Noisy phantom rail...?

I've just built another of my SSLtech 18-0-18 +48V power supplies for a project, and I'm noticing something interesting that I haven't fathomed yet... -Maybe someone here has some wisdom to offer.

With the AC power applied to the primary of the transformer, I get a "HASH" white noise on the 48V line. When I switch the power off, it disappears IMMEDIATELY.

I don't know if it can be the (LM317) regulator, because it continues to supply power perfectly silently for a couple of seconds afterwards, before the voltage leaks down.

As soon as AC power is re-connected, the 'hash' is back. It's a fine, white noise, not dull or pink, but thin and white.

I'm going to try a few things like pulling the anti-reverse diode, or bypassing the asjustment pin to ground, but I really fail to see how those differentiate between reservoir caps discharging and ongoing-charge-from-AC conditions?

I feel like I'm missing something utterly fundamental here... anyone got any illuminative thoughts?

Of course, I can filter this down, but an ounce of prevention is worth a pound of cure...

Gratefully,

Keith

Oscillations?

Zooming in with the scope doesn't reveal any obvious repetitious oscillation component... just random hash.

Keith

what does the 18V+ look like?

I know that there's a lot of network stuff going over power these days, transmitting bursts of data when the AC crosses 0.

Could that kind of high frequency content could issues in your voltage multiplier circuit?

just a thought.

/R

This lunchtime, I'll pull the ±18V pridge and see what that does. Then I'll try snuffing the 48V bridge. Then I'll try a different bridge.

The 'bipolarization' of the 'floating' caps just makes me feel leery... the added caps should be more or less permanently reverse-polarised, since the existing caps are essentially correctly-biased about 95% of the time, as I read it. -None the less, if all else fails, my 'quest for understanding' will make me try what I've instinctively avoided until now. -My avoidance of it may be the exact reason that I don't understand it.

Keith

If you have an A-P, use AC Mains check mode and see what it has to say about THD on the power line--just to eliminate the hash-on-the-line possibility.

When you switch off the line, what precisely is getting switched? Are you disconnecting from the line completely, or?...

[quote author="bcarso"]When you switch off the line, what precisely is getting switched? Are you disconnecting from the line completely, or?...[/quote]
Complete disconnect.

In actual fact, I have a fetish for Neutrik 'Powercons' (like 'speakons' which are latching power connectors. Working live has taught me to curse IEC piwer connections...) and 'unlatching' the powercon disconnects both live and neutral, while leaving safety Ground connected.

The transformer is a twin-winding E/I transformer, which I prefer over overlaid-secondary toroids because it makes me feel that i should have a better chance of avoiding HF line dirt... -Whether or not that's realistic is a different matter! :wink:

It's one of these exact transformers in fact:


Keith[/img]

See what happens with different groundings of the transformer core. It does sound as if it's just line-borne RF garbage, and you are right that the separate bobbin E-I will be a better performer in that case than a toroid.

That is a head scratcher but the regulator is IMO not above suspicion. When you remove power, presumably it is hard pegged trying to regulate to a higher voltage than it can so in a different active mode than when voltage rails are adequate.

It's always something.

JR

[quote author="mediatechnology"]Also clip D1 for test. What if it was leaky and in rev breakdown?[/quote]
I did think about that, but the reservoir should still be 'reverse-biasing' it even after the AC stops pumping the usual 120 chugs-per-second into it...

Lunch was taken with a short-notice meeting, so I'll sneak a look at this later on.

Keith

Lightly loaded. -I could try upping that a bit to see if that's a factor.

Clipped the reverse-bias diode. Still noisy. Removed the ±18V bridge. Still noisy.

-The LM317 is coming in for some serious suspicion! :? -Funny thing is that I've always trusted these to be much less noise-prone than the LM340 series (78xx type)...

Keith

Is it's output shunted as close as possible to the chip by a tantalum cap?

Oh yes.

-Well, Ceramic, but good enough. 1/10th of an inch from the regulator -And I tried clipping tantalums on as well.

So, this is a mystery.

I tried swapping the LM317. Same thing. -Actually the new one was about half as noisy again!!!

So I tried bypass-shunting the 'V-set' resistor with an electrolytic. -INSTANT improvement by 95%. A simple R/C network on the output and it is now UTTERLY silent.

I've NEVER had to do that before. -I have NO idea what it was. I thought that the V-set resistor might have been noisy (bit of a long shot, but I've seen noisy resistors before...) but then I remembered that I'd at one point switched out the resistor for a 10K pot, to see if it was anything to do with regulation margin (i couldn't find the variac, and I figured that if I can't raise and lower the incoming raw DC, I'd raise and lower the required voltage... No difference AT ALL. Hissy at ALL settings, running into a fuzzy 120Hz dropout pattern if I asked it to regulate too high.

What was THAT all about?

The datasheet talks about bypassing that resistor ONLY in terms of optionally improving the ripple rejection; it says NOTHING of noise improvement. Since the two LM317s which I tried were -deliberately- from different manufacturers, I think I should add this cap as a matter of course from this point onwards. It also requires another 'charge protection' diode, so it's definately time for a board layout redesign... I always wanted to put those regulators in a line on the edge of the board anyhow... :grin:

Well, I'll go and look at ALL of the other 48V regulators that I've built before now... I reckon they will all end up getting some components tacked on to them... :wink:

Hmmm...

I usually have additional R/C on each phantom powered input, it also eliminates ground issues. Like 100 Ohm / 100 uF is fine.

[quote author="SSLtech"]Well, I'll go and look at ALL of the other 48V regulators that I've built before now... I reckon they will all end up getting some components tacked on to them... :wink:[/quote]

I think you can replace them with a new pcb... new layout obviously... :wink:

heh...

This is for GroupDIY freeware/shareware, so don't go eyeing no diamonds there, Boyo! :green:

Of course if everyone wants to chip in and contribute to Rochey's hunny-bunny fund, I have no problem with that, but for now I'll just post some gifs...

...when I get round to it! :grin:

Keith

i thought you'd appreciate that

:green: :green: :green:

Iechyd da!



Keef

yup.. VR are not exactly the top of the food chain for low noise design. Their list of priorities has quite a few ticks ahead of low noise.

When you removed power and unregulated rail decayed, the regulator output pegged full on and was quiet because it was no longer trying to regulate to it's noisy internal reference.

Its always something...I have also found cheap VR not very good for HF output Z due to lousy GBW internal circuitry.

JR