22
\$\begingroup\$

Say I'm looking for some power connectors for a design, the current requirements are quite high and although I could get a connector with contacts rated for the full current, they're so bulky that they're unattractive for my design.

Smaller power connectors are available, which could meet the current requirement if I bond a few circuits together to spread the load and would result in a much smaller overall solution with regards to board space used.

Is doing this bad practice? Is there anything I should watch out for when trying to do this?

\$\endgroup\$
5
  • 7
    \$\begingroup\$ It's fine as long as power is off during connection and disconnection when one contact will make or break before the other. \$\endgroup\$
    – Transistor
    Nov 3, 2021 at 9:41
  • 9
    \$\begingroup\$ Your average desktop computer has plenty of paralleled contacts in it. \$\endgroup\$ Nov 3, 2021 at 9:43
  • 6
    \$\begingroup\$ Strictly speaking - yes it is bad practice but often you have no other option. One obvious problem is that a bad connection on one of of the wires could lead to a fire hazard or at least unwanted cable heating. Regarding bulky connectors though... the laws of physics usually require that the cables are as similarly bulky as the connector. What connector do you have in mind, more exactly, and what currents and wire thickness are we talking about? \$\endgroup\$
    – Lundin
    Nov 3, 2021 at 10:02
  • \$\begingroup\$ Is your design divisible, so each power connector could power an independent section with no connection between them except for signaling and GND (meaning: no paralleling of input power sources)? Are we talking low voltage barrel connectors, or 120/230V AC mains? \$\endgroup\$ Nov 3, 2021 at 20:14
  • \$\begingroup\$ Thanks for all of these great answers everyone! I had a feeling it was OK but wanted to be sure of any Gotchas. \$\endgroup\$
    – VBwhatnow
    Nov 5, 2021 at 11:40

7 Answers 7

33
\$\begingroup\$

Using parallel pins of a connector or parallel wires in a cable to increase the current capability is standard practice. It has been done for a very long time, probably since the start of electrical wiring.

It's often done so that a multiway connector can use smaller pin/wire thicknesses suitable for the majority of the wires and just use a few pins/wires in parallel for the high-current power/ground connections. That happens in cars, for example.

Another is to allow the different current requirements of different power connections to be met by using different numbers of parallel pins/wires for each. That happens on a PC motherboard power connection, for example.

It is usually important and favoured to ensure that the low/zero voltage return path gets given as high or higher current capability than the power supply paths.

This is to ensure that high supplied current cannot cause excessive voltage drops in the return path, causing excessively different ground/0V voltages at connected items that require a common ground/0V. They'll be slightly different anyway but normally tolerably so.

Some systems perform a hot connection: a connection/disconnection while the connector wires are active. Some don't, only doing so when power is off. For the former, care needs to be taken about how pins gradually meet as only one connection of many may momentarily be made. But that's part of hot swap design which is a subject in itself.

\$\endgroup\$
5
  • 3
    \$\begingroup\$ Good point about needing to consider the return path also. \$\endgroup\$
    – SteveSh
    Nov 4, 2021 at 15:40
  • \$\begingroup\$ +1 The only problem with such an arrangement is that damage to any of the individual conductors can easily push the rest of them into overcurrent without otherwise looking like anything is wrong. Additional fault detection/protection may be required depending on the application. \$\endgroup\$
    – J...
    Nov 5, 2021 at 17:32
  • 2
    \$\begingroup\$ Maybe. But that's no different than damage (such as a nick) to a single large conductor, that would cause current crowding and create a local hot spot. \$\endgroup\$
    – SteveSh
    Nov 5, 2021 at 19:18
  • \$\begingroup\$ @SteveSh True, and in a lot of applications it's treated that way. It does constitute a higher number of individual points of failure, though, so just by probabilities it's more likely to suffer a fault without particular attention to QC in any such assembly. \$\endgroup\$
    – J...
    Nov 5, 2021 at 23:28
  • 2
    \$\begingroup\$ While there may be a greater chance of a failure with many parallel connections, the power delivery system as a whole is more reliable because the other pins can pick up the current that was carried by the bad connection. That's one of the reasons for the 50% pin derating plus the added 1.25X margin. While the serial MTBF with multiple pins may, on paper look worse than that of a single pin, the system reliability is a whole lot better. \$\endgroup\$
    – SteveSh
    Nov 8, 2021 at 18:32
19
\$\begingroup\$

That's done all the time, even for hi-rel systems (spacecraft).

Here's one way that's used. To determine the number of pins needed, you take the derated current carrying capacity of a single pin (usually half the max current rating), divide that into the max current you have to carry. Then multiply by an additional derating factor, typically 1.25 and round up to get the total number of pins needed.

