Good practice for meeting input voltage spec

[okw]

Hard to write a proper subject for this thread. But I am designing a 1 cell li-ion powered thing, that shall charge directly from the 12V of a running (and charging) car. It's typically ~13.8V while charging, but many models (especially older ones), can be higher (15-16V). Automotive also have a lot of spikes, load dumps, etc, but I've have implemented input protection already.
I only need a 3.3V power rail, and this is supplied by either by a 12V->4.2V buck-, or the li-ion battery-. The selection battery/buck is handled by a simple mosfet/schottky diode power path. The power path then goes into a linear 3.3V regulator.
Previously I used 5V buck which fed a linear TP4056 charger, but then I found the BQ25606. It has several benefits (besides being cheap, reliable/good manufacturer and doesn't require a lot of circuitry). It's a switch mode charger, so more effective, and it has a built-in power path, so I can eliminate both the 4.2V buck regulator and the external power path. Seems I can tap into the SYS pin, to power my 3.3v regulator from this, and the charger takes care of everything else.
Would it be correct (as intended) to use the SYS like this?

However, it shuts off >13.5V (abs max rated at 22V). I had 10 assembled evaluation boards made before I discovered the 13.5V limit (the empirical way).
Is it bad practice to drop the input voltage a bit, instead of finding a more expensive chip?
I assume I set the battery charging to somewhere between 500 and 1000mA, and the 3.3v regulator will supply 400mA at the most.
Which methods are most appropriate/effective/safe for the dropping?

[moffy]

Just use a suitable voltage regulator or low drop out regulator, they have ranges just for automotive use.

[okw]

I'm trying to save power, and a linear regulator would generate an additional 4W of heat when charging (dropping 1A from 16V to 12V). And such a regulator would need to be quite big to dissipate the heat, and I have very limited space.

[jonpaul]

See Alternator Load dump transient protection.

Up to 78V, 100 mS many joules.

Needs a specail large die area TVS like old MR2535 across the 12V bus.

Jon

[okw]

True
Would a SMBJ16A suffice?

[jonpaul]

No on the SMD diode.

See old poosts same issues

https://www.eevblog.com/forum/projects/load-dump-circuit-reviewimprovements-suggestions/

Good ref

https://www.littelfuse.com/~/media/electronics/application_notes/littelfuse_tvs_diode_meet_automotive_load_dump_standard_application_note.pdf.pdf?la=en

Jon

[okw]

This design will actually be on a motorcycle, not a car (just said as many are more familiar with electrical systems on cars).
Motorcycle alternators are much smaller than the ones found in cars. Typical batteries are 6Ah (compared to 55-120Ah on cars). Do you know if also the load dump is smaller, so I can use smaller sized components? Space is already a premium on motorcycles.

[voltsandjolts]

I'm sure the alternators and transients on Puch 50cc and a Gold Wing GL1800 differ considerably

[okw]

LOL! Good point. Goldwing spaceships :p

[2N3055]

Look up automotive spec regulators and connected application notes and whitepapers.

[Psi]

Unless your product needs to use large currents you can always add a series resistor on the power input of your entire device to limit the current that the TVS has to deal with. That can relax the power requirements for the TVS a bit.

Keep in mind the more aggressive you make your protection the sooner it will kick-in and try to handle more of the load dump current than all the TVS diodes in all the other devices also connected to the 12V rail.
So you want to, where possible, limit the current share you have to deal with.

Also, the TVS can be a SMT device, just needs to be big, you can get SMC ones that are fine for automotive.

An automotive 5V regulator like the LM2937IMP-5.0 can handle rev battery and transient load dump surges up to 60V -50V minimum (75V typ).
So any extra TVS you might want to add only needs to clamp to under 60V / -50V, which means it only has to clamp the very top of the 75V pulse which is much easier to deal with.

[Siwastaja]

If the thing is a cheap gadget, no need to design to survive load dump (as long as it does not catch fire). Load dump issue is exaggerated by many orders of magnitude; it's an event which classically kills all your headlamps etc. 95% of car owners never experience one.

