This enclosure will rest on an ESD bench mat in a manufacturing station. It will have connectors in the back with signals, power and ground going to my PCB.
I am trying to decide whether or not I should deliberately connect the GND of my PCB to the chassis.
So, three questions:
Should I connect the GND of my PCB to the chassis? The chassis will definitely be connected to earth on an ESD mat.
If I do need to connect my chassis to the PCB’s GND, what’s the best way to do that? The enclosure is certainly designed for 75mm wide PCB’s, but there are no screw holes nor any obvious place to tap holes. It is intended to have PCB’s slid into it. I am assuming that I just need to find the symbol & footprint for a “strip” that runs along the edge of the PCB. Can’t find it, what is it called?
If I do not need to make this connection, is it “OK” to just rely on the PCB mask to avoid all electrical contact with the chassis? The board will be slid in once and almost never removed.
Personally, I’d avoid relying on an edge strip as a reliable connection to the chassis. Probably because of my training to always use a CGP (Central Grounding Post). What I’d do is attach a fastener to either the top or bottom of the chassis (the side with the largest clearance to the board) to use as a CGP. And run a wire from the board to the CGP.
If you need a flush surface on the top or bottom, use a flat-head screw with a countersink. The CGP should be positively attached to it’s mounting surface even without any wires attached, so you will need enough space between the inside of the mounting surface and your PCB (with components) for two nuts and a ring lug. (Use a ring lug on your grounding wire so the grounding doesn’t fall off if the nut gets loose.)
But, then, I was trained on medical devices… You may not need such a failure resistant solution.
EDIT:
To answer point 3, make sure you don’t have any electrically connected copper inside the slot of the chassis. Even though the board will only ever be inserted once, (HA!, never rely on that unless it is mandated that this device will be destroyed and disposed of if there is a fault. Even then, don’t rely on that…), the board will still slip and move microscopic amounts from movement and thermal cycles (as the chassis gets warm or cold with the room temperature, the metal will expand more than the PCB material so there will be movement). Soldermask is not designed for abrasion resistance. Anyone who has to bodge a wire onto a trace is thankful for that.
The answers to some of these will depend on what you are using it for. For some good references on hows, whys and whens you may want to check out Ott’s “Electromagnetic Compatibility Engineering”
Yes. Unless you shouldn’t. That is to say your default should always be to ground the enclosure. There are cases when you can’t or shouldn’t depending on your application but unless you specifically know you shouldn’t, then you should.
If you don’t, the next best thing is to have a “chassis” ground around the outside of your board and connect that to your signal ground at a fixed point.
“Best” way would be to have a set of metal screws and stand-off posts that can be hard-fixed to your board. This will provide solid, continuous contact that stands up to mild vibration. But if you’re just looking to avoid static, you can fill your top and bottom copper layers with GND and remove the solder mask along the edges. This is not a footprint but rather just a polygon on the solder mask layers.
This is the only question that has a definite answer: No. Never.
LOL. My medical devices training is rearing it’s ugly head again… I avoid using a mechanical connection as an electrical connection where I can. I understand that it is good enough for most applications, but my training makes me wary.
I would guess that housing is made from aluminium. Be aware that aluminium oxide is quite a good insulator. (well ok insulator is the wrong word. It does not really connect electricity as well as you would like but it also does not really insulate)
Some such enclosures have an artificially reinforced oxide layer (“eloxiert” in german. Not sure if i can trust the translation by google which is “anodized”.)
If you want to connect your ground then you need to ensure that there is a place where the oxide layer is replaced by something else. Most likely this will be a layer of chrome or something similar.
Or use a toothed “lock” washer. The teeth should bite through the oxide layer on untreated aluminium. May have to grind/sand off the anodized layer if one exists.
Good point! The oxide (rust) is not always visible. And it is often an insulator.
Emery cloth is a good tool to remove oxide from a metal.
The other issue with joining dissimilar metals is you need to join them using antioxidation grease; otherwise electrolysis will in time cause one of the two metals to dissolve or to oxidize, neither of which allows the junction to continue to serve as a conductor. See https://www.ehow.com/how_12178311_apply-antioxidants-aluminum-wiring.html for discussion of the details on how to do this.
SembazuruCDE is correct that proper grounding is done ONLY at one point.
Just a heads up if you depend on this procedure: Aluminum Oxide forms a 4nm insulative layer in about 100 picoseconds after you expose it to oxygen. It will happen too quick for you to avoid it. The tooth washer that @SembazuruCDE recommends will get through but you’ll need to prep the surface with an oxidation blocker to keep it from forming under the teeth over time.
I am going to connect the GND of the PCB to the chassis in the way that SembazuruCDE described. Will use a toothed washer, screw and nut (and swab it all with some DeoxIT prior to tightening down!).
I will still use the card-slot guides to hold the board inside the enclosure. Underneath the mask I have GND planes on both sides on the edges. Was thinking about masking-off the edges to expose the GND plane, but I’d like to follow the advice of keeping a single GND connection to the chassis. Now going to try pulling the GND planes in some more so that there’s no chance of electrical contact (even if mask rubs off) other than through the explicit screw connection.
