High voltage protection on low voltage data lines

[rx8pilot]

I have been designing some peripheral add-on for an existing system using 1-wire for communication. So far it is all working well and easy to use. The challenge I am trying to solve is over-voltage protection on the 1-wire data line which is bi-directional like i2c but there is no clock line.

The way these things will interface and the nature of their use gives them a reasonable chance of being connected incorrectly and the data line will see up to 28v where the max allowed is 6v. The over-voltage would not be a short spike, but a continuous application of DC. My first simple approach was to put a resistor on the line and a zener to ground, but the data line cannot tolerate much resistance at all. A suitably small resistor is around 50 ohms which requires the zener to dissipate a lot of power during a fault.

I found another part from Maxim call the MAX9940 which is an active IC that uses a MOSFET switch to allow the bi-directional data to flow until it senses an overvoltage  and it shuts off the data line before the MASTER could be damaged. This is great and the part is suitably low cost.

The remaining problem is that there could be 7-8 1-wire devices on the line and I am hoping to protect them as well. They are hopefully as small as possible and they are parasitically powered by the 1-wire data line. The solution to protect them therefore has to be passive because no power will be available (most of the time). Being that the 1 wire lines cannot tolerate much resistance at all - does anyone have any suggestions for protecting these devices from a threat of overvoltage passively? The PCB space is very limited too which add another dimension to the challenge.

The images are clipped from the data sheets.
Maxim DS2484
 max voltage on 1-wire IO pin is 6v. The 1-wire is passively or actively pulled up to 5v.
Maxim MAX9940
 can protect the DS2484 master and other downstream devices, but will leave the slaves to be fried.

This stuff will be used in remote locations and subject to a rather hard life electrically and environmentally. The goal is to make as big an effort as practical to keep it survivable when mistakes are made. Repair and replacement is usually hard or impossible until it is too late to matter.

[exmadscientist]

Treat it like a power line (which it technically is) and use an overvoltage protection IC such as the LTC4365? Since it's a low-current line and you don't need reverse voltage protection, you can get away with a very tiny single FET instead of the beefy duals typically paired with parts like this one.

[daveshah]

What about a low current polyfuse (e.g. http://uk.farnell.com/multicomp/mc36205/fuse-ptc-reset-smd-60v-100ma/dp/1861178) and zener to ground?

[rx8pilot]

What about a low current polyfuse (e.g. http://uk.farnell.com/multicomp/mc36205/fuse-ptc-reset-smd-60v-100ma/dp/1861178) and zener to ground?

That is not a bad idea. The initial 15 ohm resistance would allow a modest size Zener to survive and it would not mess up the data. I will add those parts to my next DigiKey order. Super easy to give that a try. Curious what the endurance of the fuses are.

Nice suggestion, did not cross my mind.

[daqq]

That is not a bad idea. The initial 15 ohm resistance would allow a modest size Zener to survive and it would not mess up the data.

Be wary of the power dissipation on both the Zener and the polyswitch - the fuse will have a current flowing through it all of the time. The current and voltage will still be there, at the holding current.

My first simple approach was to put a resistor on the line and a zener to ground, but the data line cannot tolerate much resistance at all

That is strange... as far as I knew 1 Wire was fairly tolerant to this sort of thing.

[ajb]

Have you looked at Bourns' TBU devices?  They're sort of like a polyfuse on steroids for data lines.  I've been meaning to try them out on RS485 bus, but haven't gotten around to that yet.  3GHz bandwidth, <1us response and protection ranging from 40-850V.  They're not terribly small, but should be able to be paired with a smaller TVs or zener due to their faster response.

[rx8pilot]

I am not a 1-wire expert at all and hoping to figure out a couple of things.

First, I disabled the active pull-up on the master thinking that I could use a passive pull-up of a value appropriate for the zener protection to be viable. When I turn-off the active pull-up, the line is still pulled up without a passive resistor pull up. I am not yet sure how strong it is, but it appears to be in the 700ohm range which is rather heavy. I was hoping to use a 3k passive pull up and a 500ohm current limit resistor with zener.

I will continue to look for something I am doing wrong on the 1-wire master setup which is entirely possible. the data sheet clearly indicates that an external pull-up is possible. Although the MAX9940 wont really care, the slaves will.

Be wary of the power dissipation on both the Zener and the polyswitch - the fuse will have a current flowing through it all of the time. The current and voltage will still be there, at the holding current.

During normal operation, the current is in the nA-uA ranges. During a fault, the fuse only conduct until it gets to it's trip point where it will disconnect. After it cools down, it will re-connect and start the cycle over again. Dissipation should not be a problem unless the fuse resets extremely fast. The data passes through the fuse.

