Noise from switch mode power supply and ground loops

Question

I've got a piece of equipment that has pretty severe noise on it that I'm trying to improve. It's an older system (built in '96 but a lot of the design is probably late 80's) and is mostly bundles of hook up wire run through it instead of shielded cables with twisted pair wires. It's also largely an analog system, lots of sensors and boards with op amps. Depending on what/where I measure on the system I can get anywhere from ~300mV to ~1.8V of high frequency noise, which seems excessive.

It has a stainless steel enclosure that has DC power supplies, breakers, terminal blocks, some AC distribution. Connected to this is another stainless enclosure that houses a VMEBus computer with various cards that have all the digital/analog inputs/outputs. These enclosures are earth grounded (a 2" copper strap tied to building steel).

It has 2 modular DC power supplies (MeanWell MP650's), one large 24V and one that has 7 outputs (+5, +5, +12, -12, +15, -15, +24). All of the DC- terminals on the power supplies get tied to the enclosure, then most of the DC+ terminals go to breakers then terminal blocks. There's also DC- terminal blocks that are also tied to the enclosure.

The +5V supply that's used to power the VME has pretty severe high frequency noise on it, ~1V pk-pk at ~10MHz. I've got a scope pic attached. From testing that I've done turning various breakers and parts of the system off I think this line is the largest source of noise on the system.

The wiring for this supply is a little weird and I'm not sure what the best course of action would be. On the power supply DC- terminal there's one wire that connects to the enclosure and one that goes to the VME. The power supplies DC+ terminal has one wire that goes to the VME and one that goes to a breaker then a terminal block to go to a couple other devices. Now the VME also has two ground straps between the DC- and the enclosure. With this I've got unequal currents running through the two main lines that power the VME because half of it is returning via the stainless enclosure. These two lines are also bundled with ~20 other wires that go to various places.

I've got a diagram attached showing these connections and roughly how it's laid out in the enclosures. I'm thinking that I should remove the two ground straps on the VME use a twisted cable between it and the power supply. I would like to also add some sort of filter onto it, but I'm not sure what would make sense.

Edit: How can I reduce the high frequency noise on this system?

Answer 1

Your text refers to a direct connection between the PSU DC- and the chassis, but it is not shown in your diagram. Running any connection other than the earth through the chassis is very, very bad, especially if it is steel and not the more conductive aluminum. Another bad practice is to run the DC- a different route from the DC+, as this creates inductive loops which both emit and pick up interference. Worst of all is poking other wires through the loop, thus creating a little air-cored transformer.

You need to separate out the various DC- lines and the earth; run each separately. Run the -ve line next to the +ve line wherever possible. Use a star layout wherever possible, typically bringing everything back to the main distribution block. PCB and PSU earths should be wired back to a common "star earth" point.

They will need to connect together at least once, at some point, and you should check whether that is already done inside the PSU and/or on any of the circuit boards. If the 0 V lines are taken to earth in lots of places, you will have to take special care over the wiring runs.

The ripple on that +5 V line is probably originating from the VME system. 10 MHz is high for a switched-mode PSU of that era, but typical of the CPU clock rates you'd find inside the VME box. Fixing up the returns and grounding might well fix that, or you just might need to filter the line inside or close to the VME box, if that has not been done already.

Hope this gives some pointers, it's just standard wiring practice.