Piss Poor On-Board Switching Power Supply - How not to design a Switching Supply

[daveyk]

Reminds me of "What can we remove and it still works" mentality. - lol

Dave

[daveyk]

The datasheet recommends a crowbar detection circuit to turn the lower fet on, shorting the output to gnd.
Definitely a poor design if it's part of something expensive.

And that crowbar detection can not be sent to the chip because I think the chip is the failure.  The control of that lower FET probably could be accomplished with a diode feeding the gate from the detection circuit.

[T3sl4co1l]

Can't say I've worked with LTC1148 before, but it's very old (datasheet says (C) 1993), and the block diagram has... not sure exactly, some combination of current mode control with hysteretic voltage regulation?  Not sure, there's the compensation network in there still, but also a Schmitt trigger?  Oh, or maybe that's the burst mode control, and it's peak-current-mode while active.

It is at least more detail than you'll see in newer block diagrams (blank function boxes galore)...

I don't think it's doing active rectification exactly; I think its operation will be full-wave (alternate N/P conduction, forced CCM -- allows negative inductor current) until topped up, then it sits and waits until Vout falls below threshold.  So, it will gladly run negative inductor current at light load, as opposed to DCM that you'd expect from an active rectifier.  Also the operating frequency, is... fixed off time, sorta?  Depending on things, but who knows.  Maybe they implement frequency foldback that way.  (The plots seem to agree.)

So eh, nothing really suspicious.  At least for normal operation.  As for abnormal, if they're not obeying voltage/current limits, or there's ambient noise getting in, or surge or something (could include hot-plugging transients?), that seems a most likely culprit.  Oh, or maybe the compensation is going wonky because of aging electrolytics, that's always a thing.

Tim

[langwadt]

Can't say I've worked with LTC1148 before, but it's very old (datasheet says (C) 1993), and the block diagram has... not sure exactly, some combination of current mode control with hysteretic voltage regulation?  Not sure, there's the compensation network in there still, but also a Schmitt trigger?  Oh, or maybe that's the burst mode control, and it's peak-current-mode while active.

It is at least more detail than you'll see in newer block diagrams (blank function boxes galore)...

I don't think it's doing active rectification exactly; I think its operation will be full-wave (alternate N/P conduction, forced CCM -- allows negative inductor current) until topped up, then it sits and waits until Vout falls below threshold.  So, it will gladly run negative inductor current at light load, as opposed to DCM that you'd expect from an active rectifier.  Also the operating frequency, is... fixed off time, sorta?  Depending on things, but who knows.  Maybe they implement frequency foldback that way.  (The plots seem to agree.)

So eh, nothing really suspicious.  At least for normal operation.  As for abnormal, if they're not obeying voltage/current limits, or there's ambient noise getting in, or surge or something (could include hot-plugging transients?), that seems a most likely culprit.  Oh, or maybe the compensation is going wonky because of aging electrolytics, that's always a thing.

Tim

the current sense resistor says 0R0402, I'm guessing a 0R resistor in 0402, what's the resistance of that and thus the current limit?

being synchronous could it be pumping current back into the input (with possibly bad electrolytics) and exceeding the max voltage?

[BreakingOhmsLaw]

"buy the book The Art of Electronics right now"

Thank you.  Ordered:

1. The Art of Electronics, Horowitz, Pau
2. The Art of Electronics: The x Chapters, Horowitz, Paul
3. Learning the Art of Electronics: A Hands-On Lab Course, Hayes, Thomas C

You will not regret it. Number 2 is actually quite entertaining with a lot of anecdotes from the authors.

[T3sl4co1l]

the current sense resistor says 0R0402, I'm guessing a 0R resistor in 0402, what's the resistance of that and thus the current limit?

being synchronous could it be pumping current back into the input (with possibly bad electrolytics) and exceeding the max voltage?

I'm assuming the resistor is a typo, or a BOM substitute out of sync with the schematic.  Or hoping.

If it's actually a jumper, typically those are around 50mΩ, so the maximum current sense of 150mV would give 3A, ballpark.  The tolerance on those is obviously quite poor, so too would the current limiting (and compensation, since the modulator gain is proportional to that).

It shouldn't regenerate any, at least not enough to matter; it shuts off at high output so even an external source applied to the output, will not cause input OV.

Tim