 V and
so should work OK.
Check the data sheets of these BJT's (Q2/Q3) to see how much
voltage you need across C-E at 1.5mA.
Let me know how it goes.
Cheers,
Terry
peufeu said:
I have only listened to straight cascodes WRT I-V stages.
I have access to Audio Precision for short periods of time
so when I last had it I made up a few versions of I-V and
followers to log measurements for when I do a DAC
and will listen then. So I have made followers and I-V, got
them stable and measured them.
We did some listening to earlier versions of follower which
were bootstrapped but not strictly "super paired" which is
a form of bootsrapping. We found the bootstrapping can make
sound less natural.
So WRT I-V I will not be surprised if super pair has same effect
ie; less natural.
Also note that the only thing the super pair does is make
up for non linear HFE and miller C effects. These issues can be
addressed without using a super pair anyway... but we will see.
I currently have DAC proto 3/4 finished on bench that has
defeatable super pair. However we are currently doing class A
power amp and DAC will have to wait a little while.
Regards,
Terry
Terry, thank you for your help. I didn't have the time this week for further testing. I'll try with a double layer pcb when I can.
How can I know (from the scematic, parts selection...) if a design is more or less prone (or immune) to oscillations? Now that I have experimented this oscillating super pair, I'm wondering which circuits I can trust or not.
How can I know (from the scematic, parts selection...) if a design is more or less prone (or immune) to oscillations? Now that I have experimented this oscillating super pair, I'm wondering which circuits I can trust or not.
Bricolo said:
FWIW, all the circuits I tested were one sided copper covered
PCB with components Xmas treed on top. No breaks or cuts
in GP. And the current loops were kept as small as possible.
Why don't you use this method first, get the thing running,
stable, tweak it for best sound and THEN do a PCB.
Cheers,
Terry
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.