Relays-only loop selector.

[R.G.]

A circuit from my technical childhood has been pestering me for quite a while now. I think I have it cracked open enough to talk about.

There is a circuit which uses the difference between pickup and hold currents in relays to let you make a one-of-N selector switch. The way I remember it is that the relays are set up with all coils fed from a current limited supply, and turned on by a momentary pushbutton grounding the other side of the coil. One set of the contacts on the relay is set up to parallel the pushbutton when the relay is pulled in, so once it's turned on by pushbutton, it latches itself on.

The problem is how you get it to release. Pushing the pushbutton on another relay, the idea is that the two relay coils try to pull more current than the current limited supply can give them, and they BOTH lose enough current to hold in, and so both of them go into dropout. Only the one which was previously held in can drop out, and when it does, the current is now available to pull in the one with the pushbutton on. Kewl, a relay-only selector switch.

Mother Nature has other ideas. It turns out that this clever setup was for a generation of relays where the hold-in current was more than half of the pull-in current. If the hold in current is way low, then you can't make a current limit that is low enough to drop out two relays but hold in one. Modern PCB relays have hold-in at about 0.75 of full-voltage current, and the will only specify "greater than 10%" of full-voltage current. You really need 40%  for the hold-in current trick to work.

I got around this with a trick I saw in EDN. A resistor/capacitor on the ground side of the coil has the capacitor initially uncharged. When you push the button to pull in the relay, the cap eats a lot of current and the relay pulls in with high current. Then the resistor charges up the cap and the relay reverts to low holding current. In this case, the cap lets the newly-on relay suck a LOT of current, forcing the other relays off better.

So I covered a few sheets of paper with math, figured out that the resistances don't need to be negative (that's always a bad sign) and did some sim work. Looks like it can be made to work with **some** common 9Vdc PCB relays.

The flys in the ointment are that
(1) a 3PDT relay is really expensive, and you need one whole section for bypassing
(2) 4PDTs are worse
(3) it takes about half a second for the capacitors to reset, so no pushing switches faster than twice a second.

I got rid of the first two by noting that you can use a DPDT to do the latching, using one section, then use the other section to latch on another relay that actually does the signal switching.  So net cost per loop in relays is two DPDT 9Vdc relays, one R and one C, and another R for limiting all sections. LED indicators are a natural, and this lets you put your signal relay away from the selector switching, if that's any use to you. It's about $4-5 per loop.

Still experimental.