RE: SISG failure report

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: <dhmccauley@xxxxxxxxxxxxxxxxxxxxxxxxxx>

Hmmm. . . Thats very interesting.  We had a problem with some TVS for a Beam
Voltage Regulator application used in a high power TWT system and literally
spent upwards of $200k and more than a year resolving the problem
only to be told by a tiger team at Protek that the response time was
directly proportional to the cross sectional area of the device and that we
needed to use larger devices, in this instance 15kW bidirectionals.

Wonder if was just a marketing ploy to get us to buy their products.  It did
solve the problem though moving to the 15kW devices.

On the context of inductance, I can't see how inductance would be greater in
the smaller package.  The loop area is much much less in a small 1.5kW TVS
then in these mondo packages made by Protek.

Dan



Hi Dan,

At 07:37 PM 6/12/2006, you wrote:
>Terry,
>
>One thing that I have found with TVS with my extensive experience in
>designing high voltage supplies and radar transmitters is that the response
>time of a TVS is directly proportional to the power rating of the device.
>This has to do mostly with the surface area of internal junctions and
>devices usually rated for 1.5kW or less, usually do poorly on fast
>transients. They do clamp, however, not without delay and usually at a much
>Higher voltage that rated on the datasheet.

They are "bulk" devices rather than junction devices.  We have used
them in very odd 13.56MHz things...  Inductances is a big killer in
general and we had to go to custom things mounting the bare
die.  They are hexagonal die too (obviously demonic ;-))  "Heat"
allows the "process" to continue and changes them much further like
in the SISG failure case I noted...   They really do sit just there
at the factory and "turn off the oven" just when the voltage is
right!!!  Smaller devices have more "inductance" in the leads...


>This is the main reason I use exclusively 15kW and 30kW TVS devices for my
>DRSSTC work.  I have even measured instances where those 220V 1.5kW devices
>that most people use don't clamp until well over 300V during fast voltage
>transients across the C-E of the IGBTs.

"Inductance", not the device material...  I think "mono" directional
TVSs react in pico-seconds if you really look way too
hard.  Bi-directional gets a lot longer into the nano-seconds.  If
your results differ, it is measurement error cause you did not spend
$250,000 on the test equipment... ;-))  I can't say much since it is
secret, but it all slow things are "outside" the silicon die
level...  Consider the dI/dT and the inductances outside the device
material.   "Leads" are the packaging guy's problem...  Big brick
IGBTs have terrible inductance design...  Greg Leyh has met that
problem ;-))  But when dealing with thousands of amps/uS, you don't
have the option of having 1 inch leads...  We had specified stress
service loops to copper strap to the "die surface" since we had to
control inductance...  It's not the "die"...  We did not put down 30
wire bonds to the die for current, it was "inductance" that we had to
defeat!!!  Many GHz guys "twist around" wire bonds to "adjust"
inductances...   I did too,  I works good :-))


>Of course, the 15kW and 30kW devices are pretty expensive, but they do
>sample!  Of course, you have to speak directly with a sales rep there.

Cut them open and find they are mostly "air"...  It is a big package
sales trick...  We got the biggest MOVs Digikey had...  Cut them open
to see what was inside... Then we got real mad >:((((    "Billboard"
space only!!!  If you need high current fast, you can't tolerate the
"package"...  "Inductance kills!!!"

Cheers,

         Terry


>Dan