Tie SMPS neg to earth?

[staze]

All,

Hopefully an easy question. I'm building an enclosure for one of the FE-5680's, and I'm putting an SMPS in the enclosure that generates +12 and +5 (yes, the FE-5680a's will lock at 12V).

Anyway, I've put an IEC C13 jack on the enclosure, and I'm curious if I should ground the enclosure (it's just a hammond aluminum case). This would, in effect, tie the SMPS negative to ground since the FE-5680's have all their grounds tied together, and they would be connected to BNCs that are bolted to the enclosure.

Is there any harm in doing this? Is it (dis)advantageous for any reason?

Thanks!

[ConKbot]

A metal enclosure with mains in it should always be connected to mains earth. There can be some exceptions in some circumstances, but for a diy, definitely do. Use normal bnc's at first (shell directly connected to the metal case at first. If that proves problematic, use an insulated bnc so you can do a star ground in the case.

[staze]

That's what I figured. So, no issue with tying SMPS neg to ground?

[Richard Crowley]

Your question is over-simplified.  There are more factors than what you have disclosed here.
I would NOT connect the supply negative to ground at the power supply enclosure.
I would try to keep the "star ground point somewhere close to (or identical with) the power ground pin of the most critical module (likely your rubidium oscilator).
Easier to go back and ADD a ground connection to an isolated connector vs. isolating a hard-wired connection.

[staze]

Okay, this doesn't make any sense reading it.

I have an IEC connector with a ground connection. If I ground that to the case, that will ground the SMPS output negative rail to earth through the rubidium (since all of it's grounds are tied together internally). 

So, you're saying rather than using the chassis as my ground plane, I should run the ground from the IEC connector, as well as the negative from the SMPS, and the "ground" from the Rb outputs to a single star point?

I guess I should ask really, my concern is whether it is going to blow out the SMPS or if it's just going to cause poor grounding if I do chassis vs star ground?

[Richard Crowley]

I have an IEC connector with a ground connection. If I ground that to the case, that will ground the SMPS output negative rail to earth...

It will?  Really?  Most power supplies have floating outputs (both positive and negative) so that you have the independent choice of where or if either side gets grounded.

[max_torque]

1)  the metal case of your device really MUST be connected to mains power EARTH!

2) The outputs of your SMPS should really float with respect to that earth.

3) Add an "earth" terminal to the front of the unit, giving you +ve (usually red), -ve (usually black) and Earth (usually, green or yellow) terminals.  That way the user can decide if they want to couple earth to -ve.

4) you could put a high value resistor between Earth and -ve if you want to avoid the -ve line floating at high voltages away from mains earth potential due to static or ground loops etc.

Most PSUs will be laid out like this, i.e:

[Circlotron]

2) The outputs of your SMPS should really float with respect to that earth.

I spent 10 years working in a SMPS manufacturer's design lab. I saw supplies that had the output floating from earth, over a period of several days gradually charge up the EMC caps that go from either output to ground. They would charge from the HVDC side through the normal gigaohms insulation  resistance of the transformer. Could give you a suprising bite.

[dom0]

The outputs from many open-frame or closed-sheet-metal-box supplies are often not floating but internally connected to PE. HF-wise they are almost always connected to PE or phase/neutral. However, if you have high HF currents flowing at your supplies in regular operation, your design is not good and misses local supply capacitors at the HF loads.

From a DIN/IEC/ISO compliance perspective you'd need to connect the mains connectors PE tab directly and mechanically safely to the metal case.

[staze]

All... Sorry, this isn't helping.

This is not a PSU project, it's a Rb standard. I'll just ground the enclosure, which will ground the negative SMPS rail through the Rb standard. I'll probably try to ground it earlier rather than having that Rb be the star point but regardless, without isolating the Rb, and the BNC's, grounding the chassis will ground that negative rail of the SMPS.

I should have just sucked it up and used an external SMPS and lived with the frustration of a wall wart. :/

[dom0]

At medium frequencies it is usually the best option to isolate the BNCs from the case and shunt their ground to case on the BNC/case "junction" with a small capacitor (so that HF on the GND will be shunted to the case), while maintaining a nice low frequency star ground.

[staze]

Okay, that makes sense. I assume 10Mhz would be "medium" freq? Also, what value cap?

[dom0]

Yeah. Value is not so critical, say a 100 pF or so.

[codeboy2k]

At medium frequencies it is usually the best option to isolate the BNCs from the case and shunt their ground to case on the BNC/case "junction" with a small capacitor (so that HF on the GND will be shunted to the case), while maintaining a nice low frequency star ground.

I agree it's best practice to isolate the BNC's from the case, and shunt the shield GND's, but 100 pf seems too small to me.   That might shunt stuff up in the GHz regions really well, but not so good in UHF, VHF. If you have anything around you in the HF, VHF or UHF bands, you might want something else that has a lower frequency breakpoint.

I don't have personal experience with this, but I've also seen it done with a high-pass filter from BNC shield to case using 5k ohm || 100nF

If you want to maintain the best star ground, as has been said, then you could forget the resistor, and parallel two caps instead, maybe 100nF and 100pF to get good bypassing at a wider range of frequencies.

Ultimately, the choice really depends on what frequencies are around you that might get onto your shields 

[vk6zgo]

Okay, that makes sense. I assume 10Mhz would be "medium" freq? Also, what value cap?

No,10MHz is High Frequency-----these are nominal labels which have existed for decades & are still adhered to for consistency.

Medium frequency (MF) is roughly 500kHz to around 2MHz
You can find a more accurate listing by Googling.

[staze]

Okay, thanks. Though, not sure how that helps discussion. :/

[vk6zgo]

Okay, thanks. Though, not sure how that helps discussion. :/

It is pertinent to choice of capacitor values.
Note ,there was a typo in my original answer--bottom end of MF should be 500kHz.(as edited)

[staze]

That may be... But it doesn't offer an opinion on the choice of cap.

As you might be able to guess, I'm new to all of this, so input is useful here.

Anyway, ordered some isolated bnc bulkhead jacks on a slow boat from China. So, guess I'll work on something else for a bit.

Thanks for input all!

[dom0]

vk6zgo: Sorry, I wasn't aware that medium frequencies is actually a band designation, too (they are all named slightly different in German, e.g. Kurzwelle, so I don't know them all). I meant, well, medium frequencies, so from everything from a few hundred kHz to ~100 MHz, which is the typical range of signal frequencies found in not-so-high-speed systems like this one.

capacitor: I actually wanted to write "a couple 100 pF", not "a 100 pF". At 10 MHz I'd start maybe with 680 pF (because I have a bag with 500 of them) or 1 nF. If that'd suffices I'd not bother any further.

[Ian.M]

I'd be concerned that if you are triggering any digital circuit off the sine wave output of the rubidium oscillator, any mains hum added to the signal will shift the trigger point and introduce unwanted phase and duty cycle modulation.  Its hard to avoid ground loops if your source has a hard ground.

It might be better to ground the chassis, insulate the module from chassis, use insulated BNCs  and add a 1M resistor in parallel with a small capacitor from 0V to chassis at the SMPSU board to prevent static buildup and shunt to ground any HF leakage of the switching frequency due to interwinding capacitance of the transformer.