Exploding IC in buck converter

[Siwastaja]

You may want, for learning if nothing else, to look at solutions where the voltage limiting element is in series with your load, not in parallel.

The parallel configuration tries to clamp the full input supply capability, such as a 100A alternator. I.e., your clamp is dissipating Vdiff * Isupply_max, where Isupply_max is limited by the resistances in the system. Adding a bit of extra series resistance before the clamp helps, but if you add too much, your load starves under normal operation; this resistance is always there, want it or not.

The series configuration only needs to dissipate Vdiff * Iload, where Iload is the actual consumption of your load. Basically, this would be a pass transistor with high enough voltage rating. Yes, it is more complex.

It sucks when you choose a voltage regulator IC without enough voltage margin. I once chose a 26V max recommended, 30V abs max buck converter IC for a system with a regulated 24V bus. Works great. Until I needed a -10V output as well, and "simply" BOM reused the same part to form an inverting buck-boost, without realizing the part now sees 34V . Things like this happen in a hurry. The problem was made worse by the fact that the few units in existence worked just fine; one blew up when a customer wanted to use a 26V bus voltage instead, finally revealing the problem. 36V was finally enough for this 30V abs.max. part.

[Teun]

Alti, I usually detect sarcasm quite well but now I have to ask, are you serious or joking?

I have been telling him I don't always agree with things for almost 2 years now. But I guess you should know him to know the whole situation here.

Anyways mzzj is right. Alternator voltage can be over 30V on 24V systems.

[T3sl4co1l]

And to generalize that -- when the load resistance is already above the generator resistance, it's better to stay on the high side, i.e. reduce the load by merely maintaining load current.  When below, it's better to stay on the low side, i.e. increase the load by maintaining load voltage.

(We might measure the load in terms of average load resistance: instantaneous V / I.  Note this is different from the AC incremental resistance of the load, and is simply the ratio of numbers measured with a meter, no fancy differencing needed.  Note also a "heavy" load is a low resistance, because we always speak of load in terms of conductance, but resistance is reciprocal, go figure.)

A linear regulator passes the same load current, while dropping the excess voltage; average input resistance is thus proportional to peak voltage, and so rises during the surge.

A TVS passes the same load voltage, while drawing excess current; average input resistance is thus inversely proportional to peak current, and so falls during the surge.

Note that a switching converter does even better than these possibilities, because as voltage goes up, current drops roughly proportionally; it has a constant-power characteristic and doesn't care what the input voltage is, as long as it's within operating range.

A switcher will still have some resistive characteristic remaining, for example switching loss tends to be significantly higher at high voltages.  This can be designed to be a fraction of total load power, making it still much easier to handle than the power absorbed by a brute force (series or shunt) method.

Load dump generator resistance is generally tested at 2 ohms or thereabouts (depends on the alternator ratings in the target vehicle), so the decision point is at a fairly low resistance, and lighter loads like this will graciously choose the first option.

I suppose very low loads, like audio power amps, might opt for a shunt method; with a customized switching controller, they might even try and turn on both switches simultaneously, shorting out the supply and clearing the fuse.  A more subtle case would be partially turning on the switches, using them as a power shunt regulator instead.  Though, transistors aren't good for absorbing much energy, so this probably wouldn't work well until you get into the really big (multiple kW?) sizes, which can pretty much absorb the full pulse anyway without special effort.

Which, the peak available power is in the kW -- so, anything in that middle range of impedance, and not big enough to just grin and bear it, can save a huge amount of trouble just by shutting off completely.  Which can still be done through a series or shunt method -- in series, the switch opens, drops all the peak voltage, and load current drops to zero; in parallel, the switch closes (crowbars), sinks all the peak current, and load voltage drops to ~zero (which eventually turns to the series case, as a fuse must open -- there will still be a 12V baseline after the surge, and we can't have that crowbar staying on forever!).

Tim

[Alti]

I have been telling him I don't always agree with things for almost 2 years now. But I guess you should know him to know the whole situation here.

I do not want to push this subject off-topic - I am no automotive expert. Also, I have no idea if the actions taken by Mr. Head (your boss) are wasteful or maybe actually you are the problem in this relationship. Or maybe your boss has his own Ms. Finance Department who is analyzing revenues giving him no choice but cut BOM cost. As long as you do not own shares - this is irrelevant.

What I point out:
1. You take no responsibility for the decisions of your boss.
2. Disobedience is the worst path one could follow. This will only lead to a frustration from both sides. The project timeline will slip and guess who is going to be blamed?

Dear Mrs. Finance,
I requested Teun to use component X and instead he decided on his own to use component that just went unobtanium because of Covid.
We are f$#%*.

Sincerely,
Dick Head

[Siwastaja]

When seen unnecessary micromanagement, I have had good results by playing along, being nice, and then, just doing The Right Thing based on actual engineering principles (like the simple calculations shown here). This pattern is easy to apply in trivial cases such as maximum voltage ratings of parts. Of course, if the manager already bought the wrong parts, then this is gonna suck.

I have never seen a case where the micromanaging boss remembers what their exact (faulty) instructions were, and you would be in trouble because you made a working device in a "wrong" way. Not saying it couldn't happen, just that I haven't seen it. Usually everybody's happy when the device works, and not happy when you have a recall. Blame shifting starts when things go wrong, and they go wrong when people don't catch the mistakes.

What is 100% sure is that exceeding the absolute maximum voltage ratings of a part severely will 100% definitely lead into troubles, recalls, and waste a lot of time, and while documenting that the blame is due to a micromanaging boss is better than not documenting it, even better is to avoid such situation completely. You can do that politely, without wasting anyone's time, and everyone's happy.

Such engineering calculations do not take longer than handwaving a design with random components, and the amount of time saved by getting the design right is immense, compared to reaching a working solution through iteration with random parts. That is exactly the job of the engineers.

Like aircraft have two pilots co-operating, taught to catch other's mistakes and correct things. This is the result of recognizing that following the one with the authority just doesn't work. In general, you should give everyone the benefit of doubt, and discuss such things openly; such engineering things are just normal daily science, shouldn't be a matter of feelings, authority, or opinions. If this seems to constantly fail with a certain person no matter how you try to handle the situation, I strongly suggest looking for a job elsewhere. Not everyone sucks.

It's quite typical that the ones obeying the orders are the ones who become scapegoats later on. You don't need to participate in that game.

[Alti]

Blame shifting starts when things go wrong, and they go wrong when people don't catch the mistakes.

Naaah, "Exploding IC in buck converter", I am sure this is not the case.