Ultra Low Noise Lab Power Supply Extender for your existing lab supply

[branadic]

Hi Wolfgang,

I do know both websites and also made some measurements and simulation using ideas from both.

-branadic-

[Wolfgang]

Hi,

I just had a look at it, but one questions remains - the LM317 is a really bad regulator regarding noise - the performance would probably a lot better if you had used an LM723 with PNP pass transistor and a Wenzel cleanup shunt afterwards - did you try this one ?

Regards
   Wolfgang

[Andreas]

Hello Branadic,

I had a bad experience with 50 Hz hum.

I was measuring 1/f (0.1 - 10 Hz) noise as usual of a battery powered reference with my 4th order band pass filter cirquit.
The output looked quite plausible: around 4uVpp for a 10 V reference.
But the signal looked somewhat "fuzzy" against similar measurements.
So I did a FFT analysis which revealed a strong 50 Hz (and 150 Hz) signal.
When zooming in you can also see a 0.8uVpp 50 Hz amplitude in the raw signal.
(Ok I should have done a noise floor measurement before actually measuring noise.)

Root cause was a switched on HP34401A near my "cookies box (tinned can)" where I put in my references and the amplifier.
Usually I would think that a 6.5 digit DMM has enough shielding around the transformer but obviously this is not true.

Also other transformers I have to put away for more than 0.5m to get no influence above noise floor.

So I would use e.g. a 12V AC wall wart with enough distance to my low noise supply or split the design into 2 boxes.

https://de.rs-online.com/web/c/stromversorgungen-transformatoren/netzteile/steckernetzteile/?searchTerm=12VAC

with best regards

Andreas

[carl0s]

That's a curious ruler. I'm trying to decide if I could use one of those, or if a tape-measure is better.

[branadic]

Wolfgang, this is about a power supply extender with variable voltage setting based on TL3045, not LM317, LM723 or fixed voltage regulators. The measurements I performed in the linked thread was about LM317 and since I had the components and the noise amplifier laying around and was interested in the answers I simply did it.

Andreas, thanks for the input. I haven't observed 50Hz hum in my 0,1 - 10Hz noise measurement yet, maybe due to 14s record time? I activated FFT on the scope, but saw nothing, so I will have a look on that. But I can see mains line hum in my 10Hz - 100kHz measurement and will need to improve the build for stand-alone configuration with the transformer in the same case as the regulator board. I do have some rf sheet metal from former rf projects to create some case seperators.

About shielding of dc rail cables, what would you suggest? Twist positive and negative rails and add a shield connected to the case or would you go for coax? I was curious to see, that a high quality coax cable between dc output jacks and DSO picked up quite a lot of mains hum, while a twisted cable reduced this significant. So I tend to use twisted but shielded cables.

However for the lab supply extender you only have to keep away mains line from it.

-branadic-

[Andreas]

Hello Branadic,

I would go for twisted pair against 50 Hz (magnetic) hum.
The reason is simple: The shield of a coax is simply not thick enough for magnetic shielding. So twisting is the more effective measure.
Low frequent (magnetic) shielding is done by the skin effect: the field generates a current in the shield which produces a counter magnetic field.
But at 50 Hz and copper you will need many multiples of 9.4 mm as shield thickness to get a effective shielding.

https://www.adt-audio.de/ProAudio_Grundlagen/Skineffekt/Eindringtiefe.html

Of course a additional shield is not completely useless: it helps against capacitive coupling.

with best regards

Andreas

[Wolfgang]

Hi,

the plausible reason for the 50Hz is that the 34401A (and successors) are always averaging over an integer number of line cycles, so this influence is averaged out and a perfect transformer shielding is not needed. This is also found in the spec sheet of the meters.

If line interference is cumbersome, you could try:

- twisted wires so that magnetic influence cancels out (works better than coax)
- maximize distance between the source of magnetic fields and the sensitive circuitry
- you could optimize stray fields by changing transformer mounting and direction
- never load transformers to the max because stray fields rise when cores approach saturation
- use toroids instead of classic transformer cores
- check rectifiers for no snap effects (use schottky or soft-recovery, snubber circuits, ...)
- soft iron boxes (or mu-metal) can shield some magnetics fields quite well.

To branadic,

OK, the LT3045 is a regulator with impressive data, I never used one of them due to the max voltage being too low. This plus a shunt noise canceller could yield good results, why not. I can try when I am back home at my lab in September.

Much luck !

[branadic]

Twisted all cables and also added MKS2-100nF on the output jacks, this gave a slight improvement in noise, but not good enough.
A seperate compartment for keeping out 50Hz hum would require at least a 1.5mm steel case. The rf sheet metal I have is only 0.5mm, not the perfect choice. So currently I stuck and can't go on.

-branadic-

[Echo88]

The only cheap source of shielded transformers i was able to find were transformers for tube-amplifiers, for example: https://www.ebay.de/itm/Rohrengleichrichter-Netztrafo-355/123279036514?hash=item1cb400d062:g:3EgAAOSw7t1a-pQz

Can someone suggest a source for double shielded transformers,preferably toroids, like Henrik mentioned?

[branadic]

The only thing I know are R-core transformers, but can't say nothing about their performance.

-branadic-

[Messtechniker]

I'd put the transformer, then not necessarily a torroid, with mains input filter, bridge rectifier
and a large cap + a small cap and a bit of series inductance in a separate metal box you can keep nicely
far away (2 to 3 m or more, use shielded cable) from your low noise regulation box. Saves you
all that possibly expensive shielding effort which even so may be insufficient.

