Relay vs. FET's

I don't have lots of horror stories to share, but I'm not clear on what problems you have with the solid state versions of relays. Which direction are the signals going? If they are driving your optos, I don't see how that is significantly different from driving your relay coils. You would have an input voltage min and max and a spec on the current that must be supplied.

For an output that is driving the remote panel, then you have a driver (contacts in the relay) that have a max voltage and current rating.

How is this different between the relay and the solid state version?

How do you see using a relay making this easier?

--

Rick C

Hi Rick, I'm not sure it would make it easier. (other than in my head, maybe)

In case I wasn't clear: Our board is what provides the "contacts" that the n get wired over to the alarm panel. (I think you got that; just making su re.) Right now, those "contacts" are FET's, but I'd rather they be relays.

Most alarm loops (provided by others) seem to run about 12 VDC with a 2.7k supervision resistor. So, alarm loop current is generally low. The wiring from our board over to the alarm panel is usually not well controlled. So metimes shielded, sometimes in conduit, sometimes just flapping in the bree ze.

The typical FET arrangement we're using will measure 0-Ohms closed, and sev eral Meg's open (with the alarm loop disconnected). But when you connect t he alarm, sometimes the "open" condition can read as low as 1 Meg-ohm (whic h frankly, even that shouldn't be a big deal).

But it seems everytime I get involved with this interfacing issue, I'd rath er there not be a direct copper connection to any of our equipment other th an via the business side of a Form-C dry contact relay.

I occasionally get into situations where I can pull the alarm wires off our interface, and the interface works exactly as it should. Then, open & sho rt the alarm wire loop (now totally disconnected from our interface), and t hey work just fine too. But hook them up as intended, and sometimes things get weird (granted, not very often -- maybe even borderline rare?)

I don't think we have enough data and cases from the field to pinpoint what is going on, but my strong suspicion is that it has something to do with o ur FET output.

Adding to this: "back in the day"..., I think every residential or commerc ial burglar / fire panel sensor I ever saw used a relay. Like a PIR motion sensor, or smoke detector. (Maybe my sample size was too low!) :)

Anyway, it's late. I appreciate the answers and everyone's input. I know it's probably been beat to death... I just have "that nagging feelin g" about this one...

Relays are great. They are isolated, tough, cheap. Ron is low, voltage and current capability are high, no likely ESD is going to zap them.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Ok, so I get that you are emulating the relay contacts. You still haven't explained what your problem is. When you say things like "things get weird", I don't have any way to translate that into a description of the problem.

You seem to indicate you are concerned about a 1 Megohm resistance for a "relay open" state. But then you say "even that shouldn't be a big deal". This seems to be an emotional issue for you; "I'd rather there not be a direct copper connection to any of our equipment" is the best you can say as to what the problem is.

Why can't you explain what problems you have seen?

From what you describe there should be no grounding problems... at least if the power supply to the isolated side of the optoisolators is done right that is. You didn't describe that. The transistor side of the opto has to have some power to control the FET. This power source needs to be isolated from the rest of your box just as the opto isolates the signal path. Do you know hwo this is done?

--

Rick C

+1 for a set of contacts for the user to do whatever they want with them. AC/DC, floating, higher current/voltage etc etc.

Providing changeover contacts gives them an "inverted" output option as well.

If they stuff up (even momentarily) they can damage a FET meaning the whole assembly needs replacement or workshop repair. Relays are more robust, they can also provide feedback to installers when they "click".

Stick a polyswitch fuse in series with the relay output common for some added protection.

BM2335 wrote: >

** OTOH there are a many drawbacks:

Relays momentarily disconnect if you bump them or chatter when subjected to vibration.

Relays *cannot switch off* when DC voltages and currents are high - like 50V and 5A - the contacts and burn instead.

Relays need significant drive power, have fragile mechanisms and are not good at switching small signals reliably.

They have their place and I like using them when they ARE the best solution.

..... Phil

All true. Depends on what the OP has connected to the contacts.

I'm guessing, perhaps wrongly, it's a low speed signal indicating an alarm triggered state, driving a strobe or siren etc etc.

We usually use opto fets rather than mechanical relays for signalling to "unknown stuff". With a birectional TVS across the "contacts", since they seem to have zero specifications for ESD.

--

John Devereux

hen get wired over to the alarm panel. (I think you got that; just making sure.) Right now, those "contacts" are FET's, but I'd rather they be relays .

k supervision resistor. So, alarm loop current is generally low. The wiri ng from our board over to the alarm panel is usually not well controlled. Sometimes shielded, sometimes in conduit, sometimes just flapping in the br eeze.

everal Meg's open (with the alarm loop disconnected). But when you connect the alarm, sometimes the "open" condition can read as low as 1 Meg-ohm (wh ich frankly, even that shouldn't be a big deal).

ther there not be a direct copper connection to any of our equipment other than via the business side of a Form-C dry contact relay.

ur interface, and the interface works exactly as it should. Then, open & s hort the alarm wire loop (now totally disconnected from our interface), and they work just fine too. But hook them up as intended, and sometimes thin gs get weird (granted, not very often -- maybe even borderline rare?)

at is going on, but my strong suspicion is that it has something to do with our FET output.

rcial burglar / fire panel sensor I ever saw used a relay. Like a PIR moti on sensor, or smoke detector. (Maybe my sample size was too low!) :)

ing" about this one...

