TVS breakdown voltages for higher currents

to handle that situation- none of which are single component unfortunately. And a load dump can put more than 1-2A into a shunt- way more.

A transistor is always an option, normally, but here I only have the space of an SMA package. A shunt device would need to be beefy and a series device (which can be smaller) would need very snappy turn-on/off and thus some active electronics up front. Plus I'd really want a shunt because the caps are also rated 100V. Of course they will stand 1.5x at the least but in this design exceeding abs max is definitely not considered kosher and would be flagged.

to handle that situation- none of which are single component unfortunately. And a load dump can put more than 1-2A into a shunt- way more.

Make a teensy "daughter" board? ...Jim Thompson

MOSFETs to handle that situation- none of which are single component unfortunately. And a load dump can put more than 1-2A into a shunt- way more.

There is already one, full to the brim :-(

On a sunny day (Mon, 24 Sep 2012 09:22:25 -0700) it happened Joerg wrote in :

Sometimes by just clampoing the positve peak you can dissipate enough of the swinging energy, the diagram that somebody else showed here, transistor zener, but add some capacitance to kill the positive side swing?

------------ | | zener | | | C | not complete, you get the picture, 100% capacitive behaviour about Vx | c ---- b e | /// Slimulate some

about Vx

That would require more intelligent circuitry than some passives on the base. Has to be more snappy. Because else the turn-off and possibly also turn-on will be too sluggish, SOA get exceeded, and ... *PHUT* :-)

Things like that are often better done with a FET.

I can't do that here, If I used a FET it would need to be driven fast and snappy.

Here is the pickle, quote "Note, even though the power-transient thermal resistance indicates reasonable junction temperatures, the device still may fail if the peak current exceeds certain values" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

That would be really nice to have in a datasheet. I am more on the left side of figure 18. But such figures are typically not part of the datasheet of SMA-sized zeners, and aren't considered guaranteed data unless it clearly says so for a worst case limit.

Reference to zeners was:

- tolerance in driving a fet gate load.

- energy absorption capabilities for simple parts.

Look for ON Semi HBD854

RL

It's more like 10-ampish but only around 50-60usec. The brief resonant spike is all I need to muffle, not the whole 100msec of load dump. The load dump is fine, the circuit handles that voltage without flinching. Even for hours if need be.

More like hi-rel stuff, so no cutting corners.

Oh, they do hold up well. The problem is the huge tolerance and poor specs. A 78V TVS won't clamp any serious amps before it gets to well above 100V, and that's too much here. So I need a TVS that is guaranteed to clamp hard (as in lots of amps) at less than 100V but will not conduct more current than it can handle without becoming hot.

I noticed you only need 1 amp in a earlier post and you are cramped for space.. How about using a TVS diode that will fit your space and rail supply wire to the board that just meets the current requirements. The rail lead can easily absorb that for that short of time.

I don't know if you're making consumer fly by night products or mill spec however, if not mill spec, that would be my resolve!.

If you look at a lot of the cheap automotive electronics, that is exactly what is done. The lead wire for the (+) terminal is just big enough to power the device and any transients will get suppressed with out over taking the clamping components with in the device, as long as they remain short lived.

btw, I use 1.5k series type TVS diodes for small medium pulses and they seem to hold well.

Jamie

On a sunny day (Mon, 24 Sep 2012 17:50:32 -0700) it happened Joerg wrote in :

I dunno WTF you are doing, and from lack of data perhaps sometimes I think you do not eitehr, but if 'all you need' (copy right beatles) is a peak limiter, why noit use an emitetr follower wit hzener in teh base?

PNP or NPN depending on Polarity INPUT ------------ c e ------------ OUTPUT | resisister b |____________| | zener or avalanche or whatever | ///// ground or earth or water or whatever

BTW I compeltely disagree with you otewhr remaork: 'that si betetrdonem wit ha FTE'. Either you have no clue what I ment, or about other things, Transisistors are cool because of their .7 (or .15) V Vbe on level, this can be made to construct a very accurate ZENER subsiustiurte wit 2 resisistors.

INPUT ------------------------------------ OUPUT | | resisistor c | |---------------------- b Silicone filling | other resisistor e | | ///////////////////////////////// center of earth the ratio .7 x (resisistor divided by other resisistor is about approxamately close somewhere in the order of very near to the LIMITING voltage where transisistor starts heating up.

This saves you board space.

If the joule/sec rating is to be conformed to, for a specific voltage of part and a specific duration, passing certain average current levels, to comply, is unavoidable - unless there's some other method of forcing energy into the part.

If your peak is less than this average, then whaty is your fear based on? Lack of bench testing?