If still not satisfied you might try two back-to-back transformers with careful
passive filtering (100 Hz low pass) in between to get rid of high frequency mush
coming from the mains.

A power transformer with a shield between primary and secondary would be preferable,
but these are expensive custom designs* often for medicinal equipment.
And might be salvaged from such..... Good of a one off but not so good if others
want to build your circuit.

Just a few ideas.

*) just ask Pikatron or Haufe in Germany for a quotation

[branadic]

Thanks, but I will go a different way:
First, because it's relative easy for me to manage, make a steel tube (1.5 - 2mm thickness) with steel washers for both ends and magnetically encapsulate the toriod transformer. Yes I know, there are some more cables carrying the mains line hum, thus...
Second, design a steel case (1.5 - 2mm) for the regulator part, laser it, weld the single parts together (for such reasons I bought kweld) and install it inside the aluminium chassis.
Third, I already ordered R-core transformer for comparison and will measure differences in different orientations, which could also be interesting for further projects.
Normal transformers have the disadvantages that they are big and don't fit into my chassis.

This will take some time for the material to arrive, but it's worth learning.

-branadic-

[Wolfgang]

Just a thought - tinplate (Schubert) boxes could be easier to handle; and they have good RF shielding properties.

[branadic]

Just a thought - tinplate (Schubert) boxes could be easier to handle; and they have good RF shielding properties.

That's correct, but since it's only tin plated steel you need a certain thickness to surpress 50Hz/60Hz. Following the picture below this gives 2mm @ 100Hz.



The curve in the above diagram uses μr=460 for steel AISI 1010 and a specific resistance of 6.9e-7 ohm*m. Other formula indicate at least 1.5mm for ST37. Now you might understand why the cases of good instruments are that massive around the transformators.

-branadic-

[mzzj]

Just a thought - tinplate (Schubert) boxes could be easier to handle; and they have good RF shielding properties.

That's correct, but since it's only tin plated steel you need a certain thickness to surpress 50Hz/60Hz. Following the picture below this gives 2mm @ 100Hz.

https://rfantennas.files.wordpress.com/2013/12/eindringtiefe2.png
The curve in the above diagram uses μr=460 for steel AISI 1010 and a specific resistance of 6.9e-7 ohm*m. Other formula indicate at least 1.5mm for ST37. Now you might understand why the cases of good instruments are that massive around the transformators.

-branadic-

Wrong pic or you extrapolated from graph? Your graph shows 60 micrometers at 100 kilo-hz?

AFAIK its even worser than that when you are trying to suppress (magnetic) near-field:

[Wolfgang]

Hi branadic,

I am not sure if the Schubert boxes are really made of steel. To me it looks like a very soft iron alloy (Weißblech). They can make steel ones, too, but you have to order them explicitely. How well do your curves fit to soft-iron alloys ?

[branadic]

Correct, I extrapolated to 100Hz but also checked with a µr of 250 for steel. Maybe an optimistic assumtion, there is not much on the net about it. I don't want to surpress near field but low frequency from mains line.

As far as I know tinplate is just tin plated steel, nothing fancy and I couldn't find any hint on carefully treatment, which would indicate some special grain orientation of the steel.

Meanwhile I started to upgrade my 10Hz - 100kHz 80dB AN83 amplifier. Since now I had 330µV/10V SMD tantalum installed, which was fine for measuring voltage references. I now upgraded to 3x AVX Corporation FFB54D0117KJC, so that I'm able to measure over the full output voltage range of the power supply and up to 75V.

-branadic-

[Harm314]

Hi Branadic,

The power supply extender sounds interesting, but I'm not certain from your posts what you intend to deliver: PCB only, PCB + parts, or fully built module.

If I read the demo board sheet correctly it is built using a four-layer PCB. That is beyond my design skills.

[branadic]

One the other hand without the transformer it is a pretty nice extender for existing lab power supplies, forming them into a low noise supply for the bench. So this is the idea. Let me ask, is there any interest in a group buy of bare boards and the aluminium chassis parts? All other parts can be easily bought in the bay and at your local electronic store or distributor. I plan to redesign the board with additional current limit setting, which is currently missing.

Wrote the answer in the first post. It will be a dual layer board only. The aluminium chassis parts get cheaper the more people are coming together. But up to know only two persons are interested.

Back to the stand-alone solution:
Found some shield from industrial cables today. Which cables should be shielded? DC cables? AC cables? What to connect shield to? GND (for DC cables) or earth (for both)? Do we have some experts on that or is it just try and error by measuring the resulting noise?

-branadic-

[branadic]

Received some goodies today 

-branadic-

[BravoV]

Received some goodies today 

Whats the SCN wire stands for ?

Btw  where did you buy it ?

[Cody Turner OKC]

woah, intresting production date on those... hahah

[Gerhard_dk4xp]

That means 9th of April in Europe.

I'd like to know if there are more to be sourced?

I have just put an 650 VA isolation transformer into a 4 height units 19" box
with some outlets and lots of filters to feed my scope and a NGT20
power supply or two. There is no safety ground on the secondary,
but lots of ferrite. It seems, my worst ground loop problem is gone.

Cheers, Gerhard

[2N3055]

Received some goodies today 

Whats the SCN wire stands for ?

Btw  where did you buy it ?

That is the screen to minimize capacitive coupling.
I'm interested how this transformers will perform...

[branadic]

Received some goodies today 

Whats the SCN wire stands for ?

Btw  where did you buy it ?

SCN (green-yellow) means, as 2N3055 already mentioned, screen and has to be connected to earth.
It's already all there and linked in the #msg1740833

-branadic-