Recent searches
Search options
ADI / Linear Tech LTC4367 power protection controller, from one of my failed 48V IBC designs that blew up from inductive spikes when I applied power.
I didn't do extensive circuit analysis but it looks like one particular net is blown out in several places. Guessing this is the same pin that cratered on the package.
Besides the damage, some other features of interest:
* Colored fringes on large top metal features. Maybe spin-on glass top metal or something.
* The LTC logo and 4367 part number at the 6 o'clock position are attached by little metal tracks. Maybe a DRC rule about no free-floating metal features below some size to avoid the risk of them floating off or something?
* A few dozen gates of standard-cell logic in the 3 o'clock area
Full res (M plan apo 100 HR): https://siliconpr0n.org/map/analog-devices/ltc4367/azonenberg_mz_mit100x
Pin 1 was blown out on the package if anybody wants to use that as a hint when doing circuit analysis. According to the last data I got before blowing out my scope frontend with unexpectedly high dV/dt transients, the inductive kick from the common mode choke on the nominally 48V input spiked to over 100V.
@azonenberg that tracks my experiences with the part. VIN blows up and GATE sometimes blows up. oddly, even with VIN blown, i was able to power a part from VOUT and the charge pump on GATE was working.
i really don't like this part.
@tubetime I mean, it was partly my fault putting a common mode choke on the input and not having any kind of snubber to absorb transients when you applied power.
But I did also switch to the TPS16630 for the next version of the design (as well as adding a TVS, zener, small ceramic cap, ferrite bead, another zener, and small electrolytic cap upstream of it to fully suppress transients)
@azonenberg @tubetime (Why did the common mode choke make things worse? I thought they were AC-blocking, thus would _suppress_ transients.)
@cr1901 @tubetime They're inductors, they fight changes in current.
When you first apply power the parasitic capacitance of the cable charges up to 48V, some nonzero current flows to make this happen.
Then it hits the 4367 and stops. So you get a flyback spike that went upwards of 100V and fried the VIN pin of the 4367.
It makes sense in retrospect but I didn't see it coming when I did the design.
@azonenberg @tubetime I got nerdsniped into wanting to solve for the step response of this circuit, which I haven't done in like 15 years.
But I recall an pure LC circuit can at max double the input voltage. So I don't understand how you got over twice the voltage spike at Vin (i.e. my simplified model is missing something).
@cr1901 @tubetime I was using an R-C divider probe so I'm not sure of the exact peak, the response wasn't flat. I was not expecting such a big spike and the dV/dt was sharp enough to fry the scope frontend through the 10x R-C probe so any data gathered during that process is inherently suspect.
But it was "substantial" voltage.
@azonenberg @tubetime Yea, thinking about it more, Idk where I came up with that "max twice the voltage" thing. I'm probably misremembering and need to review :'D...
Intuitively, I understand there will be a voltage spike, but I have trouble quantifying how big the spike will be in the context of a larger system.