Hint for @Ash

You can use CMOS Schmitt triggers when you don't need precision thresholds. That saves on component count/complexity significantly.

Typical chips: 74HC14 (in 5V logic), 4093, 40106, 4584

CD4093 is 2-input NAND, so it can perform some useful logic functions at the same time.

Medved :

The PNP transistor is pretty much what is done in my 2nd circuit (P-MOSFET here, same idea)



Driving everything with the comparator outputs directly would mean connecting everything to the same point. With open collector comparators this is definitely a problem, as the intermediate voltage on the dividers etc will appear on the gates of the MOSFETs that shut down signals. To make this work at all, voltages of the dividers would have to be carefully considered vs. the MOSFETs Vgs th.... More design constraints and worse reliability

With push/pull comparators this could work, but i dont see a justification

Small signal MOSFETs are cheap, use them to break up everything to separate lines and invert as necessary, each with its own simple design requirement, resulting in fewer constraints and compromises in the design



The protection logic paths have to remain independent, because their outputs control different things in the circuit :

 - Undervoltage and overvoltage shut down all operation modes. They short the gate drive to the MOSFETs (or if a gate driver IC is used, its inputs or enable input)

 - EOL lamp only prevent the 50% opreation - Full power with the remaining lamp is still allowed. So it shut down the switching generator IC only



RRK :
Here i do want some moderate precision for the UV and OV protections. Since i already have Vref and comparators for that, i can as well use the additional comparators in the chip for the other functions



You all may be interested to see the following part of some other project i am currently working on

@Ash

I see. As a fresh idea for your design you may consider using modern capacitive isolators instead of optocouplers. Very convenient, good isolation ratings, logic I/O on both sides, saves a lot of components/board space. Original TI's ISO series used to be expensive, now every major Chinese manufacturer like Chipanalog or 2Pai seems to have an inexpensive clone With some versions, there is even a luxury of having a tiny isolated DCDC converter right on a chip, be careful at layout though, as they work at really high frequency, some 20-50MHz!

I dont use those in my designs. Not after i seen one of them in some other commercial product short between the 2 isolated sides after one side got burned out (from an external applied vdd reverse polarity)

It can be expected that the chip will burn out, but an optoisolator would never short through the isolation

Are you sure you are talking about capacitive insulators, and not some related technologies like Analog Device's magnetic couplers? Not sure about Chinese clones of course, but at least TI's parts consist of two chip dies, connected with some tiny wires inside the case. If you burn one die, the second will stay intact, and so will the second pair of insulating capacitors.

More, many times the insulation is not employed as a safety-related feature, but mere as a means to break ground loops or to protect serial ports like RS-485 or just to shift ground voltage levels. In that case, even if you burn the insulator there is no significant safety risks.

And you can of course employ a sacrificial 5V TVS on the power bus nearby to reduce the chance of this happening.

If an optocoupler is abused to the point that the case is charred, one can expect the insulation is compromised too. Though, TBH I have not seen this happned before in practice, even if semiconductors nearby naturally exploded ) Probably can happen with some intelligent high-speed parts having extra +3.3/+5V power to the receiver side like 6N137-139.

Back to topic, browsing through the parts for work I found out that thermal fuses for some high temperatures (say 150+ degrees C) and high current of 20+ Amps are easily available. So making a ventilated metal box with a couple of 100W resistors (some power margin) and strategically placed thermal fuses will solve the problem cheaply/easily/safely and with zero worries about the interference to car electronics. Added benefit is that you are not limited to 100/50% power ratios.

https://www.setfuse.com/Products/Over-Temperature-Protection/Thermal-Link-OTCO-Organic-Type/RP-series.html



 

That isolator was Si8641, rated as basic insulation (it was in a SSOP package. The same component in a SOIC-W would be rated as reinforced insulation). It was burned internally without visible external damage. Showed some medium resistance on a megger between the sides

(And dead short between VDD and GND on the side that was burned out, but this can be expected)



For your resistor suggestion :

Connect the resistor from the minus side, and route the lamp wires such that they cannot short to each other (so the wires come right up to the lamp holder from 2 sides, and not in one cable). This will reduce significantly the chance of full 12V getting to the resistor

Protect the circuit by a motor overload breaker set to the precise current - like PKZM0-25 or similar

For the resistor construction :

 - As the resistor element use a packaging steel strip. Use the thinnest possible so it have high resistance/unit length. The element will be of large length (about 4m if folded by two in a 2m long duct)

 - The resistor is installed on busbar insulators.

