GND connections

[johndon2000]

Hi,

I am bench testing a cct board and the edge connector has 42 pins to GND (all distributed in random places over the connector!)...

I'm sure it's not necessary to connect all 42 pins to GND to simply test the board (for voltages, clocks, resets etc)...

I guess if I connect 10-15 pins to GND this should be sufficient, or is that a mistake?

If so, how many would be sufficient (as it is time consuming every time I test a different board!)

Thanks,

John.

[madires]

I think that you have to check that by trial and error :-) The GND pins/wires are purposely added to the design to deal with things like crosstalk and so on. It might work with a few GND pins connected but not very well and not being able to meet full specs.

[dfmischler]

Then again if it isn't plugged into the edge connector anyway it might not matter at all for checking voltages and clocks.  Try just hooking up power and one ground and check if you can see what you are looking for.  Or build a pre-wired test jig with a card edge connector if you have to test lots of identical boards.

[KJDS]

On a typical design, someone early on will decide that, say, an 84 pin connector will be used. There's 16 address lines, 16 data lines, four control lines and four power supplies. To handle the current each supply rail will use four pins, so that's 16 needed for the supplies. This gives a total in this example of 52 lines needed, leaving 32 spare. The supply lines will often be put on the same pins for a number of cards that could go on the same motherboard to prevent damage so the fixes their position. Another card may need a fifth supply, so that could mean four pins are left spare for that. So now we have 28 pins spare. We'll want a ground line between each data line to help shield it, so that takes 17 of our 28, leaving 11 lines that we can use for ground. Sometimes these are placed with care and consideration at the start of the design. That's usually the way, however it is very normal for the design to be tweaked a little, another couple of control lines are needed after prototype boards are made and they end up using the most convenient pins just to simplify the board relayout.

Thus what should have been, and indeed started out as, a sensible spacing of ground lines often ends up being a mess.

[ignator]

Are these pins the only way ground can be connected to the board? i.e. no standoffs that provide ground bonding, or edge plating that is in a metal card guide ....stuff that grounds the board.  They most likely are there to connect a real good ground to minimize the series inductor each pin to ground has intrinsically  (lot's of parallel ground connections).  Does this board have real high frequency clocks, as well digital hardware that switch parallel bus signals synchronized with a clock?  Fast edges, everything your trying to provide a good ground so it does not radiate (RF emissions), or have signals that get false clock noise switching.
If it runs with just one ground, and you are just playing with it, then no problem.

[johndon2000]

Are these pins the only way ground can be connected to the board? i.e. no standoffs that provide ground bonding, or edge plating that is in a metal card guide ....stuff that grounds the board.  They most likely are there to connect a real good ground to minimize the series inductor each pin to ground has intrinsically  (lot's of parallel ground connections).  Does this board have real high frequency clocks, as well digital hardware that switch parallel bus signals synchronized with a clock?  Fast edges, everything your trying to provide a good ground so it does not radiate (RF emissions), or have signals that get false clock noise switching.
If it runs with just one ground, and you are just playing with it, then no problem.

There is a chunky heatsink on the processor which is grounded... I guess I could just use that instead of any/ all the 42 off board GND pins?!

The two on-board XTALs are 16 and 66Mhz... but I am just playing with it (i.e. just performing basic checks on voltage/ clocks/ resets etc) so there shoudn't be any issues really.

I connected up 10 out of the 42 pins to GND and that seemed to work.

Thanks all