With the RC compensation change, your diagram suggests removing the 33pF,1k2R between the C and b of Q3 and placing a 100pF 330R onto the base of Q8 to ground.
OK removed 33pF and 1k2R. Installed 100pF 330R between base of Q8 and force ground. Not much change. With Vin decoupling steady state oscillation, without it still oscillates but has more harmonic content. Oscillation sits right at 15MHz.
Lets see if its in the CFP. We got to find its origin. If you make the 9140's gate stopper 60R what happens? If you take the error amp signal connection directly to the 550's base beyond its base stopper what happens?
~25mVrms / 80mVpp. Not swear time, this is interesting. I only swear when magic smoke is let out.
I have switched probes to my better one and verified results. Darn.
How about dropping gate resistance to the irfp9140?
I have switched probes to my better one and verified results. Darn.
How about dropping gate resistance to the irfp9140?
Reducing gate stopper to 60R still oscillates at same frequency but Vpp is now 130mV and Vrms is 43mV. To me it means that source is before 9140 gate.
Will move error signal input directly to bc550 base now.
Will move error signal input directly to bc550 base now.
Left gate stopper at 60R and bypassed base stopper on bc550(Q8) slight gain in magnitude same freq. Now 140mVpp and 45mVrms.
Would like to put gate and base stopper back to original and increase base stopper on Q3 which is the bc546B.
My reasoning is that with the increasing maginitude of oscillation when we reduce the stopper values further up the chain it means the oscillation is before that part. If is was that part we would see a drastic change or no change.
Would like to put gate and base stopper back to original and increase base stopper on Q3 which is the bc546B.
My reasoning is that with the increasing maginitude of oscillation when we reduce the stopper values further up the chain it means the oscillation is before that part. If is was that part we would see a drastic change or no change.
Last edited:
I would try play the Mosfet's stopper value up too, afterwards. Some nH resonates with some base or gate. Could be originating in the error amp's layout too, you are right.
How does it show if your time scale is 10us? Does it change by clipping elsewhere the gnd crock of the probe? Because it looks more like noise now.
Get much of the same when moving the gnd of the probe around. Attached is the 50ns and 10us with 47R as base stopper on error amp.
I am wondering if I should move the gate stopper back to 120R on the 9140 to see if oscillation resides in error amp or shunt.
The attached look more like noise to me now.
I am wondering if I should move the gate stopper back to 120R on the 9140 to see if oscillation resides in error amp or shunt.
The attached look more like noise to me now.
Attachments
Last edited:
At 10us should be cool if it was something with the probing only. They look the same at both time settings. Like noise indeed. But playing with the stoppers you change things. From there you will see where it likely comes from.
Ooops, it may be working OK now and the noise is not the shunt. Because periods vanished. To be sure, a test is if it has something bad it will make the phono's FFT full of harmonic grass all over. If its as clean as with the other channel on original reg its not its own noise.
P.S. Catch you later.
P.S. Catch you later.
Good Night and Thanks. Need to clean up my mess now.
P.S. Couldn't resist but replace 330R gate stopper with 120R on 9140 and oscillation returned. Replaced back to 330R and oscillation stopped.
P.S. Couldn't resist but replace 330R gate stopper with 120R on 9140 and oscillation returned. Replaced back to 330R and oscillation stopped.
Last edited:
This is 9140 with almost double Crss than 9240, and your first iteration proprietary layout, so when it works, it works with what stoppers it needs. Maybe a 220R test at a point for the sake of it. It is not the phase margin bcs both doubling the Ccomp and changing the step network's place did not change it at all, nor the termination from film Zobel to 10x value lytic did. Actually it loses margin with more base and gate stopping. Some layout inductance talked to some base and gate and it had a certain period resonance its the most probable. Watch that FFT noise floor VS other channel with V1, its tell tale on a 58dB phono. If it will feed even 1-2mV of noise, the first stage cascode will catch it. No PSRR to speak off. I wonder where the scope gets that noise when the reg is powered off though. Can it be doing something with the bench DVM which is also across the output or is there some emitting equipment near? Also does it start fast now, so there is nothing left to see about in the main CCS part? Will be nice if its fully OK and you will not have to bother with layout experiments.
Help me explain this. That noise was read across a 5W 700R DALE wire wound resistor. If I read across a 510R 1/4 watt carbon or metal film I don't see the noise. Even with DVM attached.
The regulator picks up instantly when hooked up to the RIAA. It takes its time when hooked up to 700R resistor.
I will try 243R (since I have one) as gate stopper on other channel. If it works I will install in both channels. I ran a quick scan on the noise floor of the riaa and it was same as V1.0 (or very close). I will do a more detailed analysis later when I get my work bench cleaned up. It showed absolutley no 60,120,.... Hz peaks. I still have not gotten my measured noise floor to your levels though.
The regulator picks up instantly when hooked up to the RIAA. It takes its time when hooked up to 700R resistor.
I will try 243R (since I have one) as gate stopper on other channel. If it works I will install in both channels. I ran a quick scan on the noise floor of the riaa and it was same as V1.0 (or very close). I will do a more detailed analysis later when I get my work bench cleaned up. It showed absolutley no 60,120,.... Hz peaks. I still have not gotten my measured noise floor to your levels though.