TI Smart High-Side Switch Current Limit Mode?.........

[Smokey]

TI has these "smart" high side switches with internal current sensing and current monitors, a bunch of internal protection features, and an adjustable current limit.

A couple examples:
https://www.ti.com/lit/ds/symlink/tps27s100.pdf
https://www.ti.com/lit/ds/symlink/tps272c45.pdf

While it's pretty clear from their plots that the current limit is clipping/regulating the current at the set level, and not a trip-fault that shuts down the device, they are not very clear on the actual mechanism that the thing uses for the current limit. 

They do have a note that reads:
"While in current limiting mode, at any level, the device has a high power dissipation. If the FET temperature
exceeds the over-temperature shutdown threshold, the device turns off just the channel that is overloaded. After
cooling down, the device either latches off or re-tries, depending on the state of the LATCH pin."

Are they actually driving the internal pass FET in linear mode during current limit?  That seems a little crazy considering the voltage and current rating of these devices.  I would expect something more than the above note if that's the case.  Maybe something along the lines of: "If current limit is exceeded, the device turns into a furnace and will immediately overheat.  Care must be taken not to burn yourself." 
If they are chopping the output when the current limit is exceeded, that will also indeed cause higher power dissipation from switching losses, but if this was the case why don't they just say that is what they are doing?

Anyone know for sure how these parts implement current limit?

[uer166]

Are they actually driving the internal pass FET in linear mode during current limit? 

Yes, that is what they do I believe, which obviously has to be done with SOA protection. The overtemp protection has some hysteresis, so on average the temperature shouldn't be at the trip point, but it will be quite toasty at over 100C.

[Smokey]

I haven't actually played with one of these yet, and I'm not doubting their thermal protection actually works.  It probably protects the device just fine.  Although that is a pretty tiny package to be actively dissipating a bunch of heat.

If they actually are driving the FET in linear mode, it would be really interesting to play with and see how it acts.  One question I would have is around what they are doing with the gate voltage (which of course is impossible to tell since it's buried in the case).  In normal switch-type operation you would want that FET fully on, but you need a current loop regulating the gate voltage to keep it linear during current limit mode.  I wonder if they are keeping the gate voltage just barely above full on so its faster to transition back to the linear current loop.

[Smokey]

Ha!  From the data sheet: https://www.ti.com/lit/ds/symlink/tps272c45.pdf

"Sometimes, systems require high inrush current handling as well (example incandescent lamp and capacitive loads). By integrating a dual stage current limit, the TPS272C45 enables robust DC current limiting while still allowing flexible inrush handling."
"The ILIM1 and ILIM2 pin resistors set the current limit thresholds for CH1 and CH2 respectively while ILIMD pin resistor sets a delay time for the device to operate in a higher or lower current limit during device start-up or output turn-on by retry after a fault (thermal) shutdown)."

How funny would it be if they realized their current loop gate voltage would be all wound up at full-on levels and would need some time to settle, so they turned that into a feature.  "In-rush Delay!"

[Smokey]

Oh, duh.  Sorry, I've been working on high current relay driver circuits all day.  Who cares if you PWM into a relay coil. 
These high side switches are powering all kinds of stuff that I don't think would appreciate it much if you started chopping their supply voltage.  I guess it makes sense they are doing linear current limit control.  Even though that is going to drop the effective supply voltage anyway which I don't think devices would like much either.  At least it's not interfering with SMPS regulation or something.

Nevermind.  All is well I guess.