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

The positive feedback version has a higher Vdrop across it, around 2.4-2.5V. If that isn't an issue then it seems to work fine.
The latest discrete version has a Vdrop of 0.5V across it and better performance.
I tested the latest one with a tda7297 amp and looked at the output with the 60dB LNA. I used the cap multiplier that I used on the previous LM317 boards. The cap multiplier uses a MJE15032 which is NPN and needs around 1.5V across it. But using a cap multiplier with the PNP version seems to require only 1V across it, so with the latest discrete design there would be a minimum of 1.5V across the whole thing. But it pushes to -200dB for PSRR this way.
I attached two photos, comparison between the first and latest discrete designs and latest discrete and LM317+dienoiser, 44.1kHz sample rate with full FFT resolution in ARTA, 1kHz sine from phone to tda7297 amp. Around 100mA total output current and about 11.25Vout. Cyan trace is the first discrete design, yellow is LM317+dienoiser and red is the latest discrete design. The signals are bit lower than the first discrete design and a bit higher than the LM317+dienoiser. Output impedance sits the same between them so this might explain it? All three had cap multiplier in circuit while measuring. You add -60dB to the scale on the left for the real values.
I think I need to name them both.

edit: I used 2SB649AC for pass transistor for this measurement, I have no MJE15033.
 

Attachments

  • lm317 vs discrete.png
    lm317 vs discrete.png
    77.9 KB · Views: 309
  • discrete vs discrete.png
    discrete vs discrete.png
    76.4 KB · Views: 309
Last edited:
Better tempco and higher current output without affecting the reference. This way there's a constant current through the reference.

I don't see any improvement here, the simulation looks the same regardless of whether it is a simple zener or GLED, GLED is designed for a voltage of 2.5V, for higher voltages there is another circuit with two transistors.
I recently did tests of WJ references with LEDs and if you follow the instructions they work great, much better than the LM329.
The picture shows a 19V 4A regulator that has the same input part as yours but on the bottom side of the pcb there is an AD825, GLED is a reference and is powered via CCS with 2SK246 and is set to a voltage of 6.9V, CCS holds current through GLED of about 3mA.
 

Attachments

  • gled6,8V.jpg
    gled6,8V.jpg
    388.2 KB · Views: 312
I don't see any improvement here, the simulation looks the same regardless of whether it is a simple zener or GLED, GLED is designed for a voltage of 2.5V, for higher voltages there is another circuit with two transistors.
I recently did tests of WJ references with LEDs and if you follow the instructions they work great, much better than the LM329.
The picture shows a 19V 4A regulator that has the same input part as yours but on the bottom side of the pcb there is an AD825, GLED is a reference and is powered via CCS with 2SK246 and is set to a voltage of 6.9V, CCS holds current through GLED of about 3mA.


Maybe not so obvious for Zeners, or higher voltage Zeners, but if you try it with LEDs for lower voltages it starts to help a lot. If you look closely there's some very small improvement for Zeners as well.
Also varying output current varies the current through the voltage reference. The GLED setup shunts the extra current and keeps the reference (LED/Zener/LM329) at a fixed current irrespective of the output current.
And lastly, for high output currents there's only a certain amount of mA that can run through some references before they burn out.
 
Here's a comparison between without and with the GLED setup.
First photo is tempco, second is comparison of current running through one of the LEDs from each setup, with a stepped AC current load, 10mA to 100mA.
 

Attachments

  • Screenshot from 2021-02-01 23-08-51.png
    Screenshot from 2021-02-01 23-08-51.png
    141.5 KB · Views: 379
  • Screenshot from 2021-02-01 23-06-23.png
    Screenshot from 2021-02-01 23-06-23.png
    142.5 KB · Views: 393
In the original circuit this is not a reference position at all. Try switching it to the other side to be powered from a regulated voltage and necessarily via CCS.
 

Attachments

  • reg 19V 4A.png
    reg 19V 4A.png
    46.7 KB · Views: 425
Last edited:
Yes but I used it like this as a minimum extra complication to the circuit with some extra performance and flexibility (higher output currents). I suppose it can be made more complicated with better performance, but for what it is and the fact that it can be used with simple LEDs for lower voltages for me seemed good enough.
The pcb that I made also has a capacitance multiplier, doesn't take much room on the pcb anyway. Output impedance is already very low, and noise floor is either at the limits or below my measuring setup.
The supply is meant as a sort of simple and cheap fully discrete supply that can be made with accessible components.
edit:
I don't know how it was used in the Super-Reg circuit. I took inspiration from this article:

https://refsnregs.waltjung.org/GLED...D Reference Cell _Walt's Blog 2014_092418.pdf
 
Yes but I used it like this as a minimum extra complication to the circuit with some extra performance and flexibility (higher output currents). I suppose it can be made more complicated with better performance, but for what it is and the fact that it can be used with simple LEDs for lower voltages for me seemed good enough.
The pcb that I made also has a capacitance multiplier, doesn't take much room on the pcb anyway. Output impedance is already very low, and noise floor is either at the limits or below my measuring setup.
The supply is meant as a sort of simple and cheap fully discrete supply that can be made with accessible components.
edit:
I don't know how it was used in the Super-Reg circuit. I took inspiration from this article:

https://refsnregs.waltjung.org/GLED...D Reference Cell _Walt's Blog 2014_092418.pdf

I don't know if it can be simpler than this, the AD825 can be replaced with the NE5534 or similar. It was tested on a current of 6A, the picture shows the first version with an output voltage of +/- 20V 4A in my A-class amplifier. The voltage drop is around 4V.
No problem with component selection, it works even with OS-CON and is stable.
I later made four new ones that I upgraded with Walt J. LED references and CCS.
 

Attachments

  • wj regulators 2A.jpg
    wj regulators 2A.jpg
    50.5 KB · Views: 200
  • IMG48.jpg
    IMG48.jpg
    61 KB · Views: 185
  • regulator.jpg
    regulator.jpg
    51 KB · Views: 265
  • regulator AD825.png
    regulator AD825.png
    106.5 KB · Views: 391
  • IMG3927.jpg
    IMG3927.jpg
    337.4 KB · Views: 214
  • 57.jpg
    57.jpg
    319.6 KB · Views: 202
Last edited:
With a NE5534 I get similar PSRR, and .noise sim shows a worse noise by 5 times compared to the discrete designs. For me it's not worth it.
I could get 3A or I think 5A, can't remember, out of the discrete designs in simulation. Some transistors have to be replaced with more powerful ones but I think I might be able to make it work.
I really like the idea of simple, small part count low tech with good performance. The two discrete regulators were made to show off the simple denoiser circuit performance.
 

Attachments

  • Screenshot from 2021-02-02 05-17-22.png
    Screenshot from 2021-02-02 05-17-22.png
    116.4 KB · Views: 165
Last edited:
Here's 4A at 30Vout. Stepped AC load 100mA to 4A. It needs around 1.3Vdrop min at 4A. Probably could be tweaked further.
edit: added output impedance graph. The sim results are different between this and resistive loads. Why?
 

Attachments

  • Screenshot from 2021-02-02 10-29-51.png
    Screenshot from 2021-02-02 10-29-51.png
    159.5 KB · Views: 150
  • Screenshot from 2021-02-02 10-30-38.png
    Screenshot from 2021-02-02 10-30-38.png
    145 KB · Views: 145
  • Screenshot from 2021-02-02 10-41-16.png
    Screenshot from 2021-02-02 10-41-16.png
    128.3 KB · Views: 117
Last edited: