Multiple pins per signal on a connector

[mck1117]

I'm working on a project that needs a common "brain board" plugged in to a family of "motherboards".  Something in the style of Arduino shields, or the Raspberry Pi compute module.  As such, I need some sort of connector between the two boards.  In the range of 100-125 total pins, with most of them at very low currents (<10mA digital/analog IO), but a few of them (5-10 total) driving inductive and resistive loads in the 1-3 amp range (and grounding to support these - they're all low-side switched).

Current frontrunning options for the connector include traditional mezzanine connectors like IEEE 1386 (1A per pin rated), or a card edge connector like laptop memory SODIMM (like is used on the Raspberry Pi compute module, 0.4-0.6A per pin).  However, these connectors don't provide both the necessary density and necessary per-pin current capacity I need.  Ideally I would use a connector with a heterogeneous set of large and small pins, but these are pretty rare.

Is there anything wrong with doubling/tripling/quadrupling up pins on a connector in order to increase the current capacity? Are there any risks with this method?  Does it increase the rating to n * per-pin-rating, or should it be de-rated?

Thanks in advance for the help!

[Rerouter]

I will begin by warning the maximum current per pin, is generally BS marketing so that its the maximum current on any 1 pin assuming no other pins are used.

Some datasheets give you derating values and similar, plan on derating at least 30% less than the max rating, and make sure you don't have too crazy an amount of capacitance on any supply rails, every time they are connected and spark it really wears down the life

Equally almost every small signal connector your likely to find will have an awful connection cycles spec, most connectors can handle more, but they only garentee the other specs up to that number, under 100 cycles is not uncommon for spec sheets.

[mck1117]

I will begin by warning the maximum current per pin, is generally BS marketing so that its the maximum current on any 1 pin assuming no other pins are used.

I'd sort of figured this, as a DDR4 SODIMM has 260 pins rated at 0.5A each, and 130A total seems a little optimistic.

and make sure you don't have too crazy an amount of capacitance on any supply rails, every time they are connected and spark it really wears down the life

They shouldn't ever be connected/disconnected while under power, so I'm not worried about erosion from arcing.

[T3sl4co1l]

Yeah, scales up with derating.

Sometimes you'll see current ratings for N pins in use (selected values of N, or even all-inclusive), sometimes you'll see just one pin.  In the latter case, you really need to measure it yourself.

If you don't mind reliability or temperature rise much, you can go considerably over the rating, too.  It's a very wishy-washy thing, contact resistance is.  On the one hand, when you need reliability, you're hard pressed to get it; but when you don't, you can play fast and loose with it.  Not that we get many opportunities to play that game, though...

Reminder to leave plenty of grounds.  There's nothing worse than awful signal quality on a tons-of-signals connector -- even moreso when you've got a few amps of switched loads going through the same connector!

Grounds should be interleaved, for example ribbon cables are usually wired every other ground, which makes the two-row header have one row of all grounds.  Nice and simple.

Tim