Sounds like FETs are costing you due to faults. It comes down to what is th e extra cost of relays, what's the cost of these faults. If you know fets c ause issues and plan on dealing with them, put a place for a relay on the b oard & make it easily accessible, then at least you can solve them quickly & reliably.

NT

By "weird", I mean sometimes the alarm panel sees our output and goes into the desired alarm condition, and sometimes it ignores our output. The last two times this has happened, the board (our board) was simply replaced and allegedly this fixed whatever the root problem is.

As to the optos in our build, I have the schematic at the office and will p rovide more detail on this problem this afternoon. I think I have it memor ized, but let me verify it anyway, and maybe grab the latest version from i nventory and take a good hard look at the circuit board itself to confirm i t matches the schematic exactly. The short answer is I think the opto isol ation is done correctly, and yet...

In reply to Phil and BM2335, yes - small signal. Signal output (from us) to be used by the alarm panel (them).

I guess the question is what was found when the "defective" panel was tested?

Is the alarm state a contact closure or a contact opening? I know alarm systems have fail-safes built in, so it could be either way depending on which was deemed more reliable.

--

Rick C

Did you get the "bad" board back and diagnose it? it could be ESD, or there is a lot of variation in opto-coupler's tranfer ratio, so some could be too weak to show as a solid "closed" contact. On an electromechanical relay, closed should be way less than 1 Ohm. On the other hand, relays could get dirty (oxidized) contacts and not show as closed the first time you did that after a year of sitting open.

And, of course, most opto's are polarity sensitive, so the wiring must get the positive from the alarm control to the positive of the opto's output.

Jon

Both examples above indicate poor design. There is no reason why an opto-isolator can't do the job as well as a relay.

They don't need to be polarity sensitive. You can buy solid state relays that work for AC as well as DC. It just needs two output FETs rather than just one.

--

Rick C

Surely you mean Form-A contacts..?

To emulate a truly universal SPDT (Form-C) contact, you need two optos (driven alternately), and four transistors (each is an antiseries pair). Maybe you meant this; just unsure.

In installations like that (long wires, in a facility, conduit or wiring-panel environment), I'd worry about surge ratings. Is it acceptable that your contact may suddenly, without warning, fail as a three-way short?

Even if you provide transient protection (usually, a MOV rated at

0.3*Vds(max), or a sufficiently beefy TVS rated for about 0.7*Vds(max)), it's a tricky proposition, because the TVS device will also eventually fail shorted -- given enough large surges. And, during the surge, all that current is passed to the load, i.e., your switch momentarily looks cloesd. (Maybe not for very long, a few microseconds perhaps; but maybe that's enough, who knows?)

You also have a lot of capacitance (particularly with the TVS present), but that shouldn't be a problem around most contact-closure situations, which are DC or mains frequency.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

One of our products needs to interface to various, unknown, ever-changing, commercial building alarm panels. Our designer is using opto-isolators to drive a pair of N-Channel FET's and this arrangement supposedly is identical to the "old-fashioned" way of just using a general purpose small signal relay to construct a Form-C output. (Similar to: Digikey Z773-ND, which is an Omron G5V-1-DC5.) Maybe I can draw the circuit in ASCII ART if needed. Although our board and the alarm panel are usually in the same room, there is no guarantee they share anything else in common. Might be the same "ground", might not be.

Anyway, I'm the guy who usually gets stuck dealing with this interface board when problems creep up in field installations. Sure, we can figure them out, but that takes time, labor, and typically.. aggravation.

I say that because we simply don't know (and therefore cannot control) whatever voltages and currents are present in the alarm panel loops (or god knows what else!? - third parties connect some pretty weird stuff sometimes!), that we should just use relays and skip the FET approach entirely. To me, the added benefit of using a relay is it at least establishes a clear demarcation between where our obligation ends, and the third-party world begins. (But maybe I am just old and stupid? I do like FET's for many things... just not this.)

BTW: These alarms are for things like AC-Fail, charger fail, high temp, etc... It's not like we need the speed or lifetime cycles that a FET arrangement provides. Also, build quantities are low enough that a few general purpose relays per build isn't a deal breaker.

Thoughts? Relays or FET's. (or does it matter?) Horror stories appreciated, especially if they can be used as ammunition to advocate for a relay solution. Thanks!!

OK. So as promised, I have the schematic in front of me and will describe the circuit in more detail.

We'll start with the 74AC14 (Hex Inverter w/ Schmitt Trigger input). Everything before this point in the design is not relevant. Each inverter on the 74AC14 can sink/source 24 mA and I calculate we're at around 17 mA or so.

We eventually construct (3) Form-C's equivalents from these 6 inverters. I'll describe one such arrangement; the other two are just cookie-cutter.

