Current rating of pin header?

[Faringdon]

Hi,
We are using this type of  pin header to carry 6A of current onto a little daughter PCB containing PFC sense resistors….

Pin header
https://uk.farnell.com/harwin/m20-9990845/header-1row-8way/dp/1022256

Its datasheet, says  each pin is rated for 3A of current. However,  that presumably means when mated with  a socket housing such as eg….

Socket housing
https://uk.farnell.com/harwin/m20-1060800/crimp-housing-8way/dp/865667?MER=sy-me-pd-mi-acce

When soldered direct from one PCB to the other, we presume the current rating per pin is far higher than 3A?

{EDIT: Please note we are not using the socket housing, we are soldering the pin header into PTH's at  either end. END-OF-EDIT}

[Terry Bites]

RTFM 3A is 3A. Why psuh your luck? It wont't melt at 6A but you will are risking a 20mV drop or 20mW of lost power. The mated value may be higher or lower in contact reistance.
6A .03R go figure.

[Jeff eelcr]

Soldered to PCB on both sides would be higher, by how much?
Test one connector terminal at 6 amps, it should not be hot
or get hot after time. If it does try 2 terminals.
Jeff

[Faringdon]

My apologies, i should have made clear that we are not using the socket housing, we will be soldering the pin header at  both ends, as Jeff eelcr kindly alludes.

[alm]

I'd find a table of wire sizes to max current based on temperature rise and voltage drop, and see what the max current would be for a wire of this surface area. How much voltage drop you can tolerate will depend on your application and the voltage. A 24 V rail is obviously less sensitive to voltage drop than a 1.8 V rail.

I'd think for the 10mm or so you could get away with more current, but I haven't done the math.

[Siwastaja]

2-3A is very typical maximum rating.

This is indeed related to the mating of the socket. The contacts are tiny and do not have much force.

But if you just use short pin header, pins directly soldered on PCB both ends, I see really no problem with 6A. You can easily qualify the heating by a simple test since the uncertainty of the contact resistance is removed.

There still is a risk that if you buy another brand, they use different alloy with worse electrical conductivity, but if the PCB-PCB separation is just a few millimeters, it's quite far-fetched. The pins are thick, after all, and the PCBs will act as heatsinks. I wouldn't be surprised if you could pull of something ridiculously high like 20-30A, given the PCB tracks can handle this!

[langwadt]

My apologies, i should have made clear that we are not using the socket housing, we will be soldering the pin header at  both ends, as Jeff eelcr kindly alludes.

then it just like a piece of wire, not a header/socket combination

[JustMeHere]

You can always use more than one pin to conduct the power.   You should derate.  If it says 3A then back off 20%. 

[eugene]

Bottom line physics is that the current carrying capacity of a conductor is purely a function of heat dissipation (IR heating due to the resistance of the conductor.) If you can remove the heat, then that pin header can conduct 10000A.

To put it differently, it's about temperature rise. The current ratings of different gauge wire is based entirely on the presumed allowable temperature rise. If the temperature of the pin when conducting 6A is acceptable to you, then the pin can handle 6A.

[Doctorandus_P]

I know it's an old topic...
I'm looking around for current ratings of 2.54mm headers & sockets too (which is of course how I found this).

Some datasheets state as low as 2A, most hover around 3A, and Samtec goes up to 6.4A, but only for very specific header and socket combinations, and for "single pins". I assume that when using multiple pins for higher current loads, the heat is more concentrated and the whole thing gets hotter.

[Siwastaja]

I know it's an old topic...
I'm looking around for current ratings of 2.54mm headers & sockets too (which is of course how I found this).

Some datasheets state as low as 2A, most hover around 3A, and Samtec goes up to 6.4A, but only for very specific header and socket combinations, and for "single pins". I assume that when using multiple pins for higher current loads, the heat is more concentrated and the whole thing gets hotter.

Yes, that is all correct, it is quite usual to see different ratings (or derating factors) when single pin vs. multiple pins carry the same current exactly because they heat up their neighbors.

Using a specialized socket with standard dimensions but more current capability carries a risk that someone does second-sourcing on that BOM item, but whether it's a problem depends on how well your BOM is being controlled.

Paralleling multiple pins for power is a good approach unless you are very space constrained, but if you are then connector selection for high current is quite tricky anyway.

[tooki]

Paralleling pins is something I’d do only as a last resort, to be avoided if possible. Using a connector that supports the required current on a single pin is preferable. Why? Because if one of the paralleled wires should break, the contact fail or get corroded, or anything else cause one of the contacts to stop carrying current, the remaining contacts will dutifully carry the failed contact’s current, potentially resulting in overload. Because the device doesn’t stop working, one doesn’t notice that it’s actually compromised.

(Kinda like an old incubator cabinet I fixed some months ago that had a bank of 10  fans, and nobody noticed the fans failing one by one over the years — not until the very last one failed, which is when we were asked to look at it!)

