Noisy phantom rail...?

[bcarso]

Yeah that did sound awfully high as a spec.

One thing to look at is correlation---adjust the scope to just barely trigger on the noise and crank out to faster horizontal sweeps and look for things oscillatory.

EDIT: I see early on in the thread that this was already examined. I'm beginning to side with the way-high out-of-spec noise contingent. What has/have the brand(s) been so far?

 

[JohnRoberts]

Thanks, I always wondered what % means... Mea culpa, I was rounding off, and doing the math in my head, so yes, it looks like I stopped thinking when I got an answer close to the result. Today my meat computer agrees with your math, .003% of 50V=< 2mV

While it may be a semantic distinction whether the high noise level is evidence of instability, it is clearly undesirable. I am suspicious that their data sheet is playing fast and loose with only a "typical" spec. I have run into noise issues with other regulator makers, that weren't so easily mitigated.

Without looking at an internal schematic of the 317 I can only imagine everything that's going on inside but they talk about ripple rejection being improved by decoupling at that adjustment port so apparently crap on the input couples through to there also, and is amplified by the ratio of resistors used to set output voltage (not unlike negative feedback in an opamp multiplies opamp input noise).

As others have noted, and Keith's experience confirms, that port needs to be cleanly decoupled to ground. Without cap decoupling I suspect not only the ratio but value of voltage setting resistor string matters very much.

JR

 

[Samuel Groner]

As others have noted, and Keith's experience confirms, that port needs to be cleanly decoupled to ground.

I think it might be some sort of positive feedback back to the input--adding the capacitor reduces noise gain to unity for high frequencies so cuts that path. I guess this is only a problem for high output voltages where noise gain is high as well. Just wonder why the TL783 appears to be perfectly immune to this?

Samuel

 

[JohnRoberts]

This is perhaps more like PSRR in an opamp...

The 783 looks similar. Like it was designed to drop in for 317, but perhaps improved. Is the pass element DMOS in the 317?

JR

 

[Guest]

Reference voltage source is noisy most probably.

 

[Samuel Groner]

This is perhaps more like PSRR in an opamp...

Well yes, the input is the power supply pin, so I actually mean PSRR I think...

Is the pass element DMOS in the 317?

Nop, it's BJT.

Samuel

 

[Svart]

I used a 1uf polyester cap to bypass the 8k to ground.. works like a charm. no more hash on the phantom. my 9k pres are now -82db quiet!

 

[Samuel Groner]

Just out of curiosity: which brand where those models that showed the instability problem?

Samuel

 

[SSLtech]

Oh boy... Motorola here, and I forget what the discarded one was...

Keith

 

[Svart]

I think mine was TI

 

[SSLtech]

Roger,

Thanks, I'm flattered!

That pic was just one that was online. It was the prototype of the 9K mic preamps (with top-groundplaned boards) and the prototype of the PSU board. Also the two meters in there are the prorotypes of the bargraph board that I did.

Nowadays I think I might use the pico board for any future projects, it's simpler to get that board than etch my own! -Maybe one day I'll make a dual-cascaded board for a 20-LED extended range version, unless you're planning something similar.

The current project is something which should put that one to shame... -if you thought THAT was good, this'll floor ya! :twisted:

:wink:

Keef

 

[JohnRoberts]

The classic conundrum in VR design is that you want low noise and low output impedance at all frequencies, but at the same time expect to terminate with as much capacitance and you can afford/fit. This is an obvious conflict with using lots of negative feedback to enhance that output characteristic as the capacitance will surely lead to phase shift and instability.

For this reason most VR exhibit rising output impedance (not unlike opamps). If interested it isn't that difficult to characterize the VR output impedance. I think I used something like a few hundred ohm resistor driven by a few volt sine wave (cap coupled), when I looked into this way back when. Note: the output Z can vary with load so consider nominal loading.

While the subject of this post was a fairly extreme case of noise, in general I would target getting the VR decent within the audio bandpass, and rely upon local decoupling closer to the ICs for HF impedance, and perhaps additional noise abatement if needed. Keep in mind PSRR of opamps is referenced to the IC device input, so noise gain or closed loop gain of that stage will impact sensitivity to PS noise. For a very high gain mic preamp this is more of a concern than for a unity gain stage.

JR

 

[JohnRoberts]

While, I have mentioned this before, back when I researched this I found the impedance characteristic of a 1,000uf aluminum electrolytic, nicely complemented the rising output impedance of 78/79 series 3-terminal VR. YMMV

JR