9
\$\begingroup\$

Long story short, I'm designing a little device to monitor a sensor or two in my car. I'm not an EE, but I'm a stickler for knowing things will work right and that I won't have wasted my time and money only to have my device fried by less-than-ideal power sources. :)

I've managed to cobble together a power supply circuit that incorporates both a surge stopper (LT4356) and a linear regulator (LT1963AEQ), with a nice TVS (SMAJ40A) sitting in front, to give me regulated +5V and hopefully withstand the mythical load dumps referred to in SAE articles and TVS datasheets all over the internet. :)

What I'm stuck on now is... how do I verify what I have is good? What could I do better?

I've had a few ideas on testing it: really mess with the circuit and attach high voltage DC supplies to the inputs... touch wires briefly and what not. I'd really like to know if this thing can survive an honest load dump, though.

As far as the design... is my TVS diode stout enough? Is there another approach I should take like shoving in a big cap or two to slow down the spikes and give the TVS less work to do? I've asked in IRC channels and perused the web... and I've seen everything from DC line filters to inductors to huge caps to diodes to tranzorbs and a million other things. I'm just slightly confused on what is considered a good baseline approach to protecting a circuit from the rigors of an automobile's DC power system.

\$\endgroup\$
4
  • \$\begingroup\$ Before giving a shock to both you and the device: do your circuit and/or the components in it have some protection mechanism? Clamping diodes, smoothing capacitors/inductors, fuses... \$\endgroup\$
    – clabacchio
    Commented May 11, 2012 at 9:32
  • 1
    \$\begingroup\$ There is a TVS diode between the input and ground (SMAJ40CA, SMT, Littelfuse) and after that, per the datasheet, there's simply the current-sensing resistor and the MOSFET to switch the load. From the MOSFET, it goes right to the linear regulator which has a 0.1uF/1uF/10uF set of caps on the input and output. I still need to add a fuse to the input althought he LT4356 is set up to cut the input when it detects an overcurrent situation. \$\endgroup\$
    – Toby Lawrence
    Commented May 11, 2012 at 11:54
  • \$\begingroup\$ 400W TVS may not be enough for some load dumps electronics.stackexchange.com/a/17125/763 \$\endgroup\$
    – Martin Thompson
    Commented May 11, 2012 at 12:36
  • \$\begingroup\$ Readers may find this question of interest as well: electronics.stackexchange.com/questions/40419/… \$\endgroup\$
    – Tim Williams
    Commented Oct 28 at 4:42

1 Answer 1

Reset to default
10
\$\begingroup\$

The "correct" way to test your circuit is with a "load-dump generator" - for example, the LD200N load dump generator.

Load-dumps are actually pretty rare though - you need to worry much more about the other ISO pulses (eg the +-200V fast spikes which the TVS won't have chance to clip much) - don't know how fast the LT4356 will switch.

Those fast pulses tend to happen on many journeys as they can be caused by things like switching the wipers on and off, or the air conditioner compressor switching, for example.

EDIT: I've looked at the TVS datasheet:

Your TVS claims a "response time" of 1ps. I find that very hard to believe given that the package is 13ps long at the speed of light! I'd be interested to know how they measure it.

Anyway, the amount of series inductance there will be in circuit with it just from wiring and strays will slow that significantly.

Also, check out figure 1 - a load dump pulse can last a few hundred ms, two orders of magnitude to the right of that graph, which looks to halve the power for each 100x on the x-axis, so is more like a 200W device. It may well be a good device for the fast pulses though!

(This Passivated Anisotropic Rectifier (PAR) Vishay SM8A27, using a kind of Junction passivation is the sort of monster fitted to some automotive ECUs: )

Vishay have what looks like a reasonable read on the subject of automotive transients.

As a teenager I built a bar graph RPM counter that survived fine for many years without load-dump, or indeed much else, protection. Now I'm a proper engineer designing automotive products, I shudder when I think about it :)

\$\endgroup\$
8
  • \$\begingroup\$ Alright, so it seems like this is dipping back into the design aspect: how am I to protect against these +-200V spikes? The TVS datasheet says it has a response time of 1ps from 0V to Vbr and the LT4356 says it has a maximum turn off delay of 1us when it detects an overvoltage situation. I plan on putting in a decently sized electrolytic cap right at the input to absorb some spikes better.... but aside from that, is there anything else I need to consider? \$\endgroup\$
    – Toby Lawrence
    Commented May 11, 2012 at 11:58
  • \$\begingroup\$ We usually have some 100s pF of ceramic across the power terminals. Assuming the supply is fused at a reasonable current (I'm guessing you won't need more than 1A?) \$\endgroup\$
    – Martin Thompson
    Commented May 11, 2012 at 12:32
  • \$\begingroup\$ So, it sounds like maybe I need a stepped approach here. A big honker of a TVS at the front door, to limit the monster spikes, then a current-limiting resistor followed up by a smaller TVS to possibly clamp things down more. Is the response time of the TVS still a concern here? Would ceramics on the input really be able to dampen those spikes enough? \$\endgroup\$
    – Toby Lawrence
    Commented May 11, 2012 at 12:56
  • \$\begingroup\$ The ceramics also help keep some of the nasty RF out as well. Belt, braces, string, gaffer tape, etc. :) I should think that big TVS+small TVS+some caps would be OK \$\endgroup\$
    – Martin Thompson
    Commented May 11, 2012 at 14:20
  • \$\begingroup\$ So forego a current-limiting resistor between the TVS? Sorry if my questions sound like they're asking for a full answer flat out... but I just want to make sure I understand your answer properly. :) \$\endgroup\$
    – Toby Lawrence
    Commented May 11, 2012 at 14:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.