Rob

It is based on the fact that there are usually no such ratings in the datasheets:

http://61.222.192.61/mccsemi/up_pdf/SMAJ4728A-SMAJ4764A(DO-214AC).pdf

Some have surge power but not for pulses under 100usec and they won't say what the zener's voltage will be at during such surges (which is the whole reason for this thread):

http://61.222.192.61/mccsemi/up_pdf/SMAJ4728A-SMAJ4764A(DO-214AC).pdf

Think about what this will do when the voltage is close to but not quite at the zener voltage. Think about what will happen if the regular 1A operational current flows through there, what resistor value you'll need to keep the BJT deep in saturation so as not to melt down. When spikey-spike roll along it has to come out of saturation within very few microseconds, something a power BJT will like as much as I like pea soup.

FTE'.

You are drifting towards Swahili :-)

And what's wrong with being in disagreement on occasion?

resisistors.

approxamately close somewhere in the order of very near to the LIMITING

Which will turn into a smoldering hole when the transistor comes on too slowly. Then there's saturation, the turn-off delay and such fun stuff. Problems that FETs don't have. It is one of the reasons why BJTs are no longer used much in switchers. Last time I did was 19-sumpthin'.

P.S.: One also has to think about what happens if the input goes low and is very low Z but your caps on the ouput side are still charged ... *PHUT*

I just had that situation here a the lab, courtesy of a 3min power outage. Nothing happened because there was a FET in there and those have a sturdy body diode. One of the many reasons why I prefer FETs.

r whatever

you can an NPN with a build in diode, something like:

-Lasse

This wasn't the issue with the originally quoted SMAJxxx series from General Semiconductor/Vishay. The issue was that there could be some unanticipated waveform with an 80V peak that caused excess dissipation in the SMAJ78CA.

Neither the 78CA or 75CA versions of this part demonstrate >1mA conduction at 80V. The 25A clamping voltage of either is 126 and 121V respectively, so they will not do what you want at 100V. Energy absorption is graphically depicted between 100nS and 10mS.

By taking the zener large scale characteristics out of the equation, normal tolerance and slope resistance can be used to tailor other clamping (or disconnection) methods.

Motorola quotes empirical surge characteristics of it's zeners in figure 18 on page 52, based on years of experience producing the generic axial slugs, in the quoted app note. The fact that these parts are no longer manufactured domestically by Motorola does offer some pause for thought about what they will actually do today.

I don't see that accurate bench testing can be avoided, using a wider range of potential components and circuits than you are currently restricting yourself to. An accurate threat simulator ( based on the application) that can adjusted for overstress to failure would obviously come in handy, if you're attempting to establish margins.

Your circuit application may not be generic enough for somebody else to do this for you on a regular commercial basis, selling you the value-addd part. That's just the way it is. That's why they pay you the bucks - not to make false reasurances, but not to spook easily either.

RL

Where do they guarantee that in here?

http://61.222.192.61/mccsemi/up_pdf/SMAJ5.0~SMAJ440CA(SMA).pdf

I can't base a design like this one on bench test results, there need to be values stated by the mfg.

If I had the space, sure, then I'd make an active clamp.

I don't easily spook but in the hi-rel world I know where the limits are. Winging it or even bench testing doesn't cut it there. I am pretty positive that I will get answers from one of the mfgs. But have to get back to the office first, tough to do that from the road.

General Semiconductor / Vishay (since 1975).

Parts are all tested at It. This shows up on the MCC data sheet, as well, but not as a 'test value', or with other definitions. This is often a sign that a data sheet is cribbed incompletely from another source. Microsemi called it Ibr in their data sheet. Microsemi and Gen Semi pulse current for maximum clamping levels was actually below 5A.

I'm not sure that Micro Commercial Components is actually a manufacturer - it's been possible to out-source the supply of these part numbers from other original manufacturers for some decades. It was often a requirement for house part number sales - that these part numbers be multiple-sourced - generating a raft of new mfr/distributors with data sheets.

You have to validate them in the application....... Just because they claim to sub, doesn't mean that they will. A data sheet is an advertising material warranty for replacement of defective materials, not a contract. Parts often exhibit non-advertised characteristics, both good and bad.

If you want a mfr to do it for you, you'd best get used to waiting for it.

RL

I know, I have plowed through pretty much any mfg's datasheets. As mentioned before, I need this data at It=10mA or thereabouts. Also higher if available. If they have a good handle on the process it should not be a problem to furnish such data.

A datasheet as a formally released document with min-max values listed in a table does carry legal responsibilities. Just like other published documents such as fuel economy statements do (one car mfg found our the hard way).

We'll see. I was plastered with other work this week but I've had a similar need with a zener a few years ago. The wait time for hard data (QC limits to which they test) wasn't too bad, it all arrived here on my desk within a week or so. Problem is, engineering support in many companies has shriveled to some extent, probably due to budget cuts, the economy and so on.