Now if you decide to handle load dump, don't try to shunt it away because it's a long event, hence you have to dissipate a lot of energy. Instead, easiest way would be to pick a switch mode converter with maximum input voltage specified to 100V or so. Result - the thing just operates through the load dump!

[okw]

Unless your product needs to use large currents you can always add a series resistor on the power input of your entire device to limit the current that the TVS has to deal with. That can relax the power requirements for the TVS a bit.

At most, it will charge a 2.5Ah Li-ion battery (bq2xxxx charger from Texas, rate 500-1500mA, but mostly just maintaining as the battery will almost never be used). Along with powering a few hundred milliamps for the rest of the circuit. But for the few times it operates solely on Li-ion, it has to conserve power almost IoT-like. So I've gone through great lengths to find suitable a buck and a linear regulator that is very efficient (both in datasheets and by my own experiments).
What is a good starting point for the input resistor (resistance and effect)?

Also, the TVS can be a SMT device, just needs to be big, you can get SMC ones that are fine for automotive.

Most bq24xxxx (charger ic) are 28-30V abs. max.
Could I first TVS clamp the high dumps (60-75V), then clamp to e.g. 20V?

An automotive 5V regulator like the LM2937IMP-5.0 can handle rev battery and transient load dump surges up to 60V -50V minimum (75V typ).
So any extra TVS you might want to add only needs to clamp to under 60V / -50V, which means it only has to clamp the very top of the 75V pulse which is much easier to deal with.

It's linear, so it wastes a lot of energy. I prefer to make it as efficient as possible (motorcycle batteries aren't that big).

[okw]

If the thing is a cheap gadget, no need to design to survive load dump (as long as it does not catch fire). Load dump issue is exaggerated by many orders of magnitude; it's an event which classically kills all your headlamps etc. 95% of car owners never experience one.

It's not gonna be cheap at all, so I prefer to design it as good as possible.
The 12V is only needed for charging the internal li-ion. During surges, 12V can be absent (product automatically switches over to battery in absence of 12V).
That might make it easier, than to operate through the surge?

[Jon_S]

It's not gonna be cheap at all, so I prefer to design it as good as possible.

In that case, possibly consider one of TI's SurgeStopper devices. I have had a lot of success with them. A large portion of the range fully enhances a high-side N-FET, so you can get excellent performance. At higher currents it becomes very tricky to meet the SOA with the clamp-type ones and the over-voltage-disconnect variants are a better idea.

The old MELF-and-TVS approach can be extremely effective and reliable though! 

 

[Wolfram]

I would start with a buck converter with a 40 or even 60 V input voltage rating, to make surge clamping easier and cheaper. I have good experiences with the AP64xxx series, the 200 and 352 could be good candidates. Regulate the incoming voltage down to 5 V, then you can power both the battery charger and 3.3 V LDO from that. One more stage of conversion, but it would likely lead to a cheaper overall solution if you want to make it robust enough to handle automotive-level load dump. Don't forget polarity protection as well

[Siwastaja]

The 12V is only needed for charging the internal li-ion. During surges, 12V can be absent (product automatically switches over to battery in absence of 12V).
That might make it easier, than to operate through the surge?

Might or might not. All you need is some 100V rated semiconductor switch (MOSFET or load switch IC) which can switch off quickly, and some logic to drive it. My point is, a 100-V rated switch mode converter has this switch as its main switching element, and can demonstrably survive input voltage rising quickly to 100V, which is normal operation for it. It actually operating through the surge is therefore just a (maybe not important, but positive) side effect. The reason to do so would be ease of design. Although the switcher's capability to run through such wide Vin range must be confirmed, specifically minimum duty cycle.

[okw]

I agree. An automotive (40-60V tolerant) buck is probably the way to go. I found this to be quite good: AP64352QSP-13.
12V -> 5.0V, then to the PMIC (either bqQ25606 or bq24195L).
How is it to power a buck PMIC with a buck? Oscillations / instability / etc?

[coromonadalix]

wide input buck regulators / modules  etc ... you can find some going up very high, i saw some going up to 48 volts for a 5vdc output at 3 amps ??

it really depend on your design, and the prices nowadays are   ishhhh