 - Cable duct made of metal with ventilation slots as the outer housing. Install a unit of ~2m length (as much as can fit) under the car

It may be possible to fit 2x or 4x the resistor element running in parallel along the duct. The 4x configuration is the best here, as it allows to fit 2x 4m long resistors (so the resistors for both lamps) in a single 2m long duct

The heat dissipation <100W in a 2m long metal duct is not much, it is not going to get to any high temperatures

Okay, the devil is in the details. Skyworks does not appear to to be overly honest on which parts are certified for which purpose in their datasheet. Well, even Chinese friends put it more clearly! For QSOP part the isolation rating is very minuscule (down to 1kV!) so it will not qualify and will be plain dangerous even for basic insulation, which requires a mandatory ground connection anyway.

We of course are using wide-SOIC reinforced 5kV class chips for anything between live and low-voltage circuits, SO8  for functional insulation only.

Still, you are right that digital insulators are at somewhat elevated risk of getting shot-through on a several abuse. TI even has a special application note on this.

Driving everything with the comparator outputs directly would mean connecting everything to the same point. With open collector comparators this is definitely a problem, as the intermediate voltage on the dividers etc will appear on the gates of the MOSFETs that shut down signals. To make this work at all, voltages of the dividers would have to be carefully considered vs. the MOSFETs Vgs th.... More design constraints and worse reliability

With push/pull comparators this could work, but i dont see a justification

Small signal MOSFETs are cheap, use them to break up everything to separate lines and invert as necessary, each with its own simple design requirement, resulting in fewer constraints and compromises in the design

Obviously I meant combine that way those that lead to the same response - like overvoltage, undervoltage and thermal shut down.
The extra transistors may be cheap, but it still means extra components to deal with. Plus discrete MOSFETs seems to me rather sensitive devices to handle, way moer than even the complex LSI CMOS IC's use to be. Apparently the desire to maintain the process simple and cheap does not allow any decent ESD protection on the gate. And compare to the power elements, they are lacking the high gate capacitance the power devices inherently have, which allow the large devices to swallow quite some charge without reaching any dangerously high voltages.

 - EOL lamp only prevent the 50% opreation - Full power with the remaining lamp is still allowed. So it shut down the switching generator IC only

Well, the other protections (OV, UV, TSD,...) still need to remain active and effective, even when in this "open lamp" mode.
For that I would make the open load detector to just stop the oscillator when the VDS at switched off transistor remains low. That way it keeps the other side full power, but still allow automatic recovery once the fault is fixed (= the lamp being replaced).
And because it goes via the chip in a normal way, other protections are working the normal way.

I did not expect this thread to blow up like it did. however I appreciate that it did. Lots of circuits and ideas to try. within the coming weeks I will likely have all the parts around to try at least one method. I would like to try them all tbh..

I am not super fond of the resistor idea anymore due to power waste/inefficiency

however I could build that too and see what happens I suppose. This is a separate project but eventually I want to also try and make a slow ramp up circuit for the halogens to prevent such a hard toll on the alternator when all that wattage comes on at once.

I am assuming I can make something utilizing a MOSFET, and a capacitor charging slowly via a resistor. I will have to mess around with that at some point. I don't want to mess with micro controllers if I don't have too mostly because coding is not my strong suite...I suppose better time than any to figure it out. But I do like the idea of simple "Dumb" circuits. It is what I grew up messing with and where my (limited) knowledge actually resides.

UPDATE:

The circuit for that specific function was actually pretty easy to find. Seems this one ramps up at 500 µs which seems quite fast for a halogen. So I probably will need to modify it/ look around more but the idea exists out there in the hobbyist space which is good.

https://www.instructables.com/Soft-Start-and-Soft-Finish-for-P-Channel-MOSFETs/

"The circuit for that specific function was actually pretty easy to find. Seems this one ramps up at 500 µs which seems quite fast for a halogen. So I probably will need to modify it/ look around more but the idea exists out there in the hobbyist space which is good."

What that is supposed to be about?

To turn ON a cold halogen, you need to turn ON the current fast, even when the resulting current spike is high. You need to warm up the filament as fast as possible, with minimum energy dissipated elsewhere.
Because the power dissipated in the lamp is proportional to the square of the current, but the total energy consumed from the battery just straight to the current, the higher current the higher percentage of the battery energy ends up in the filament. And because the energy you need to pump into the filament is always the same, the higher currents you use, the less total energy you need so even less energy gets dissipated in the wiring/switching elements.

