Relay weirdness - silence detector and EPU

[arjoll]

I've just finished building a silence detector and EPU (emergency programme unit) for Cave FM in Gore.

This unit is similar to a commercial ex-Rhema one we use for Life FM to cross to a backup programme on an MP3 player when the satellite drops out. Occasionally the Raspberry Pis I supplied for Cave FM's studio-transmitter link crash, and this will save a little bit of panic when that happens.

The circuit isn't anything too fancy. A couple of variable gain input buffers each for main and backup audio, main audio summed and going to a precision rectifier and then a comparator (just using the other half of the dual op-amp that I'm using for the rectifier). The output of this goes high when audio is present.

The logic for silence detection uses a couple of 4017 decade counters, clocked by a 555 astable running at around 0.67 Hz. The output of the comparator, running through a couple of gates on a 40106 Schmitt inverting buffer, feeds the reset line on a 4017, holding it in a stopped state. When audio is lost, the 4017 starts counting, and after about 15 seconds of no audio output 9 comes up, which is connected to Clock Inhibit and the Set line on a 4043 RS latch. This energises a relay, changing to backup audio.

The audio present line also feeds a second 4017's reset line, this time via one of the NOT gates, so this 4017 counts when there is audio. Output 4 is connected to the clock inhibit line of this 4017, and also to the Reset line on the 4043. This doesn't normally do anything, but when the audio was lost, this gives a 6 second delay before the main audio is restored - this means that random clicks and pops on the audio won't result in audio switching back.

The output from the relay heads to an inverting output buffer with unity gain. The output is balanced - hot is the output from the unity gain output buffer, cold is from the relevant input buffer via the relay.

There's some other switching to select auto/manual and, when manual is selected, main or backup.

One thing that has confused me though is the action of the relay. I found a really nice little one at element14 - less than $2 and all the pins were on a 2.54mm pitch so it was ideal for stripboard. It had a 1.8k coil and was advertised as "high sensitivity" (I didn't notice that bit when I ordered it).

The relay acted perfectly normally when fed with 12V - clicked nicely, contacts changed over, everything you'd expect a relay to do - but in circuit it acted strangely. There was no audible click when it should have, and the relay appeared to stay in the NC position. With the power on and audio present it appeared that both the NO and NC contacts were closed - at least that's what the beep of my DMM told me, until I realised the continuity beep was modulated with the audio. As things quietened down between tracks, the DMM quietened down too.

I scoped the voltage across the relay, and it seemed smooth. No ripple or noise. Taking the relay out and testing it, it acted perfectly normally.

In the end I gave up and used a chunkier relay I had in stock - resulting in some pretty horrible looking jumpering, extra holes and cuts to the stripboard to make it fit. This worked perfectly.

The output from the gate fed an NPN driver using a PN100 transistor, nothing overly fancy and a configuration I've done heaps before. Any ideas what was going on? I'd find it hard to believe that a relay like that couldn't handle a couple of volts of audio!

[TonyDuell]

A few questions :

Do you have an (accurate!) circuit diagram of this device?

What state should the relay have been in? In other words was it pulling in when it
shouldn't have been or dropping out when it shouldn't have been?

Did you fit an anti-back-emf diode across the relay coil? (HV spikes from a relay coil
at turn-off can confuse logic circuits if not actually damage devices. The downside
of putting an antiparallel diode across the coil is that it slows down the release of the
relay, but that may not matter here).

[GMB]

Maybe it only works properly if the polarity is as shown in the diagram?

I'm thinking things like is has a built-in diode, or maybe the high sensitivity version has some magnetic bias so doesn't work so well in reverse.

[julie_m]

I've encountered "high sensitivity" relays before. They had a permanent magnet, not quite powerful enough to pull in the armature from the NC position or hold it in the NO position, and required for the coil to be energised with the correct polarity so the fields from the core and the permanent magnet reinforce one another.

Quickest fix if you got the coil polarity wrong would have been to mount the relay on the other side of the breadboard .....

[SiriusHardware]

Built in diode: I've been caught by that once or twice as you can get left-hand and right-hand versions of otherwise identical relays - one with the built in back EMF diode one way around, the other with the built in back EMF the other way around.

What exactly does the drive from the output of the 4043 to the relay look like? is it an NPN transistor with a series resistor between the 4043 output and the transistor base? Is there also a resistor from the transistor base down to ground?

