"quiet" +-15V (~0.5W) isolated supply?

[max_torque]

I have a project that takes a slowly moving analogue signal (say 10 Hz sine wave) and uses a high speed comparitor to produce a fast edge when that signal cross a certain threshold.  At the moment, the supply for the analogue front end uses a small off-the-shelf 1watt isolated DC/DC, (Traco TES1-0523). However, the common mode noise of this unit is causing significant ground bounce, and that is causing the output edge to jitter, depending on how the switching of that device hetrodynes with the incoming signal.  As i need low latency and not too much phase shift (output edge to input signal), i have a pretty fast comparitor (~10ns) and only a limited amount of low pass filtering in the signal path.  Looking at the output edge vs the input signal, i have around 300ns of jitter, depending when the DC/DC fires it's switches.  In the worst case, the comparitor triggers, and the ground bounce from the DC/DC cause it to un-trigger again a few ns later (i have a limited amount of hysterisis around the comparitor)

So,  what options do i have for a low noise +-15v supply?  I already have both differential and common mode filters on both sides of the DC/DC (as i suspected it was probably pretty noisy) and they are significantly reducing the conducted noise into the power and grounds, but not completely (obviously).

Are there any off-the-shelf devices to generate low noise rails at these voltages and powers? 

[max_torque]

PS, should have added, just in case it wasn't obvious, the supply the device is 5vdc, so i need to convert 5vdc to + and - 15vdc 

[Kleinstein]

Those normal DCDC converters are rather noisy, as they use a tiny transformer driven from a sharp edge rectangular drive. From my experience especially the common mode noise is a nasty problem and hard to filter if one does not have a good ground.

There is a chance a Royer converter could work, but I don't know of a source for ready made modules. So it would be about building your own. This may include winding your own small transformer. A mechanically slightly larger transformer can help to keep coupling capacitance small.

There are special chips (e.g. from LT) to drive small capacitors with not so sharp edges to produce less noise spikes.

[Wolfgang]

PS, should have added, just in case it wasn't obvious, the supply the device is 5vdc, so i need to convert 5vdc to + and - 15vdc 

Hi, I made some converters to power high sensistivity probes, like here:

https://electronicprojectsforfun.wordpress.com/power-supplies/a-usb-powered-15v-supply/

The trick is to make "slow" switchers.

[radar_macgyver]

Consider the LT3439, which has a slew-rate limited push pull driver, with LDOs on the outputs for regulation.

[David Hess]

An inverter using a pulse transformer will have much lower noise and might even use sine wave drive for even lower noise yet.

[T3sl4co1l]

How isolated do you need it?  That is, what common mode impedance (at what frequencies) could you tolerate?

A comprehensive answer to this question, actually, would solve your problem at the same time -- where is the common mode noise getting in, in the first place?  Answer: at whatever input connections have CM-diff conversion gain, or in-circuit nodes that are picking up ground loop voltage/current

Knowing a ballpark figure of those impedances and coupling factors, tells you exactly how much CM noise you can tolerate.  Knowing the input CM voltage/current (what does it connect to? how much does it care/matter?) also tells you how much impedance you can tolerate.

If the impedance can be quite low, past some point (typically, this arises when the isolation is galvanic only), then use a "Y type" cap across the isolation barrier, to short out the noise source.  Put a CMC in series with the DC-DC converter (one side or the other; 2-line chokes are much easier to find, so, that'd be the 5V side), and probably put a Y-cap between the converter's outputs directly.  So you get a C-L-C filter, in the common mode, confining that noise to the converter itself.  (A few extra bypasses and ferrite beads, on the local supplies, is usually warranted, too -- those things are usually nasty in both DM and CM.  You've probably already tried this; consider this a note for completeness sake.)

If you need high isolation at most frequencies*, then you at least need to minimize capacitance, which will probably eliminate commodity isolator modules, at least the cheapest ones.  Do consider the high voltage, low capacitance, reinforced or medical grade, types though -- they may offer value, even at the much higher part cost.  Don't underestimate the total cost of a custom solution!

*All frequencies, simply isn't possible of course.  There's always impedance somewhere, even if it's just self capacitance to free space (as would be the case for, say, a high-CMRR probe that's isolated with fiber optics).

If you do find you have to go with a custom solution, then consider CMCs again, this time as transformers.  The coupling factor is poor (typically 0.98), but it's adequate for this power level.  The capacitance and isolation voltage are great, and the price is phenomenal!

For a 5:+/-15V application, of course, the 1:1 ratio sucks.  You'd have to consider, say, a 5:15V boost first, followed by a current-protected chopper/inverter, followed by a full wave doubler.  All in all, it can be done in about a dozen transistors, or less with ICs, but probably not less than double the total PCB area of a commercial module/brick.  Hey, you're bound to pay a price somewhere, just be ready to deal with it, whenever and however you do...

Tim