A signal originates at 74AC14 pin-13 (input). The output (pin-12) goes to a 510-ohm resistor, and then to Pin-1 (anode) o f a TLP222A-2 (2-channel opto-relay). The cathode (pin-2) is grounded.

Also on the 74AC14 output (pin-12) is another 510-ohm resistor in-series wi th an LED to ground (for use as an indicator light).

So far, we have one-half of the Form-C. (i.e., the "normally-closed" side)

74AC14 (pin-12) also connects to pin-11 (input to another inverter). Pin-10 is the corresponding output, which goes to a 510-ohm resistor and th en into pin-3 (anode of the 2nd channel) of the opto-relay. The correspond ing cathode (pin-4) is grounded.

This is the other half of the Form-C. (i.e., the "normally-open" side) And we do this arrangment a total of 3x for three Form-C's.

So to recap, we use one inverting, and one non-inverting, output of the 74A C14 to each drive a single anode of the TLP222A-2 opto-relay.

One 74AC14 output will always be negative, and the other will be positive. This action either enhances (or does not enhance) conduction on the associa ted opto-relay outputs (i.e., for Anode-1, TLP222A-2 drains are on pins 7 & 8; and for Anode-2, the drains are on pins 5 & 6). All drains on the TLP2

22A-2 are left floating (i.e, they do not connect to anything but the termi nal strip we provide to the alarm company).

Honestly, looks OK to me. (?). So, WTF? :)

I will have a known defective (new) unit in my hands on Tuesday.

And now I have exhausted my honey-do excuses and must now go clean the gara ge. Uggh..

Links:

Should I assume the 74AC14 has a 5 volt supply? I don't get the same current. 5V (Vcc) - ~1.2V (led) ~= 3.8V. 3.8V / 510 ohms is roughly

7.5 mA (the recommended drive from the data sheet). Times two gives 15 mA, not 17 mA. Did you assume 0.7 volts forward drop on the LEDs?

Up until now I had the impression you were using an opto-isolator and separate FETs. I think using a opto-relay is the right way to go. Is there no protection on the output of the opto-relays?

Do you know the nature of the circuits this will be driving? If driving relay coils you would want a bidirectional TVS or other means of absorbing the inductive kick back.

I don't agree with your assertion, "Everything before this point in the design is not relevant". Is the inverter driven directly by a digital output, like an MCU? The devil is in the details.

I assume the devices you drive only use one side of the output contacts at a time? Does it matter if the relays are make-before-break or break-before-make operation? There can be brief times during switching when the common contact is connected to both the NC and the NO contacts. It looks like break-before-make could be assured by the opto-relay itself, but the specs are not complete. They give a max time for turn-off of 0.5 ms, but no minimum time for turn-on. Turn-on typical is

0.8 ms.

Any idea if this is a potential problem?

Also, you still haven't explained what the boards are failing to do. Have you? Do they just stop driving the outputs? Are both the NC and NO contacts open? Both shorted? Do they chatter when operated? Knowing these details is probably more important than knowing the circuit details.

--

Rick C

hence we'd need the circuit diagram. A text description doesn't really work.

NT

rk.

The 74AC14 is driven directly by a MAX934 (quad comparator w/ precision ref erence). The comparators are measuring various resistor divider networks, each triggering at a different voltage threshold. Other than the tolerance of those networks (1%), I don't see anything here that could remotely caus e a "no output when expected" condition. It is as if the opto-relays are b eing told to "fire" but they are not. Keep-in-mind, supposedly these passe d inspection and burn-in before being shipped! The field installers don't really have the right test equipment to test for chattering output, other t han the LED's, which are not changing as expected.

I doubt timing of the external world is an issue (TBD), even if both NC and NO were somehow connected to common for an instant. There is no microproc essor or other software/firmware at work in this design. It is just gates, logic and comparators. Nothing fancy.

I am honestly hoping for an "A-Ha" moment on Tuesday when some known recent install failures get back to the lab. For the record, I never liked this circuit anyway. Even when it works (which is nearly always), part of it is not technically doing what it needs to do -- but that's a long story invol ving the difference between a direct measurement of something, and an indir ect inference of an operating condition. I won't bore you with those detai ls. :)

Link:

/MAX934CSE--ND/948242

Where does the error display come from, your equipment or the equipment it is driving with the relay outputs? Is this some sort of self test process in the equipment your relay outputs are driving?

Saying the comparators are measuring resistor divider networks, I assume you mean the references are generated by resistor dividers and are being compared to some voltage levels on the inputs, the idea being to detect when the voltages cross the thresholds.

So why exactly do you suspect the 74AC14 relay drivers or optorelays rather than the circuit driving the 74AC14?

I don't know enough of the inputs to the comparator circuit or the circuit itself, but this sounds like an area where there could be some issues, especially if the issue had to do with the interface from what is being measured.

Obviously it would be hard to diagnose this problem from afar. Just trying to convey all the info is not easy. This is one failure you will need to duplicate on the bench to figure out. I don't see any reason to suspect the relay outputs themselves though.

Do you have a means to duplicate the conditions for the failure?

--

Rick C