Here's an example-

  1. Current to be carried = 10 amps
  2. Max current of a single pin = 2 amps
  3. Derated current capacity for each pin = 1.0 amps
  4. (10 A)/(1 A) = 10 pins
  5. 10 pins x 1.25 = 12.5 pins
  6. Round up => 13 pins

Also look at the discussion here: Stacked PCB connections, split up current?

\$\endgroup\$
6
\$\begingroup\$

A nice example of how smaller contacts can be used with high currents is the SATA connector: its power part has 15 smaller pads which are used to provide 4 high-current connections (GND, 3.3V, 5V and 12V). Having several small pads allows to easily implement some nice-to-have features such as precharge:

enter image description here

\$\endgroup\$
5
\$\begingroup\$

Things I would watch out for.

  1. Differences in contact resistance, you can mitigate this by using separate wires for each pin. The resistance of the wires will add to the resistance of the pins making their ratios closer.
  2. Bad pins, you can mitigate this by using at least one more pin than your calculations say you need.
  3. Connection and disconnection. If a connector is connected and disconnected under load then at some point one conductor is likely to take the whole current before finally breaking, many connectors aren't designed for disconnection under load even with only a single pin in use. One possible mitigation is to use connectors with "pilot contacts" that are designed to make last/break first and can be used to turn off power before the connector is fully pulled out. Alternatively you may decide that simply telling users not to connect and disconnect under load is sufficient.
  4. Faults, you want to make sure that if a wire shorts to ground or another power rail you disconnect the power rapidly rather than leaving it to cook.
\$\endgroup\$
1
  • \$\begingroup\$ In many of the applications I was referencing in my example individual wires are not involved. Our usage typically has power supplies and user assemblies plugged into a backplane (a large PWB) with a dedicated layer (or layers, depending on the current) to move the power around. \$\endgroup\$
    – SteveSh
    Nov 5, 2021 at 19:21
2
\$\begingroup\$

Whether or not it's bad practice really depends on the specifics, including who will be making and breaking the contacts.

At one extreme: I have often seen parallel sets of 400A conductors on 480V 3-phase systems, terminated with "CamLok" connectors. Up to 6 sets for 2400A capacity is not unusual for mobile generator connections. These are usually handled by trained electricians following procedures to make sure it is safe. (E.g. test for voltage, make sure you haven't crossed the phases on one parallel set to create a short, etc.) Even then, the access door to the connectors may be interlocked with a circuit breaker to make sure connections are only made or broken when de-energized.

At the other extreme: With a consumer product, I would probably assume that if there exists a connector that will mate with the ones on the product, it will be attempted. Make reasonable assumptions about what will be at the other end, and take measures to make sure it fails gracefully.

\$\endgroup\$
2
  • 1
    \$\begingroup\$ In 480V land you are under NEC or CEC. Paralleling of this sort must be done using equipment rated for purpose on the origin (fused) side, NEC 240.8. I have one of these; each conductor is fused individually. So 6 fuses for 3 phases x 2 parallels. \$\endgroup\$ Nov 3, 2021 at 20:14
  • \$\begingroup\$ @Harper-ReinstateMonica Not sure what context you've seen individually protected parallel sets. NEC 240.8 only says that parallel CBs or fuses are allowed, not required. The requirements for parallel conductors are in 310.10(H). You would probably only need parallel fuses for small wires (smaller than AWG 1/0.) In the 2400A generator output in my example, the output overcurrent protection would likely be a single power circuit breaker with multiple sets of lugs. \$\endgroup\$
    – Theodore
    Nov 4, 2021 at 14:36
1
\$\begingroup\$

The balancing factor here is that resistance of normal contacts and conductors grows with temperature. As long as the principal connection of the separate connectors is all in the same ballpark, the currents will lean towards self-balancing.

The antithesis is the thermal secondary breakdown in power semiconductors that happens at an operating point where resistance drops with temperature. As a result, current swiftly concentrates on a rapidly heating low-resistance stretch of material that ultimately melts/burns.

For normal contact materials, the thermal coefficient of resistance is positive, meaning that current will rather be drawn to cooler parts of the material.

\$\endgroup\$
1
\$\begingroup\$

If the parts to be fed have different limits then the circuits should reflect this and have current limitation build in.

some mobos only output 500ma and the USB chips shut down above that to protect against over-current - in this case manufacturers often supply cables with multiple USB connectors carrying just the supply which when plugged in resolve the issue by supplying enough current each to support the need.

(as an aside) another old-fashioned approach to feeding power was to consider the size of the carrier in strands. i.e. the number of strands within the cable. when you have a cable of 7 strands you can peal off a strand to do 1 strand worth of current like to run a 3 amps bulb. 2 would run a 6 amp job etc so the full 7 would run a 21 amp job. so you can see how a ring main and lighting circuit could be powered from such a cable.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.