Electrolytic cap, automotive, vibration concerns 1

[CarbonWerkes]

Hello

I have a need to implement a pre-voltageregulator capacitor for an automotive computer. I need to address ESD not only on the sensor inputs (already implemented), but also on the main power input. Having done some modeling, and consulting with ESD specialists, it appears that 1ns-duration peaks will slip past MOVs/TVSs/MLVs. So, the advice I have received is to integrate the spike with a 50V 500-1000uf cap, which should keep the voltage level supplied to the regulator within its tolerance, and also have the head room to deal with load dumps. Problem is, very few if any vendors rate their 105C electrolytics for vibration- which is likely to be the killer in this application. Those devices which have a spec tend to be made to order, and not in a supply chain. Not a good option for me at present.

So, assuming that a standard 105C-rated electrolytic is the only real option here based on capacitance and voltage requirements, is there some experience out there with reducing strain/fatigue on the leads? Im wondering if it would be as simple as using a polyurethane-based adhesive to glue the body of an axial or radial can to the PCB, and then using some type of encapsulant on the leads to reduce their resonance? Seems like many ECUs out there just solder in 105C radials and leave it at that, but my application will see higher vibration loads- and if possible I would like to reduce the probability of a fatigue-related event.

Any thoughts are appreciated-
R

 

[Thecardoc]

Historically many PCM manufacturers potted the assemblies to attempt to cope with vibration and heat generation issues. That of course made for difficulty in servicing them as well. Is it possible you could split the capacitance into a small group of caps that would have the total rating you desire, and build some redundancy in with the same move? That way if one fails, the others still function and protect the circuit?

Beyond that using an adhesive, and some type of mechanical support would make sense.

This brings out another topic which from my POV is one that so much of the public just does not comprehend. As a technician I just might have to troubleshoot a failure in the system you arw working to build. To them it comes down to "let's try the computer" and see if that fixes it. For us it may come down to using a scope to monitor serial data lines, vref, and control circuits. When serial data on a vehicle randomly gets scrambled by a transient spike, it really makes our (my) job interesting.

 

[CarbonWerkes]

Hi doc-

The challenge with this initial filtering layer is that we have to deal with load dumps (which as you know can be in the 70V range for some milliseconds, plus 2000-7000v esd transients in the 1ns range, plus deal with the standard 14.4V stuff), while also handling severe vibration and some nasty heat loads. To do this with tantalum or niobium caps, or ceramics even, would require 50+ parts (and many types like niobium are not happy with higher voltages regardless of caacitance). So realistically, from a cost and layout perspective, 1 or a few larger electrolytics seems to to be the way to go given current technology (no pun).

One additional downside of some caps is that their failure mode can be closed- which is to say, a short circuit. If you have 50 small caps, your probability of a failure increases significantly, and so this may not be helpful either.

Yes, unfortunately with ECUs and various related systems, we now have to deal with multiple protocols for OBDII, plus CAN and SPI/I2C for external comms, plus sometimes even 232/485, multiplexed sensor inputs, etc. And all of those have the potential for software or hardware design problems, and can also bring in transients or overvoltages (or even reverse voltage from bad wiring or weird shorts)- so making sure that none of these modes can affect the core is difficult. Plus, with so much tasking, the MCUs are often running very dense code which is difficult to write and to debug. But your point is well taken- we should put some additional effort into debug modes, or even into supervisor processors that are looking for IO errors and logging that for you, etc. Would suck to replace an ECU only to find that the problem is a sensor throwing transients at it...

R

 

[WGJ]

Glue and sensible mechanical layour is what commercial engine module manufacturers have done in the past, where physically & electrically large caps were deemed nucessary.


Bill

 

[sreid]

I don't think that an electrolytic cap is going to stop 1 ns spikes. Small inductors are usually used for this.

 

[CarbonWerkes]

Sried-

I tend to agree. The spice modeling is likely not accurate, but I just assumed worst case per the MLV/TVS specs, and setup a waveform that the cap/vreg will likely see. Based on that, it appears the cap will integrate the spike and keep the voltage to the vreg < 40V, which is tolerable. But I do not have confidence in modeling at these time slices; there is a lot going on with ESR and inductance on the board etc that is not represented even in the best case (i.e. perfect component models). Would you then suggest an inductor pre vreg for the logic supply, and then allow the power devices (things like high side switches) to remain exposed (just paralleled with MLVs/TVSs)? Since some of these power components are 4-5Amp per device, I might have difficulty locating a high-value inductor with current capacity in the 10-12A range (presumably, single inductor for the whole subsection) which is not going to present additional problems (mass on the PCB, footprint).

Kind regards,
R

 

[Thecardoc]

How about a work around on this. Actively clip any voltage spikes by using a driver, behind the driver, where you apply system voltage to the just opened circuit. This would allow system capacitance, (AKA the vehicle battery) to do the job as the cap.

 

[Warpspeed]

One nanosecond spikes are pretty fast, all you need is a ferrite bead to create some series impedance, then a fairly small ceramic capacitor to ground after that. This will slow things down, and a TVS should then easily take care of whatever excess voltage remains.

The trick is to turn your high voltage spike into a lower voltage drawn out hump, or ringing. That should then be easier to deal with.

Think of it as a low pass filter followed by a voltage clamp.