So slowly turning ON the transistor is the worst you may do in driving a filament bulb.
If you need to handle the surge load of the transistor, way better method (and THE method normally used to control incandescents whose cold inrush current exceeds the drivers current capability) is to turn them fully ON at maximum current the transistor is capable of (here practically the short circuit currenmt dictated by the wiring resistance, Ron and filament cold resistance), then if the current is too high, switch it off (so it will be ON for the overcurrent detection time, which will be in the few 10's of us), let it relax to allow the heat to spread across the transistor (for some 100's us or so) and repeat until the lamp warms up so the detected current does not exceed the overcurrent threshold anymore.
The problem with this method is, it prevents the classic fuse to trigger in case there is a real short circuit, because by the pulsing even the rather high current spikes lead to moderate rms curent, but the transistor is getting stressed. So the system needs to implement some sort of time out, when the lamp should be already warmed up, so when the thing gives up the retries and enters "Overcurrent fault". All this becomes too complex for a "dumb" analog circuit, it is possible either with more complex digital control (all the timers, counters and the state machine), so usable only when it is integrated within some dedicated bulb driver "smart-FET" IC, or to some extend in a microcontroller.

I don't think it is realistic to have such complex functionality it in the "dumb" circuit you are aiming for here. Here I would just size the transistors so they can handle the short circuit (so turn ON fast enough) and rely on the classic fuse upstream to provide the short circuit protection. That is in fact the second reason (after the normal power dissipation) why to use such beefy FET switches - to ensure they can handle the repeated cold filament inrush (~150..300A per bulb for few 10's ms), as well a few short circuit current events (200A..1kA for some 100ms; what the expected 30A per lamp fuse allows). The proposed MOSFETs (1mOhm combined in 3xTO220 or so) should be able to handle that...

that's honestly interesting. I figured a soft ramp would be easier on electrical not harder. very well. hard start it is. less complicated anyway.

Absolutely no way of doing this linear way with the parts suggested. May be possible with a group of more robust MOSFETs or by allowing some PWM.

@Medved, please consult SOA graph from the datasheet. Although the part looks robust, drain current is limited to just ~8A @ 10V!

10ms curve won't fit for sure, we are talking a massive filament with a long time constant here.

Where do you came to the 10V?
Cold filament resistance of a 250W 12V halogen is about 40m Ohm, with 14V battery the current will be about 350A (neglecting wiring resistances). With a 1m Ohm MOSFET, so voltage drop on it will be about 0.35V.
With 6mm^2 5m long wires there will be another 14m Ohm, so we are at about 260A and 0.26V across the MOSFET.

Given the SOA of even a single FET, we are just fine - staying on the left "Ron" limit line (the voltage is dictated by the device Ron). And because I would expect at least 2 or 3 in parallel, we are talking about 80A or 130A for each MOSFET at the Ron line.

Yes, a short circuit may exceed the limit for just two in parallel for DC, but there the decisive factor is the fuse characteristic, where for a 30A fuse the break time is about 20ms for the slower fuse, while single transistor will withstand about 560A (calculated for the same i^2t when extrapolating from the "10ms" SOA). Using 3 transistors in parallel will bring the load way below the SOA...

If you are paralleling, put some 4 IXFN520N075T2 ($34 each) in parallel and you'll be able to soft start the lamp with the original linear circuit proposed.....



But i have another question about the "hiccup" start :

The whole question about soft starting was to reduce the effects of load transient on the alternator. Which i can understand as 2 effects :
 - The electrical brownout that will result from the essentially brief short circuit
 - The mechanical braking momentum, which could maybe lead to as much as a momentary belt slip ?

We are still talking about "hard" PWM on the car electrics here

Could the "hiccup" start (so pulsing the lamps on and off multiple times before they heat up) potentially be worse for the car electrics than a single event ?

 - The main filtering capacitors in the input of electronic devices will probably deal better with a few short brownouts (with recharging time between them) than 1 long brownout

 - But won't the repeated brownouts cause potentially more interference in other ways ?



Consider that with the addition of an inductor and diode, this circuit becomes a complete buck converter that can do all startup and dimming functions with no restrictions....

The magnetics will be a bit huge for low frequency, but probably a sensible frequency can be chosen to make it work

Or go back to the resistors, first connect through resistor limiting to 3x...5x In, then (after fixed time delay) to full 12V