If not, any voltage over 0.6V from the 4043 output will provide some drive to the transistor and therefore the relay, which, being a high sensitivity type, doesn't need a lot of drive to make it operate. You could therefore have the seemingly bizarre situation where both logic 1 and logic 0 out of the 4043 make the relay operate.

This is a very common problem where 741 op-amp outputs are used to drive a relay via a transistor: The output of a 741 can not go any lower than about +1.2V on a single rail supply, so the output 'on' state and the output 'off' state both provide enough voltage to turn on a transistor.

The simplest solution in both cases is to put an extra resistor from the transistor base down to ground, with a value chosen so that the two resistors divide the IC logic 0 / off state output voltage down to something less than 0.6V.

Of course, this might have no bearing on your actual problem at all. But it would be interesting to know if your (now working) relay has any voltage at all across its coil when it is OFF.

[arjoll]

Thanks for the ideas everyone - when I posted that I'd given up working out what was wrong, and put in a "relay" I had sitting in stock - you'll see it in the pic. It meant some ugliness on the bottom of the board, but I was kind-of over the whole thing and just wanted to make it work!

Originally Posted by TonyDuell

Do you have an (accurate!) circuit diagram of this device?

Sort-of. I have the logic drawn out in Visio so I could get a second opinion from my dad (who built his own CMOS-and-op-amp stuff in the 70s and 80s), but with notes etc scribbed on it, while the analogue sections and clock are all just hand-drawn in separate "blocks" - stuff like the balanced output are things I just do off the top of my head, but I thought I'd sketch them out anyway.

Originally Posted by TonyDuell

What state should the relay have been in? In other words was it pulling in when it shouldn't have been or dropping out when it shouldn't have been?

It appeared to be not pulling in when it should, but as I mentioned when I checked it with the continuity "beep" on my meter's ohms range it appeared to be stuck on AND off at the same time - until I noticed the "on" appeared to be modulated with the main audio.

Originally Posted by TonyDuell

Did you fit an anti-back-emf diode across the relay coil? (HV spikes from a relay coil at turn-off can confuse logic circuits if not actually damage devices.

Absolutely

Originally Posted by ajs_derby

I've encountered "high sensitivity" relays before. They had a permanent magnet, not quite powerful enough to pull in the armature from the NC position or hold it in the NO position, and required for the coil to be energised with the correct polarity so the fields from the core and the permanent magnet reinforce one another.

That'll almost certainly be it. I must admit, I didn't even notice the polarity indicated on the datasheet, and I have not found this before. I didn't specifically go out to buy a high sensitivity relay - I just put in "12V DC DPDT solder" into element14's website, sorted ascending by price and scrolled down until I found one that looked "normal" and was in-stock in Sydney (2 days from there vs 6-7 days from UK, US or Singapore). Comparing the datasheet to my layout, I did have it in backwards - simply because I didn't know there was such a thing I wish I'd noticed it earlier...

I'm guessing I fluked getting it the right way around when I tested it out of circuit!

Originally Posted by ajs_derby

Quickest fix if you got the coil polarity wrong would have been to mount the relay on the other side of the breadboard .....

...yeah...nah...wouldn't have fitted, but two breaks and two jumpers would have been far easier than the butchering I had to give that poor stripboard!

Originally Posted by SiriusHardware

What exactly does the drive from the output of the 4043 to the relay look like? is it an NPN transistor with a series resistor between the 4043 output and the transistor base? Is there also a resistor from the transistor base down to ground?

Just a resistor from the 4043 to the base. Nothing down to ground, but when I tested it, it was showing basically 0V across the coil at logic 0 (the meter was showing something like 0.01), and 12V when logic 1 - and nice and smooth.

Originally Posted by SiriusHardware

The simplest solution in both cases is to put an extra resistor from the transistor base down to ground, with a value chosen so that the two resistors divide the IC logic 0 / off state output voltage down to something less than 0.6V.

I'll certainly remember that for next time - it makes sense.

Thanks
Andrew

[arjoll]

And here's the unit, finished and installed.

[arjoll]

Forgot this one - the rack in its entirety. From top to bottom - Life FM: cavity filter, amp, exciter, stereo/RDS coder, silence detector/switcher, EPU (black panel), satellite receiver; Cave FM: transmitter, processor, silence detector/switcher. In the shelf there's the PoE injector for Cave's link, the Raspberry Pi that receives audio and MP3 player for backup. Monitoring is cheap but works - a set of Logitech 2.1 speakers, and an old Akai tuner out of shot at the bottom of the rack. Cave's filtering is zip-tied to the side of the rack.