What kind of DC power supply has least noise

Question

I operate an MEG system (extremely sensitive magnetic sensors for measuring brain activity) and I recently found that the DC power source that feeds into our magnetic shielded room (power used for moving the MEG/chair/bed - thus required), produces 60 Hz line noise detected by the sensors. I'll detail my understanding the sources of this noise, but my question basically is: what kind of supply or what specific features should I look for to avoid this noise?

1. Noise from currents in wires. The current supply is a typical cheap switched-mode floating DC supply, 24V 0.5A. I did a few tests to confirm that although the difference between + and - looks "clean" and constant, each wire has a large voltage oscillation w.r.t. ground. And this "voltage" seems to be the result of a current source, so even with no load, there will be an oscillation of charge density in the wires, enough to generate a magnetic field detected by our system. (In the kOhm to MOhm range between + and G, I get on the order of 0.1 mA peak current.)

2. Noise from ground currents. Unfortunately, if I tie the negative to ground, of course the oscillation w.r.t. ground is gone, and that mostly gets rid of the detected noise, but the "offending" current is not eliminated, instead going through the ground. That also causes some noise, probably primarily because our shielded room is itself grounded. Worth mentioning that this noise is much smaller.

Edit: Price is not a primary concern. I realize a well designed low-noise supply would likely be much more expensive.

Answer 1

If you really got a cheap switching power supply, then most of the noise you're seeing is probably not due to a lack of clean regulation, but because the AC lines couple back into your supply lines after the switch mode supply; your observations under "1." match that. It's really hard to avoid that – only a very dedicated device design and excellent measurement shielding can elimate that.

However, there's a couple of things that are often done to overcome this:

1. Make sure you don't have any power lines running into your measurement chambers. That especially applies to ceiling lights etc.
2. Make your ground better. If your ground level starts oscillating when you connect your supply ground to it, you simply have a high impedance w.r.t. to "earth".

Generally, if you want to build a somewhat efficient, cheap, low-noise power supply, you're often in for a Switching-Mode Power supply, like the cheap one you're using, followed by a relatively large capacitor that guarantees that load changes and slight fluctuations don't matter that much, followed by a low-ESR capacitor (typically, ceramic), followed by a linear regulator, followed, again, by a buffer capacitor. That way, you can get the "non-energy-burning, fast-to-adapt-to-load-changes" behaviour of the SMPS, and the low-noise properties of a good linear regulator (it might be worth looking a minimal bit further than the usual, ancient, 780X and LM317 regulators – there are a few more modern linear regulators on the market, and some of them have better noise immunity).

Answer 2

The "offending" current that you mention is a leakage current inside your switching power supply. This parasitic current comes from capacitive coupling between primary and secondary winding of the high-frequency isolation transformer, and creates parasitic currents at AC mains frequency. There could be a smaller (intentional) DC coupling, to avoid total isolation from earth ground. This AC leakage will be on non-switcher PS as well, since there will be still a transformer in between. It is not possible to avoid this coupling, one can only reduce it by making special isolation transformers with grounded shield between the windings at the expense of efficiency. Funny, I just looked into specifications for so-called Medical Power supplies, they list the approved leakage at 0.5mA, five times bigger than you have measured.

To reduce the effect of leakage currents, you need to design a good grounding scheme for the entire setup, to separate return currents from power electronics from grounds on sensitive instruments, and to return the leakage current into ground before it enters your shielded room. This is always a challenging problem. Good luck.

Answer 3

I am going to take a slightly different approach to this.

Firstly, are you sure the pickup is magnetic? If so, then there are essentially two things you need to do:

1. Connect one side of the DC supply to the rooms shielding right where the cables pass into the room (Both legs of the supply cable must pass thru the same hole in the room shielding), this ensures that circulating ground currents due to the EMC cap in the supply stay outside the room.

2. All the internal wiring inside the room for stuff must be tightly twisted, this will ensure substantial field cancellation. Worst case use a group of four wires twisted around a common centroid (Used in microphone circuits and known as "Starquad" in that application), wiring is by paralleling opposite conductors, and it gets you maybe another 10 or 20dB of suppression.

If the pickup is E field rather then H, then screening the cables (and bonding the screen to the room shielding is an easy fix).

To answer the question asked, a linear supply built with a transformer having a grounded interwinding screen will be almost as quiet as a battery.

You might find that opening up your switcher and removing the cap across the isolation barrier, and then grounding one leg of the output to the room screening is useful, the supply will no longer meet EMC, but that may or may not matter to you.

Good luck, this stuff can be a bear to track down.

Answer 4

If you want lowest noise and are not too worried about cost ,size ,and losses then the linear supply is the best .They still make them at a price for lab use .If you want to make one yourself there is less to go wrong than with a switcher .If you still want to do a switcher then look at a soft switching topology or some resonant scheme .I did some Apples with Apples conducted EMC tests and found that my soft switch resonant circuit was no better around 150KHz but was much much better in the MHz range .The conducted EMC plot fell off fast and sudden with frequency like the Cliffs of Dover while the orthodox hard switched PWM was like a rolling hill .This implies that some switching loss reduction scheme will help you for radiated EMC .Remember that this is off course not as quiet as a linear ,and also note that many chinese power supplies have fraudulent EMC compliance stickers .Just test one and see!