Low noise AC-DC, DC-DC switchers for mixed signal analog I/O?

[DarkMattrHole]

Hi all. So is anyone having any luck with sw. power PCB mount supply modules for mixed signal? I have been experimenting with little DC-DC +5, +-15 and the TDK-Lambda AC/+5/+-15 power module. I have a mixed signal design that has a small size and low weight factor in my design.

I have about 3 mV of clocking noise and >3mV spikes at the BNC analog output, but the BNCs are not on the PCB - more like a breadboard setup. I know it's theoretically possible to contain the spikes. I know i can do better than this with better lay out. So performance is close to required, but i'd like to cut the spikes down to less than a millivolt at the output BNC, so by a third more at least, to < 1mV. I think that may be possible with a redesign of the BNC grounding. Anyone had much luck with these modules? How about smaller scopes? Are they still primarily analog powered? Thanks for your opinion.

 

[Baluncore]

Mixed signals with an undefined layout does not make it possible to be guess wgere the problem lies. You must identify where the noise is coming from, and then why it is appearing at the output terminal.

If better layout is possible then you should move in that direction. A ground plane may be needed. You might start by lowering the ground impedance, while increasing the bypass capacitance of the supply to ground.

Look for ground loops and cases where power currents flow through analogue commons. Keep the power supply to the DC-DC converters separate from the analogue common.

Move the DC-DC converters from the analog board so as to reduce magnetic interference. Use linear voltage regulators for analog supplies and see if that removes the problem.

 

[essenmein]

The evidence is out there that you can have quiet operation with switching supplies. A quick peek at the back of our tek scopes tells me 45-440Hz, 100-240V operation, they are clearly SMPS.

Regarding your circuit, when you say "clocking noise" do you mean you see the switching edges of the converter or is it an actual clock from a micro?

Working through noise issues is basically a combination of trying to figure out if the noise is being measured correctly (ie is it real and not being picked up in the scope probe loop for example, are you using the right ground reference etc etc) and then if so, what is the nature of the noise (common mode or differential) and then what mechanism is causing it (di/dt, dv/dt, dB/dt etc), once you know these things the solution is usually pretty straight forward.

 

[berkeman]

Hi all. So is anyone having any luck with sw. power PCB mount supply modules for mixed signal?

I have used DC-DC converters (my own design instead of pre-packaged modules) for commercial radio designs (much more sensitive than just mixed-signal circuits), and have had success. But it took some learning of some tricks and techniques to get them to work really well.

First, you generally will want to follow the DC-DC output stage with an LDO linear regulator, to eliminate the switching noise and ripple. So choose a DC-DC output voltage that is a couple of volts higher than you need in your analog section, to give the LDO linear regulator headroom.

Second, if you have a digital and an analog section on the PCB, you should place the power supply circuit in the middle, and have separate power feed paths to each circuit. This minimizes the amount of shared ground impedance, which further helps to reduce the coupling of digital noise into the analog circuitry.

Finally, beware of magnetic field coupling from the DC-DC magnetics into your analog circuitry. If you have explicit magnetic components in the analog section (inductors, transformers, etc.), the coupling from any open magnetics in the DC-DC section can be significant. And even if you have no explicit magnetic components in the analog section, you still have to be careful with any loops in your traces and signal paths. Since I do my own DC-DC converters for these circuits, I can choose shielded inductors for the switching magnetics, or at the very least use mostly-self-shielding magnetics like pot cores for the switching inductors.

I once built a prototype for a biomedical customer that was in a small plastic enclosure with a top and bottom, where I had the power supply board attached to the bottom and the low-level analog card attached to the top. I built and debugged it with the package open, and then a couple days before I was to deliver it to my client, I closed up the package for final tests. Ack! It turned out that the potted pre-packaged DC-DC converter module (maybe from TDK or similar) used open magnetics, and the coupling into my analog circuitry destroyed my prototype's operation! I then spent the next 2 days figuring out how to shield that dang module's stray magnetic noise so that it would not interfere with my analog circuitry. For that prototype I didn't have time to re-design the power board with my own DC-DC converter and shielded magnetics, so the overall shield that I had to come up with was much bigger than I would have preferred.

For the final product, I was much more careful about my power supply design (no more pre-packaged converters), and I also tested it with the two halves of the package closed up much earlier in the design cycle! Lessons learned.

 

[jim hardy]

(no more pre-packaged converters)

They've bitten some of my colleagues.
It's not only their noise , the cheap electrolytics just don't last.

for a nuke plant your customers will really appreciate something like Sprague 135D



old jim

 

[DarkMattrHole]

The evidence is out there that you can have quiet operation with switching supplies. A quick peek at the back of our tek scopes tells me 45-440Hz, 100-240V operation, they are clearly SMPS.

Regarding your circuit, when you say "clocking noise" do you mean you see the switching edges of the converter or is it an actual clock from a micro?

Working through noise issues is basically a combination of trying to figure out if the noise is being measured correctly (ie is it real and not being picked up in the scope probe loop for example, are you using the right ground reference etc etc) and then if so, what is the nature of the noise (common mode or differential) and then what mechanism is causing it (di/dt, dv/dt, dB/dt etc), once you know these things the solution is usually pretty straight forward.

Hi. Thank you. The 'clock' noise I refer to is the switching frequency of the DC-DC modules. In the models I've tried are about 40KHz. switching rate - internally generated rate.

 

[DarkMattrHole]

Mixed signals with an undefined layout does not make it possible to be guess wgere the problem lies. You must identify where the noise is coming from, and then why it is appearing at the output terminal.

If better layout is possible then you should move in that direction. A ground plane may be needed. You might start by lowering the ground impedance, while increasing the bypass capacitance of the supply to ground.

Look for ground loops and cases where power currents flow through analogue commons. Keep the power supply to the DC-DC converters separate from the analogue common.

Move the DC-DC converters from the analog board so as to reduce magnetic interference. Use linear voltage regulators for analog supplies and see if that removes the problem.

Thanks. The modules are isolated DC-DC and the board power is from AC adapter. So the only place the analog and digital grounds actually meet is at the A/D/A converter, and no supply currents will return through this common ground, only signal I/O currents, theoretically. I read about people finding that such isolation techniques end up making digital and switcher noise and ECM worse, but i wonder if that's sometimes because of other factors not being addressed like common mode transmission of high frequencies and antenna affects.