The board has 22n+33R as compensation. I tested it a while ago just briefly so my results are not conclusive. And I believe the 317/337 PS is used for testing had somewhat suboptimal layout which may have influenced the results.
What I found out was that if the original 317/337 PS has lots of mains related crap on the output (e.g. due to poor layout) adding d(i)enoiser does not turn it into a SOTA ultra-low noise PS. It clearly lowers the overall noise level but does not do much about the mains crap. D(i)enoiser may have magic powers but it does not perform miracles
Based on this the approach of designing the whole PS from ground up may be a safer option.
At this stage I'm somewhat hesitant to post the gerber files since I would like to test it more. The BOM contains nothing fancy. I used BC817-40/807-40 BJTs. The SMD resistors and caps are 0805 size. All from Mouser.
You could try without the 33R
Yes, if the charging pulses of the reservoir caps stray into the "noble" tracks of the sensitive part, nothing can be done.What I found out was that if the original 317/337 PS has lots of mains related crap on the output (e.g. due to poor layout) adding d(i)enoiser does not turn it into a SOTA ultra-low noise PS. It clearly lowers the overall noise level but does not do much about the mains crap. D(i)enoiser may have magic powers but it does not perform miracles
In a good context, the denoiser provides 30 ~ 33dB improvement on all 3 key parameters: input noise, internal noise and output impedance.
I know this little pcb I designed from the Dienoisator goes in a different direction from what some members here are taking.
But it's an idea to simplify and cheaplify building these little add-ons.
Solid copper wires, isolated, would link the board to the proper points on the 317/337 regulators.
Please note the values are for a +/-30v supply, so some resistors should be adjusted following the published tables.
As you can see, the parts placing and traces are designed to follow the proto board existing pads. But it can be used to design a simple pcb. That I will soon do too.
But it's an idea to simplify and cheaplify building these little add-ons.
Solid copper wires, isolated, would link the board to the proper points on the 317/337 regulators.
Please note the values are for a +/-30v supply, so some resistors should be adjusted following the published tables.
As you can see, the parts placing and traces are designed to follow the proto board existing pads. But it can be used to design a simple pcb. That I will soon do too.
Attachments
Right now I am also testing the NCP59800 and we also have the same success (with 1 transistor) . Output noise is around 0.88uV (AP tested) with 5V noisy input (4mV) and 3.3V output
I have tried a dozen LDOs and those are the only ones that worked . Also the 2 transistor units were unstable (but I will do more testing)
So here it is folks , in the spirit of this thread and tireless work by Elvee and others . Thx you for a great (and cheaper) replacement of LT 3042
I have tried a dozen LDOs and those are the only ones that worked . Also the 2 transistor units were unstable (but I will do more testing)
So here it is folks , in the spirit of this thread and tireless work by Elvee and others . Thx you for a great (and cheaper) replacement of LT 3042
That's very crudely equivalent to 5.5nV/RtHz. There should be a lot of room for improvement. What transistor type did you use?
I’m happy to confirm that my single rail dienoiser design is working good. First tested deliberately without compensation capacitor at 60 mA, 120 mA and 11 to 17 V output voltage. Everything was stable and no oscillations spotted.
No min. load resistor was used as voltage divider and denoiser circuit draw around 10 mA, which is more than needed.
With compensation capacitor tested at 60 mA, 120 mA, 370 mA and 1,1 A. Pass OK.
I have no equipment for measuring noise levels but, based on Elvee’s measurements it should be around 2 µV with BC transistors.
There will be, very soon, second revision of PCB with pilot led at the output (noise proof), place for 18 mm diameter reservoir capacitors, some different pad dimensions or pitch and one good idea borrowed from Mark Johnson’s VRDN (you can always find some good ideas in his designs). Output capacitor will have in series position for ESR adjusting resistor or jumper. Therefore, it will be possible to use big capacitors or long life ones that usually have low impedance.
No min. load resistor was used as voltage divider and denoiser circuit draw around 10 mA, which is more than needed.
With compensation capacitor tested at 60 mA, 120 mA, 370 mA and 1,1 A. Pass OK.
I have no equipment for measuring noise levels but, based on Elvee’s measurements it should be around 2 µV with BC transistors.
There will be, very soon, second revision of PCB with pilot led at the output (noise proof), place for 18 mm diameter reservoir capacitors, some different pad dimensions or pitch and one good idea borrowed from Mark Johnson’s VRDN (you can always find some good ideas in his designs). Output capacitor will have in series position for ESR adjusting resistor or jumper. Therefore, it will be possible to use big capacitors or long life ones that usually have low impedance.
Attachments
Excellent Tombo.
Your board I ordered is still on it's way and JLC haven't quite finished my onboard design yet.
Worrying to hear the Sziklai pair aren't as stable with the LDO's. Hopefully not the case with the LM317. Either way, the second transistor can be easily jumpered.
I'm sure Diego will turn out with a solution to that apparent problem.
Maybe it's a problem with LDOs only, because Tombo didn't see any oscillations.
Are there any other BC transistors I can safely try until I finally buy a 337 or a ZTX?
I wanted to assemble the proto pcb and try it on the regulators I already have here. Those others I would have to wait till they get here.
Bohrok 2610 also experienced oscillations using the Sziklai pair on LM317.
My first test was with no compensation capacitor across transistors and yet, regulator was stable in varied voltage and load range (though not tested to full capacity) and no measured oscillation. With compensation capacitor, it should be very stable.
I wonder have cdsgames used in test new board with optimal layout or add on module. Maybe, Kelvin connection and overall layout has greater effect on circuit stability when using transistor pair?