Surge Capable Chip Resistors

170V and 22k is 1.3W, which is far over the power rating for small SMD resistors. You could use Panasonic's thick-film ERJ 2010 (0.5W, risky) or 2512 (1W) types - they allow 2.5x over- power. Or better (and smaller if mounted standing up), use a through-hole with appropriate construction (not a film type).

The Ohmite TFS part you found will be expensive and painfully hard to get (Mouser). It has substantial excess capability (3kV and 6J, vs your 170V and 26mJ). I wouldn't bother.

There are two problems with these:

1) Actually getting them 2) Paying for them

--
--
kensmith@rahul.net   forging knowledge

Do you need it to properly resist for that pulse, or just survive? Something like a ESD protection diode or TransZorb could probably save the resistor. EDS diode data probably doesn't cover over the us-range, but picking a random TransZorb datasheet on Vishay's site shows graphs into the 10s of ms.

--
Ben Jackson AD7GD

http://www.ben.com/

It's needs to somewhat resist properly.. R=22k+/-5k and the resistor can be nonlinear. (Oops left out these possible useful details..) The resistors function is just to limit current to about 8mA during the 20mS 170VDC pulse. I may be able to use a device other than a resistor but can it compete with 2512 footprint?? 6.4mmx3.2mm

D from BC

Probably 0603. Certainly a 1206 wouldn't notice such a pulse. The thermal tau is much longer than 20 ms.

It takes a couple of seconds for 1600 volts to blow a several-megohm

0603 resistor, suspended in free air.

We recently tested various precision (0.05% thinfilm) resistors for power capability, measuring resistance and thermal imaging. Turns out that the pads are dominant heat dissipation mechanisms. Soldered to big slabs of copper foil (which I assume you can't do) an 0603 can dissipate about as much power as a 1206, half a watt at reasonable hot-spot temperatures.

But why don't you just try it? Find the destruct limit and then back off by 5:1 or whatever. We do that a lot; sometimes we decide that we can overload parts a lot (common for passives), sometimes we decide we can't even run them at the mfr's ratings (typically semiconductors.)

John

Thanks..I'll use the ERJ 2512 1W. If space allows, I might resort to paralleling chip resistors. But I'm curious if a current limiter transistor circuit can beat the

2512 (6.4mmx3.2mm) footprint ? At this scale, maybe a transistor is better at handing a power pulse beyond it's continuous power rating compared to resistors..?. I don't know much about hot spots with pulse overloading devices. I'll guess resistor wins.. I'll check this out later. For now a resistor is fine. D from BC

This is certainly food for thought.. Testing time vs trusting data sheets. Heat sinking copper area vs resistor footprint. Survivability vs reliability. Interesting...

I think this is going to be my plan... If I can't squeeze in a 2512 in the pcb design, I'll drop the chip size down to 1206. That way when pcb size goes down...risk goes up.. :)

Based on the above...sounds like a metal core pcb could provide some good pulse overpower tolerability with tiny chip resistors.

Thanks for the thermals.. :) D from BC

Datasheets almost never give typical destruct limits, and usually don't give thermal taus or SOAR equivalents (except for some fraction of power transistors, and not even all of them.)

And blowing up parts is fun.

Really, this modest pulse power is zero risk on a regular thickfilm

1206.

In 20 milliseconds, most of the hot-spot heat will remain trapped in the part, so the extra foil won't help much. The thermal mass of a larger resistor will absorb a shot of joules better than a small one.

John

This can be true for constant power-handling, which has much higher energy-handling capability than the transient energy- handling case.

The issue for short power transients, of course, is the thermal mass of the power-dissipating element, which can be a thin film without much thermal mass. Additional thermal mass located at a distance isolated by a thermal-resistance pathway, necessarily comes into play later. But 20ms is a fairly long time frame as these things go.

Testing in your exact setup is not a bad idea, because the datasheets are often excessively conservative. But you do want to use a huge safety margin like 5x, as John suggested, because of production variability, and because there may be some long-term reliability issues that may not show up in short-term testing. I've seen too many failed SMD resistors in commercial equipment to be sanguine about this issue.

