D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based VReg

[tombo56]

Is your design stable both for 317 and 337 versions? Where is it?

Negative polarity dienoiser is not stable and can’t be made stable. Positive dienoiser is perfectly stable.
My design is listed in the index at the first post of this thread.

 

[Nick_D]

There might be a middle term: you can blunt the denoisator by reducing the gain of the correction. The simplest way to do it would be the addition of an emitter resistor to the transistor.
If the correction is reduced from 30dB to 15~20dB, it will still make useful job, but it should be more stable.
I didn't test it, but it should be done in the real world, because sims are useless for this kind of tweak.
Start with 150 ohm, and see what happens

I have just tried this with the standard denoiser, 337 application. Have built the denoiser with 337 on perfboard, and experienced the dreaded oscillations. Starting with 150 ohms, the oscillation was completely cured. I was able to lower this to 5 ohms, but opt for 10 ohms for safety. I also noted that the 100u output ground placement on the perfboard is sensative; tacking onto another component lead going to the perfboard ground bus garunteed instablility, direct routing lead to occasional stable operation prior to the emitter blunting resistor. FYI, this is an STMicro LM337SP, set to -25V, and no damping on the 22n cap.

 

[Elvee]

Thanks for your useful experimentations and results: I am sure that they will be useful for other DIYers

 

[Elvee]

With a 10 ohm emitter resistor, the performance reduction should not exceed 7~8dB, thus a small price to be paid for a greater peace of mind