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

[Elvee]

And this is the more evolved, 154dB@100Hz variant. Some work will be necessary on the two compensation capacitors. Note that not all the protection components have to be present in all situations; depending on the context, some might be dispensed with (and it applies to the two variants), but as drawn they should be reasonably bullet-proof.

 

[Elvee]

One more remark: none of the circuits presented has been tested in reality, and the fact that they appear to work in sim says nothing of their real world performance.
Stability aspects in particular could prove especially tricky...

 

[jackinnj]

A looking at stability, switching V1 to DC 420V and hanging an AC1 current, I1 on the output (with 25mA DC load). This graph is from 10kHz to 10MHz. Looks pretty stable to me. (This is the non-invasive analysis method from Omicron-Lab, https://www.omicron-lab.com/applica...tional-and-non-invasive-stability-measurement )

The impedance troughs around 10 microOhms at ~100Hz which is impressive.

 

[bravi]

Eagerly waiting for this design to be finalised and tested.

 

[triode_al]

Why no Zener diode at the base of Q1? It will increase mains ripple rejection ("PSRR at low frequencies") by many decibels.

Or a CCS (JFET) in the place of the resistor R5. [I did that with following Darlington. Initially also intended as slow-start-up...] - I'm late to the party I see. I was asleep for a long time. Really.

 

[carlmart]

I have used a Dienoiser in the sensitive section of a frequency synthesizer, and I noticed that for 6V the values from the table published by Diego are not optimal.
This is the modified circuit:

View attachment 1295143

With the original values, the the Vce of Q1 was <1V, which is not very healthy: the transistor works in quasi-saturation mode, and the possible correction amplitude is limited.
With R3 at 470 ohm instead of 1K2 and R5 connected differently, the collector voltage is more reasonable, and the current through Q1 is higher, which is beneficial for the noise performance

Am I wrong or there's a slight displacement to the left and up, to the impedance curve in LTSpice after these changes are made to the original Diego's regulator? Psrr is not affected, and noise we can't get reliable results of.

 

[Elvee]

Yes, having the modifications slightly altering the behaviour is normal, and not really catastrophic. Better to have a circuit that works well in reality than a beautiful sim