ESD zap

That, too, but the usual thin film parts can also suffer substantial parameter changes:

Can you place the clamps right at the pin and then a resistor to wherever it needs to go? That way a few hundred ohms would suffice and there is no overvoltage on the resistors.

--
Regards, Joerg 

http://www.analogconsultants.com/

OK caveat first, don't listen to me and do whatever Joerg says. But I stuck diodes(clamps) right on the pins (maybe at the end of wires) and zapped a laser diode with one of those peizo sparkers in butane lighters. (all over the place.. both ways...) I guess I figure the ESD diode's are big enough to take the energy.. (actually, I thought nothing like that at the time.. it worked and I moved on.)

1/2CV**2~(call V = 10kV to make math easy.. and ignore 1/2..-10+8~10mJ Why the 10k ohm resistor? (Hey maybe I should add bit of resistance?.. I could afford a few ohms.)

George H.

I also need to be able to stand a +-30 DC overvoltage, so I want to keep the R values up, and I can't clamp the inputs lest people blow up the clamps.

The pins lead, through the resistors, to the diff input of a 24-bit delta-sigma ADC. Ultimately, I'm measuring RTDs.

I guess I could put capacitors to ground at the input pins. 10 nF or so maybe. That would make a divider/lowpass with the 100 pF body model capacitance.

I wonder if a 10K 1206 would mind a 4 KV zot. Probably not. I guess I should try it. I've got an old Kepco 4KV bench supply somewhere around here, with one tube inside.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Why not clamp the input pin at 50V or such, and have your 10K after that?

--
Thanks, 
Fred.

That would fix it.

Carbon resistors may be an option, they can take such hits easier. But ideally it should be thick non-trimmed ones, assuming tolerance is not an issue and you run it in ratiometric fashion.

That is not high-tech, too pedestrian :-)

The rail for that doesn't have to be actively managed. It could be "pumped up" by pulses only when the system gets hit with voltage levels past its normal supplies.

--
Regards, Joerg 

http://www.analogconsultants.com/

4kV will nominally jump 40mils... add some San Francisco "fog" and who knows. Maybe use a carbon? ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

The 10K limits the current into the ADC chip inputs if the customer applies 50 volts. The ADC is rated for 10 mA max input current into its ESD diodes.

If I clamp at the input pins, I'd need +-50 volt clamps so I don't vaporize them with the specified DC overload.

I'm trying to be zen about this, not clamp/fight the overloads, but limit the currents. So, if a 1206 series resistor can't stand the ESD zap, maybe a cap at the input pin is the best way.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

That would take a bidirectional transzorb, and I'd have to be careful about its equivalent resistance to ground.

Or use two diodes to clamp rails that are weakly biased up to, say,

+-12 but also transzorbed to +-50.

Input caps sound easier.

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Well as long as whatever is driving it can stand the 10nF caps. (and whatever time constant that means.) For protection carbon comps are nice (as others said), but I'm guessing they don't come in surface mount.

(We use through hole carbon comps as termination resistors for an RF coil because they have non-magnetic leads.)

There's always R's in series. George h.

For lightning protection some time back I fielded two schemes for protecting hi-z differential analog mux inputs: - 10K carbon comp. in series with both inputs, 1uF differential cap, diode clamp - diff. gas discharge arrestor, 100k 1/2w series (both leads), 100nF differential cap., rugged CMOS mux

The idea was to limit worst case Vin to something sane, then limit mux current to something safe. (I made an IEEE surge waveform generator to blast it.)

Field failures stopped. (Except for the one, a direct hit. Lightning vaporized a bunch of the heavy steel box, and a big part of the board.)

Your zen solution is distribute the 4kV zap over a series resistor string that can handle it. Or gobble it with a cap, if you can stand the CMRR hit.

Cheers, James

If you have enough space, just put 10 1206 resistors in series.

I am not sure I have seen 400V rated of these but I think I did once (usually they are 200V).

Dimiter

------------------------------------------------------ Dimiter Popoff, TGI

------------------------------------------------------

I just looked at a few--1206 are rated 400V peak. I doubt that's the honest limit, but it sure ain't 4kV.

Here's a possible not-too-messy but still kosher idea:

C1 +5v o--10k--+---||---. | | - D1 === ^ | Vin >-----+-----/\/\/---> | R3 10K v D2 - | C2 -5v o--10k--+---||---. | ===

The diode clamps are the only per-input requirement, the rest is a quasi-adaptive ESD-gobbler. Much smaller than 10x 1206.

That's only for ESD. Sustained over-voltage protection would need a 2nd clamp to hard rails, after R3.

Cheers, James Arthur

Oops, D2 corrected.

Cheers, James Arthur

The ADC can stand 10 mA into its ESD diodes, so the 10K will protect it. As long as it doesn't arc over, which your clamp thingie will fix.

Or, I guess, just caps across the inputs.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation

In theory, the chip is already designed for human body model. So yeah, a little capacitance will do the trick.

You could probably go a little lower on the resistance if need be. Most chips are designed for 40ma injection, so even 10ma wouldn't be that bad. You need to insure that your power supply won't be pulled up with the external current.

All that is required under injection is the part won't latch up. There is no assurance that the part won't need to be reprogrammed, assuming it has internal registers. Or that the internal logic won't be hosed after injection. If you really want a rugged system, you should detect the external overstress then reset the system. Or use an analog watchdog chip. There is nothing worse than need to boot some measurement system.

On a sunny day (Mon, 28 Apr 2014 15:23:58 -0700) it happened John Larkin wrote in :

Yes why not?

Yes, extra capacitance slows rate of rise to allow time for input clamp diodes to conduct.

piglet

On a sunny day (Mon, 28 Apr 2014 15:23:58 -0700) it happened John Larkin wrote in :

Yes why not?

Actually whats wrong with a simple series resister (of high voltage capability), and a neon to limit volts, and then an other resistor?

in --- Rhv ------ R ----- out | 0 about 80V pp 0 neon | /// The advantage is no leakage in normal situation. There are many gas based surge arrestor types..

Good idea. 1206, 15kV:

(The 0805's are 10kV.)

Aren't neons a bit slow?

I nearly suggested tight-spaced pads on the layout--a home-made spark gap--just for fun. But that would be a nasty, carbonizing, metal-spalling dirt-magnet.

Cheers, James Arthur