Input protection: is a tri-state buffer overkill?

[Pack34]

I'm currently reving a design to add additional I/O for the specific purpose of providing the customer with additional signal lines to control the device (OEM level stuff).

Typically for protection I'll just use three levels of protection: 1) ferrite bead, 2)RC filter (100R-110p), 3)TVS diode array.

Now, since in this situation the customer would be generating their own signals from whatever device their using to control my product, I was thinking of adding a fourth level consisting of a tri-state buffer array just as a precaution in case they manage to put together a situation to blow up a pin, it would damage this cheap intermediary board and not the expensive main control board.

However, all of the pins on the MCU are 5V tolerant so I may be adding additional cost and complexity that I do not need.

Thoughts?

[T3sl4co1l]

So this is a board that goes in front of the control board?

What's your risk profile look like? ESD from clumsy techs? Lightning strikes or nuclear EMP?

You'll be fine against the former; the latter, you'll need a few more stages, plus a lot of shielding...

Tim

[wraper]

Have you thought just using optocopuplers instead of all of that?

[Pack34]

Have you thought just using optocopuplers instead of all of that?

I'm sort of wishing I thought of that. The layout probably would be cleaner. However, the current design has a CE mark and I want to keep the same design and only add to it so I don't have to go through the CE mess again.

[Ian.M]

If you add buffers it will definitely invalidate any EMC testing you have done, so you are gambling if you CE mark it without retesting.  If you add optocouplers, you *should* retest, but as simple optos are relatively slow, and the input side is of course isolated, the chances of their EMI performance being worse than for the 'naked' input are minimal.

[Tomorokoshi]

Is there any information available about failure rates, warranty returns, etc. for similar enough products from that customer?

Will the interconnection of these boards be done at a controlled factory or in the field?

If it's in the field how likely is it that they will be hot-plugged?

How far away will the controlling device be located? Will is use the same power supply? How will grounding be routed?

A tri-state buffer could still get blown through by high-voltage static discharge.

If this is some kind of industrial / robotic / PLC installation opto-isolated inputs would certainly be nice to have.

[Pack34]

Is there any information available about failure rates, warranty returns, etc. for similar enough products from that customer?

Will the interconnection of these boards be done at a controlled factory or in the field?

If it's in the field how likely is it that they will be hot-plugged?

How far away will the controlling device be located? Will is use the same power supply? How will grounding be routed?

A tri-state buffer could still get blown through by high-voltage static discharge.

If this is some kind of industrial / robotic / PLC installation opto-isolated inputs would certainly be nice to have.

This is going to be a new standard product to make it easier for a company to integrate my product into whatever solution their building. It'll most likely be used in an industrial or pharmaceutical environment.

Interconnection will likely be done inside a controlled environment. The use-case is to take our product, mount it, and plug a wire harness into the instrument. Most likely a single board computer.

I am a bit concerned with hot-plugging. But I was planning on including a warning notice in the documentation and pack-in instructions to only make connections with the instrument un-powered. I can see the potential to add a sticker next to the connection points warning of this.

I'm aware that the buffer would blow but the hope would be that this interconnection board would fail and prevent damage to the rest of the system. This would be a quick and cheap swap out in an RMA process on our end if the customer did cause damage. If they manage on damaging the main control board it would add $500 to the repair BOM.

[Tomorokoshi]

What is the TVS array rated to compared to the input levels of the processor?

[Ian.M]

Here's an example of an optoisolator board designed to be as bullet-proof as possible and to be field serviced: http://www.industrologic.com/uob8man.htm
All the optos and their input resistors are socketed.   Something similar might suit your application, but with a dedicated connector and permanently mounted pullups for the MCU interface. 

If the input connector is a screw terminal block, your installers will thank you if you use a readily available pluggable one.

[bson]

Like tesla asked, what are you protecting against?  Are you looking for an on-board or off-board solution?  If spinning a new PCB isn't an option, then off-board is already chosen for you.  However, either way I would add TVS diodes to all non-AC connections on the next board rev since this will protect against both static discharge and polarity reversals...  Of course, a TVS diode isn't going to protect against an inattentive technician who attaches 12V to a 3.3V signal input.  That's going to let the smoke out no matter what and the protection against that comes in the form of a premade harness and connectors that prevent such mistakes.

[David Hess]

I like using separate (and replaceable) buffers or I/O expanders for external interface circuits for the very reason you identify however assuming low current inputs, raising the impedance of your RC filter while adding a little bit of parallel capacitance across the series resistor to preserve high frequency performance will go a long way to preventing damage from overdrive.  Typically input voltages up to the breakdown voltage of the parts in the RC filter become tolerable.

High impedance oscilloscope and frequency counter inputs work this way.

That sucks that you cannot add optocouplers without going through CE certification again.  There is practically no way they could compromise EMI performance without other design changes.

[nctnico]

Here's an example of an optoisolator board designed to be as bullet-proof as possible and to be field serviced: http://www.industrologic.com/uob8man.htm
All the optos and their input resistors are socketed.   Something similar might suit your application, but with a dedicated connector and permanently mounted pullups for the MCU interface. 

If the input connector is a screw terminal block, your installers will thank you if you use a readily available pluggable one.

Sockets suck because they often are the source of problems (poor contact). I strongly second the suggestion for using pluggable terminal blocks. It will make swapping the module much easier (and also production and repair testing).

[Ian.M]

Sockets suck because they often are the source of problems (poor contact).

Good quality turned pin sockets, with a contact material correctly matched to the plating on the IC or resistor module pins suck less, and provided there aren't any adverse environmental factors like high levels of corrosive gasses or particulates, significant temperature cycling and vibration, can be reliable.

If possible include status LEDs on-board to ease troubleshooting.

It may be preferable to treat the whole input protection board as disposable rather than making it field serviceable - you'll have to run your own cost/benefit and risk analysis.

[David Hess]

Some sockets are a lot worse than others as Texas Instruments aptly demonstrated as only they can but in a high vibration or temperature cycling environment, I would want positive retention which is easy enough to provide if necessary.  Actually in such an environment, I would expect as many failures do to the poor mechanical properties of solder, especially modern lead free solder, as from good sockets but then of course the sockets have to be soldered into place as well.  And with the exception of the mentioned TI sockets, I have seen more solder joint failure than socket failures.

I have never seen a collet pin socket fail and they are what I use for routine repairs when replacing soldered in DIPs and other parts.

[Berni]

Best solution i find for input protection is series resistors and a TVS diode hanging off it followed by another series resistor before going to the digital input pin. Such a solution only really works for low frequency signals like low baud rate UART as the resistance slows down your signals. But such a solution can survive an idiot feeding a 24V power rail in to a 3.3V input. Very large overloads like 220V AC in to it might blow up your diode and resistor but the digital IC behind it should be still alive.

A very low amperage fuse can be added in if you want it to handle huge overloads with a better chance of your digital chip staying alive while only that one input blows up at the servicing cost of a few passives (Like when your digital chip is an $100 CPU that must not stop because its also doing some other mission critical task).

[Kalvin]

You have resistors and TVS's connected to the input pins. Just add resettable fuse / PTC in series with each input, which will cut off / limit the current flow if someone is really torturing the poor device.