This isn’t a hypothetical problem: it’s suspected of being one of the causes of the smoldering power connectors on some recent nvidia graphics cards (Google “12VHPWR melting”.)

[Siwastaja]

Paralleling pins is something I’d do only as a last resort, to be avoided if possible. Using a connector that supports the required current on a single pin is preferable. Why? Because if one of the paralleled wires should break, the contact fail or get corroded, or anything else cause one of the contacts to stop carrying current, the remaining contacts will dutifully carry the failed contact’s current, potentially resulting in overload.

Interesting way to think, because I arrive at the complete opposite conclusion. If the single contact you are using fails - a typical failure mode is increased contact resistance after all - you are seeing the very same overload. Just with much much higher probability because all you have is the single pin.

Paralleling contacts averages out the risks of failure. If any contact increases its resistance, it still only sees the voltage limited by the drop over the other pins, contributing as much as it can but not overheating. Others are sharing the extra load.

So even if the design is marginal (meaning e.g. using three paralleled 2A pins to replace a single 6A pin), I'd still say it's more reliable. And paralleling pins offers you a attractive way of flexible derating. Use four paralleled 2A pins to replace a single 6A pin and now one can fail completely open; or two can fail partially to higher resistance; and the design is still within limits.

You see paralleled contacts all the time: in server power supplies, Schuko plug earth contacts, whatever. And I think it's for reliability reasons.

I can see how your logic works, but it would work in a model where (1) failure is a binary function of exceeding current rating (below: OK, above: fail), (2) failure mode is full open circuit. If that model was coupled with not providing any extra pins beyond the minimum calculated number, then surely one failing would cascade into all of them failing, and probability would be increased by the number of paralleled pins. But this model does not match the reality.

[MathWizard]

This is an old thread but I might as well ask here as it looks like the typical pin headers everyone uses. So what frequency are these good to ? Like if you have a few rows of them on a breakout board for an MCU, what sort of impedance do they look like as you go past a few MHz ?? Next time I have my sig gen going, I might compare a breadboard, to these, see how much capacitive coupling their is.

[m k]

I guess FastATA is pretty close to top.
But even faster connections used quite similar connectors.

[tooki]

This is an old thread but I might as well ask here as it looks like the typical pin headers everyone uses. So what frequency are these good to ? Like if you have a few rows of them on a breakout board for an MCU, what sort of impedance do they look like as you go past a few MHz ?? Next time I have my sig gen going, I might compare a breadboard, to these, see how much capacitive coupling their is.

They can handle much, much more then a few MHz.

Remember Ultra ATA/133, the last parallel ATA before we switched to SATA? It’s clocked at 66.5MHz (133MHz/2). Now, this required special cable-side connectors that connected 40 grounds within the female plug (so that every other wire in the ribbon cable is a ground), but the headers are bog-standard, and I don’t think the female contacts themselves were special in any way other than bridging the grounds.

And they routinely get used for the front-panel USB 2.0 jacks on many PCs, which are 480Mbps.

So really, it’s the cable that matters more. I wouldn’t use them for GHz, but they’re definitely good enough for quite substantial frequencies.

[Siwastaja]

Remember Ultra ATA/133, the last parallel ATA before we switched to SATA? It’s clocked at 66.5MHz (133MHz/2). Now, this required special cable-side connectors that connected 40 grounds within the female plug (so that every other wire in the ribbon cable is a ground), but the headers are bog-standard, and I don’t think the female contacts themselves were special in any way other than bridging the grounds.

And I think they would do a lot better if the pinout was redesigned. It's surprising it works so well with so crappy pinout with centimeters long return path to ground pins on the connector. But due to compatibility reasons the pinout was what is what and all they could do is to connect every other wire to those far-away ground pins.

If you use signal-ground-signal-ground arrangement it will have some characteristic impedance from the geometry and you could theoretically match that on the PCB, especially if you don't have to use an insulation displacement connector but PCB-to-PCB connections only.

[tooki]

Well sure, but my point is simply that they’re good for more than just a few MHz.

The only reason the internally-weird connector is needed is for backward compatibility. They could have accomplished the same performance (maybe better) with a standard connector with twice as many pins.

[m k]

I forgot those USB thingies.
Back in the day I was like "watta" when I saw them first time.

Those male connectors were almost exactly like LSI-11/23 serial ports, only color was different.
A bit different speed grade.

But are conducting materials today so much better?
Like purity, surface smoothness etc.

[tooki]

Those male connectors were almost exactly like LSI-11/23 serial ports, only color was different.
A bit different speed grade.

But are conducting materials today so much better?
Like purity, surface smoothness etc.

If anything, I’d suspect modern headers of being worse than older ones, as manufacturers look to cut costs.

Color doesn’t have any inherent meaning. Different manufacturers use different colors for different models sometimes.

I don’t think standard 0.1” pin headers come in “speed grades”. I’ve never encountered such a thing, and I’ve spent ungodly amounts of time reading connector specs. Certainly a PDP-11 serial port would not even begin to tickle the limits of a pin header, not by orders of magnitude.