I'd try 2010 resistors before 1206, or use two 1206 parts.

That would be expensive! And likely not so effective for short very time scales.

Any idea what the thermal tau of typical surfmount resistors is like?

I did measure the thermal tau of a thinfilm platinum-on-ceramic RTD, which is probably a reasonable model for a surface-mount resistor. It's easy to fire a shot of energy into an rtd and then measure the resistance versus time, directly measuring the film temperature. I have that data at work and I'll try to find it monday/tuesday and post.

I don't think we've ever seen a surface-mount resistor fail, unless some unusual thing overloaded it, or, in a few cases, it cracked from some stress. They seem pretty tough.

I'd tend to agree. Heatsinking won't help much in 20 ms.

John

I was just doing a *very* similar thing (190 volts, 100uS) and came across a datasheet in my collection. I can't remember where on the Philips site I found it so I've stuck up here.......

Some handy graphs.

--
Gibbo

This email address isn\'t real.

How about:

and a resistor?

--
--
kensmith@rahul.net   forging knowledge

I just mentioned metal core pcbs because I make my own.. I was just trying to find another excuse to use it.. :)

So... it's not very effective to heat sink (by Cu trace or mcpcb) a

2512,2010 or 1206 chip resistor in this case to handle a 26mJ energy burst.. I'll go for that. :) Thanks..

I guess my weak spot in heat dynamics is showing :) All I know is this: Heat has a propagation time. Materials have thermal resistance. Most materials increase in resistance with heat. Materials decompose if spot is hot enough. Heating an object is like charging a capacitor??? At first I didn't quite understand what was meant by tau...It clicked in about 1 hour later... :) Gotta do the google heat review someday... D from BC

Welwyn have some data for short pulses here:

--
-----------------------------------------------------------------------
To reply to me directly:

Replace privacy.net with: totalise DOT co DOT uk and replace me with 
gareth.harris

I noticed a SOT 23 in the table.. Found app note on:

mmmm depletion mosfet I limiting...There's an animal I haven't played with yet..

I'm going to dodge the Supertex reading& math for now... As posted here, I should be ok with a 1206 chip in my app.. A SOT23 + small chip resistor at first guess has a larger footprint than the 1206.

By the way, the resistor I'm looking for is part of a smps startup cct.. But I like my homebrew cct. more than the Supertex sol'n.

Thanks...I might use DMOS someday... D from BC

Join the club. Even if you know the theory, the practical issues - finding reliable data, modeling 3D distributed diffusive systems, estimating reliability vs temp - are still fuzzy. [envision global warming rant here]. At least most thermal conduction systems are linear, sort of.

Yes. There's a direct similarity between thermal and electrical circuits. You can model thermal things with Spice, making the simple equivalences

1 volt == 1 degree C

1 amp electrical == 1 watt thermal

1 farad == 1 gram aluminum

1 ohm == 1 degc/watt

1 second == 1 second

which is good to about 5%. There's no thermal equivalent of an inductor. The bitch is that, while circuits are usually lumped, thermal systems are often distributed, so diffusion equations apply. Yuk.

Thermal time constant, how fast a thing transitions to its steady-state temperature. Can range from nanoseconds to kiloyears. They estimate time of death by how much a body has cooled down.

If a device has a thermal tau of, say 1 second, it will jump about

1/10th of the way to its steady-state temp if you pulse it for 0.1 second.

This is excellent, and free:

John

Cool. Shows an 0805 being good for about 1.2 watts at 20 milliseconds. Sounds conservative to me. But the 1206 data looks strange.

John

Looks like thermal taus in the ballpark of 40 milliseconds.

John

D from BC, if the timescale was a bit longer, the next best thing to increase the rating, for one of, would be to increase the condution by immersion. Beside making the layout with extra big pads, put a drop of heat conducting epoxy over it, and perhaps solder it upside down.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net