Need help to suppress interference in a power supply.

[sekhon]

Hello friends, I just modified 110~220v 12V 100 amp JD200 server power supply(from dell poweredge6800) for my high current needs.I had to reroute the AC input from front of the supply to the back.Its working good but i am having strange problem.It works fine during the day but sometimes it shuts down and error led lights up (LED displays over voltage,under voltage,over current or fan error)during night(after ~7-8pm).
I am not an expert and could be 100% wrong but suspecting that interference in the power line is max at that time and because i rerouted the power cord and its very close to the circuit board which monitors voltage and current errors and somehow its tripping the error.Please check the attached picture showing power cord runing close to the daughter board which monitors the voltage and current.
So could anybody please suggest me how to remove interference?Can i use a snapable ferrite core on the ac input cable?Or any other way?Thanks

 

[Skogsgurra]

It looks like the cable clamp (the one that holds cable down at entry point) is supposed to connect a shield to chassis/ground. If the cable is shielded, I think it is a good idea to restore that chassis connection by removing the outer insulation and clamp the shield firmly to chassis.

Otherwise, a ferrite usually works well. Make it large, put it outside the box and stick the whole cable through it several times. The effectiveness increases with number of turns squared. So, two turns are four times more effective than one turn. And three turns are nine times better.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[sekhon]

Thanks Skogsgurra for your reply!! That cable clamp was made by me and its connected to the chassis on the side by brass screw.Power cord is not shielded.So it means getting a larger ferrite core so that i can pass cable through it couple of times better than putting a split type snap-able ferrite core.Also due you think that the 2 copper bus bars i installed for output could be the cause of any problems?

Any spec recommendation for ferrite core?Asking because i searched for ferrite core at digikey.com and there are so many factors to choose like inductance factor,material etc.Could you please guide me?thanks

 

[zappedagain]

The metal case on your power entry module (PEM - the thingy the AC cable plugs into) suggests that it may be a filtered module. A filtered PEM needs a very good connection to the chassis ground (surface to surface contact, not that tiny edge by where the AC cable plugs in), otherwise it will not filter and will not stop interference from going out the cable. You'll either need to improve the clamping (better), or you might get away with adding a dedicated ground strap (18 AWG might work but I would go larger) from the PEM case (secured under your clamp will probably be adequate) to the chassis (good). See what the data sheet for your PEM recommends (best).

You also might try a higher quality AC cord if you are unsure of the one you are using..

Z

 

[sekhon]

Thanks for the reply zappedagain!

To hold the power entry module,the clamp you see is made out from aluminum and that clamp is connected to the chassis via 2 brass screws.Do you think i still need to try grounding the PEM?I will try to find better shielded cable as you suggested.

One more question, Is it my just assumption or its possible to have interference problems during night time when thousands of electronic devices are running?

 

[Comcokid]

I don't think that there can be enough noise on the AC power cord to create issues for the internal circuits. Power supplies are noisy to begin with and the internal circuits are designed for that.

Power supplies for computers sometimes are intended to have a minimum load applied when they are on. If you don't have enough load at night that might create an issue.

 

[itsmoked]

It could be real. There are lots of stepped regulators on neighborhood power poles. Often with times set at 7 or 8pm. It could be your actual power has a step discontinuity every day at that time you're noticing. During that time the voltage jumps up or down and it gets detected by an overly sensitive unit made slightly more sensitive by your mods.

I'd check your wall power closely and watch for a line voltage step change and see if it's related to your trips.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[VEBill]

The image provided seems to show signal wiring to or from the board indicated being 'dressed' right alongside the power cord. Zip tied to it!! That's never good practice.

Given the symptoms, those signal wires should be separated from the power cord.

Best solution is probably to redo your power cord routing, keeping it away from such signal wiring. Up against the chassis is often slightly better than through the middle.

 

[MacGyverS2000]

V, Those wires aren't zip-tied to the power cable, but the black surround of the two cables make it appear that way on first glance. The signal wires DO route next to each other for a short distance, but since everything looks like twisted pair, it is probably not an issue (who knows what CMRR is at the circuits those wires are plugged into).

Dan - Owner

http://www.Hi-TecDesigns.com

 

[VEBill]

@Dan. You're correct. I couldn't distinguish those on my smaller screen. Thank you for the correction.

Still, should be re-routed.

Especially given the symptoms and the reportedly one-and-only mod from the original design.

 

[zappedagain]

Yes, you need to ground the PEM for the internal filters to work properly. Threads are notoriously bad high frequency conductors, and the approx 0.020"x 1.000" surface area where the PEM contacts the chassis is inadequate. Both look like highly inductive paths and degrade the filter performance.

It appears there are tabs on the mounting edge of the PEM; that is typically where you need to make good contact and enclosures will be masked in the mate to that area when anodized or painted.

Z

 

[sekhon]

Thanks everybody for the replies.Those twisted wires are going to the switch(to turn supply on) and 2 potentiometers (one for voltage adjustment and second to control fan speed).The other black shrink tubed wire is going to the 3 LED's (one for power in,second for supply on and third for error)in the front.
Will try to ground PEM and reroute power cable and see how it goes.