Damn, burned out my 100-watt 50-ohm resistor!

On Thursday, August 18, 2016 at 5:49:27 PM UTC-4, Winfield Hill wrote:
> dcaster@krl.org wrote...
> >
> >On Thursday, August 18, 2016 at 5:08:42 PM UTC-4, Winfield Hill wrote:
> >
> >>  Next I'm going to tear up the PCB layout, double
> >>  the power capability of the heatsink + fan, and
> >>  use two of the 100-watt resistors in parallel.
> >
> > Still confused.  ...
> >
> > Do not worry about replying.  I am sure you found a better
> > way like using 100 ohm resistors.
> 
>  Right, although 100 ohms is the highest value available
>  and there's low inventory at Digi-Key, two 25-ohm in
>  series is better.  I'm beginning to like these guys.
> 
> 
> -- 
>  Thanks,
>     - Win

Do those have a ceramic back?  
We had issues with production over torquing the mounting 
screw and cracking the ceramic... split lock washers
were also removed and replaced with expensive belleville.
(besides the torque wrench/driver.)

George H. 


 

George Herold wrote...
>
> Do those have a ceramic back?  
> We had issues with production over torquing
> the mounting screw and cracking the ceramic...
> split lock washers were also removed and
> replaced with expensive belleville. (besides
> the torque wrench/driver.)

 Oh, boy, yes they do.  I've just been using
 a screw w/o anything, and making it tight.
 Sounds like trouble brewing.  What if I'm
 tightening down on a flexible Sil-Pad?


-- 
 Thanks,
    - Win

>> Do those have a ceramic back?  
>> We had issues with production over torquing
>> the mounting screw and cracking the ceramic...
>> split lock washers were also removed and
>> replaced with expensive belleville. (besides
>> the torque wrench/driver.)
>
> Oh, boy, yes they do.  I've just been using
> a screw w/o anything, and making it tight.
> Sounds like trouble brewing.  What if I'm
> tightening down on a flexible Sil-Pad?

You may want to use something other than just a screw through the
resistor, so that the pressure on the top of the device is spread
out.  Possibly a large washer?  Some people prefer to avoid using the
screw entirely, and like to clamp the device to the heatsink... run a
flat, rigid metal bar across the top of it and screw down the ends of
the bar, or use a clamping spring.

If you do use a screw, respect the device manufacturer's torque
specification for the part.

If I recall correctly, Sil-Pads have rather poor thermal conductivity,
compared to some other solutions.  They might reduce pressure stress,
but make it harder to keep to the "25C case temperature" part of the
rating spec.

On Thursday, August 18, 2016 at 8:50:17 PM UTC-4, Winfield Hill wrote:
> George Herold wrote...
> >
> > Do those have a ceramic back?  
> > We had issues with production over torquing
> > the mounting screw and cracking the ceramic...
> > split lock washers were also removed and
> > replaced with expensive belleville. (besides
> > the torque wrench/driver.)
> 
>  Oh, boy, yes they do.  I've just been using
>  a screw w/o anything, and making it tight.
>  Sounds like trouble brewing.  What if I'm
>  tightening down on a flexible Sil-Pad?
> 
> 
> -- 
>  Thanks,
>     - Win

I wouldn't use a sil pad, unless there was some voltage 
isolation involved.  Belleville washer and proper torque
was caddock's recommendation.  If you are doing them all,
you could do the torque by hand.  Production people like 
to make things "tight", and some of them are strong as an ox.

George H. 

George Herold wrote...
>
> I wouldn't use a sil pad, unless there was some voltage 
> isolation involved.  Belleville washer and proper torque
> was Caddock's recommendation.  If you are doing them all,
> you could do the torque by hand.  Production people like 
> to make things "tight", and some are strong as an ox.

 Thanks, George!


-- 
 Thanks,
    - Win

On Thursday, August 18, 2016 at 2:05:10 PM UTC-4, Winfield Hill wrote:
> A major meltdown last Friday for my 8ns 1kV SiC MOSFET
>  pulser.  After set of tests at 400V and 4MHz repetition
>  rate, with everything working fine, I had to increase
>  the frequency.  A loose hand slip and I was at 10MHz. 
>  That's when the smoke appeared.  Blew out a 1-ohm gate
>  resistor, both TO-247 MOSFETs, both 24V gate drivers,
>  the flying regulator and the 150Mbps 50kV/us isolator,
>  and a 0.1-ohm current-sense resistor!  Plus a scorch
>  damage region on the PCB.  In the painful process of
>  getting it going again took out another gate driver. 
>  (No more stock now for those.)
> 
>  So yesterday after Tuesday's rebuild, I took it slow
>  and easy, checking temps with the Flir IR camera as I
>  went, looking for the operating limits.  Can do 10MHz
>  in long bursts, check, 8MHz continuous with HV off,
>  check, 5MHz at 500V, nice. 
> 
>  500V, 5MHz, 5A square waves anyone?
> 
>  OK, saved scope screen shots, turned everything off,
>  went home.  Now I find the burned out 100-watt 50-ohm
>  non-inductive resistor, mounted on 100W heatsink + fan.
>  MP9100, "Caddock's Micronox® resistance film fired
>  onto flat ceramic substrate, 100 Watts at +25°C Case".
>  (Need 50-ohm back termination to drive output coax.)
>  Sheesh, but thankfully my 150-watt attenuator is OK.
>  Those are expensive, even on eBay!
> 
> 
> -- 
>  Thanks,
>     - Win

Now you know why they have such things as interlocks and overstress protection circuits built into the high power stuff.