I made a diagram that seems to make sense to me. The zeners would need to be fairly stout to handle the current spike leading up to the fuse activating, but not too bad.

[rx8pilot]

Have you looked at Bourns' TBU devices?  They're sort of like a polyfuse on steroids for data lines.  I've been meaning to try them out on RS485 bus, but haven't gotten around to that yet.  3GHz bandwidth, <1us response and protection ranging from 40-850V.  They're not terribly small, but should be able to be paired with a smaller TVs or zener due to their faster response.

Never looked at those. A quick look seems like they are intended for transient events - not DC. Should probably read further. The first data sheet I found was marked 'Obsolete'

[max_torque]

I solved a similar problem on a CAN based system by having the host micro being able to sense the voltage on each wire and be able to disconnect it during continuous overvoltage condition!

(with CAN, you have a 120ohm terminating resistor, and on my 48V system, that meant worst case of 48W! )

You could either measure the voltage with an ADC, or use a comparitor to trip an overvoltage interupt, or use a current shunt amplifier to trip at a known current etc.

In my case, because size wasn't an issue, i used a relay to open the I/O lines under fault condition (which also resulted in the unit going physically off buss if it lost power, which is a good thing)



Having said all that, another option, depending on your bus speed/ signal & data bias is to AC couple your data lines!!

Use a large resistor to control the DC component of the signal, but bypass it with a capacitor to allow AC current to flow and maintain the fast signal edges!

either do some simulation in Spice etc, or breadboard up a test solution, which would show you if it is practical in your case

[kony]

I'd say consider using LINbus PHY. 

[rx8pilot]

You could either measure the voltage with an ADC, or use a comparitor to trip an overvoltage interupt, or use a current shunt amplifier to trip at a known current etc.

On the controller/master side of the circuit I have the option of using an IC that essentially does exactly that - MAX9940. It measures the voltage and disconnects when a fault is detected. No problem.

The real problem is the various slaves on the bus. 1-wire is a unique challenge because there are only 2 wires total - 1 wire for data/power and 1 wire for GND/return. This means that there is no power available for additional active circuits to deal with protection. No comparators or anything else. A 24v DC fault on the line (almost exclusively operator error), will easily be detected and dealt with for the master, but each slave device must have it's own passive solution to deal with the fault.

Another note about 1-wire (I am learning a little every day), is that it requires a substantial pull-up to work. The reason the pull-up must be strong is that the slaves are pulling power from the data line as it idles high. In scenarios where the data line is saturated with data and there are a good number of slaves - they are all trying to steal power everytime the the data line is released to a high state. Because of this 'hard pullup', an opposite pull down is needed to pull the data line below the threshold for a low to be detected. Only a small amount of resistance can be tolerated on the data line of around 50 Ohms.

All of this makes traditional line protection not workable.

Low resistance.
Passive.
Small. (-ish)

[max_torque]

Surely therefore the protection of the input line can be split into a DC and an AC component.  The data route can be protected by being AC coupled, and the DC route  perhaps with a crow bar or TVS?  Does the slave have to be operational during the over voltage event?

[rx8pilot]

An over-voltage event can disable all slaves in the name of survival. No problem with that. The reason the signal cannot be split is because there would no longer be any way to pull down the data line which is actively pulled up. The slaves do not have Vcc input at all. The power comes entirely from the idle time of the data line. AC coupling the data would essentially turn off the slaves.

A TVS would sink a TON of current and blow the traces off the PCB unless some resistive current limiting is employed. The protection on the MASTER is about 20 Ohm and I would only have about 50 Ohms total. Each slave protection scheme could be at most 30 Ohms. With 24v fault going through 30 Ohms - 19 watts would have to be dissipated for as long as the fault exists. Unfortunately that would make the solution HUGE.

One of the previous suggestions was to use a very low current PTC fuse on the data line before the zener diode. During a fault, the zener would sink more than the PTC trip point, but only for a brief period before it disconnects and the zener is no longer doing much at all. In theory, it would allow the use of a fairly small zener assuming the reset of the fuse is slow enough for the zener to cool off. It is an easy experiment to try it out. The requirements make this a tougher challenge than I originally gave it credit for.

The prize is a network of simple peripherals that only need a single pair of wires and a simple connector.

[Siwastaja]

One of the previous suggestions was to use a very low current PTC fuse on the data line before the zener diode. During a fault, the zener would sink more than the PTC trip point, but only for a brief period before it disconnects and the zener is no longer doing much at all. In theory, it would allow the use of a fairly small zener assuming the reset of the fuse is slow enough for the zener to cool off.

You could try thermally coupling the zener and the PTC. In this case, in addition to working from the overcurrent, the PTC would also function as a thermal overload switch, switching off from the heat generater by the zener.