On 18 Aug 2016 11:04:56 -0700, Winfield Hill
<hill@rowland.harvard.edu> wrote:

> A major meltdown last Friday for my 8ns 1kV SiC MOSFET
> pulser.  After set of tests at 400V and 4MHz repetition
> rate, with everything working fine, I had to increase
> the frequency.  A loose hand slip and I was at 10MHz. 
> That's when the smoke appeared.  Blew out a 1-ohm gate
> resistor, both TO-247 MOSFETs, both 24V gate drivers,
> the flying regulator and the 150Mbps 50kV/us isolator,
> and a 0.1-ohm current-sense resistor!  Plus a scorch
> damage region on the PCB.  In the painful process of
> getting it going again took out another gate driver. 
> (No more stock now for those.)
>
> So yesterday after Tuesday's rebuild, I took it slow
> and easy, checking temps with the Flir IR camera as I
> went, looking for the operating limits.  Can do 10MHz
> in long bursts, check, 8MHz continuous with HV off,
> check, 5MHz at 500V, nice. 
>
> 500V, 5MHz, 5A square waves anyone?
>
> OK, saved scope screen shots, turned everything off,
> went home.  Now I find the burned out 100-watt 50-ohm
> non-inductive resistor, mounted on 100W heatsink + fan.
> MP9100, "Caddock's Micronox&#4294967295; resistance film fired
> onto flat ceramic substrate, 100 Watts at +25&#4294967295;C Case".
> (Need 50-ohm back termination to drive output coax.)
> Sheesh, but thankfully my 150-watt attenuator is OK.
> Those are expensive, even on eBay!

People are making surface-mount resistors up to 200 watts

http://www.emc-rflabs.com/Passive-Components/High-Power-Terminations/Chip-Terminations/SMT-Surface-Mount

but not on my circuit boards!

There are flange-mount versions too

http://www.emc-rflabs.com/Passive-Components/High-Power-Terminations/Flange

Of course, 1KW into 50 ohms is only 220 volts RMS. They may not be
good for kilovolt pulses.

The author of AoE can surely get samples to try.

I have a potential application for a 6KV or so high-current pulser. I
looked at SiC but they seemed to have high gate resistances, which
would make them slow. GaN is too low voltage, so I'm left with
stacked-and-racked mosfets, with insane gate drivers.

Doncha just hate it when some mistake blows everything up?


-- 

John Larkin         Highland Technology, Inc

lunatic fringe electronics 

On Thu, 18 Aug 2016 11:20:34 -0700, Jim Thompson wrote:

> Back in my amateur radio days (K7ZAE, 2M, ~1964), I made a dummy load
> with two sheets of FR-4 and 40 2K/2W/Composition resistors in parallel
> between the two sheets (and a fan of course).  Gave an outstanding good
> VSWR at 144MHz:
> 
> =================
>   _|_  _|_  _|_
>  |   ||   ||   |
>  |   ||   ||   |
>  |_ _||_ _||_ _|
>    |    |    |
> =================

Thanks for providing the little diagram, there, Jim. I was finding it 
kinda hard to picture parallel resistors. ;->

John Larkin wrote...
>
> I have a potential application for a 6KV or so
> high-current pulser.  I looked at SiC but they
> seemed to have high gate resistances, which
> would make them slow. GaN is too low voltage,
> so I'm left with stacked-and-racked mosfets,
> with insane gate drivers.

 The SiC transistor I'm using doesn't show the
 gate resistance in the datasheet (other manuf
 mention 10 ohms, etc), but I had to lower one
 gate resistor from 4.7 ohms to 2.2 ohms, to
 get my 7ns rise/fall time, so apparently it's
 not high enough to badly damage the risetime.

 I also got better results with +16V, -3V gate
 drive rather than the recommended +20V, -4V.
 The gate plateau was +10V.  So I could have
 used a 20V driver IC instead of a 36V part.


-- 
 Thanks,
    - Win

On Thu, 18 Aug 2016 11:04:56 -0700, Winfield Hill wrote:

[...]

I wouldn't worry about it, Win. I blow things up on a daily basis. 
Something went **bang** only a couple of hours ago, actually (stray crock 
lead hit a 240V live line). Fortunately the neighbours have got 
accustomed to all the explosions and fires over the years and the police 
don't even bother to call any more.