Hi Roman,

Thanks greatly for the comprehensive overview of protection
cktry you use on automotive devices. This is a great wealth of
practical information, useful for design, reliability, and
testing. It took me a while to digest it all. Anyone who has
seen the underbelly of a transistor knows electronics design
involves at least as much practical as theoretical.

Good rules of thumb - simplicity, overkill, bipolar devices
to buffer CMOS. I favor using the "smoking finger test" for
determining whether a device is running too hot. If I can keep
my little pinkie on it, then it is probably ok. Otherwise, I
re-design for cooler operation. [sounds like you would favor
using a baby's pinkie - figuratively speaking, of course].

I have one question. You mention using zeners in several places
for spike suppression. Have you any experience using transzorb
TVS devices [low-inductance, fast-breakdown diodes] to protect
I/O lines? Should be superior to std zeners there.

thanks greatly,
- Dan Michaels
Oricom Technologies
http://www.users.uswest.net/~oricom
===================================

ORIGINAL FOLLOWS (definitely worth re-reading):

Roman Black wrote:
>Dan Michaels wrote:
>>
>> Roman Black wrote:
>>
>> >Of course, if you wan't the *ultimate* in product reliability
>> >you might want to use the extra safety parts. I do this in our
>> >automotive applications.
>>
>> Hi Roman,
>>
>> Automotive environments are so incredibly noisy and spike prone.
>> Could you give a brief rundown on the basic protection schemes
>> used on the v.busses and I/O lines?
>>
>> best regards,
>> - dan michaels
>
>
>Hi Dan, now I'm not going to pretend to be some type of automotive
>electronics expert, as that is not my main field. But, I will say we
>have never had a product fail in service, and since many units we sell
>are used in racing motorcycles which often have no battery(!) and
>butchered charging systems I must be at least doing something right.
>
>I am a big believer in simplicity and overkill. Simplicity lets you
>use a big resistor as opposed to a smaller more expensive semiconductor
>or MOV spike suppressor etc. For the same parts cost you often get
>greater durability. Overkill to excess, I will use a 2w resistor in
>a circuit that dissipates 0.1w or even less. Resistors are cheap
>compared to the cost to you of a product that fails.
>
>My theory behind this comes from 20+ years of servicing TVs, etc,
>and it has been my experience that the most unreliable parts are
>any semiconductors, any varistors, any components full of corrosive
>(electro caps) and reliability is directly proportional to how cool
>any device runs. With the new tvs, I can touch the heatsinks for the
>the power semis and give you a pretty good estimate whether that semi
>will fail in 1 month, 1 year, 3 years, or never. Any service guy who
>fixes 50+ units a week will know what I am talking about. :o)
>
>For power filtering in auto applications I prefer simplicity of two
>large series resistors, with a cap to ground at the point between
>them. This is ideal for PIC apps which draw a small and regular
>current. I only use metal film resistors of course. Carbon resistors
>are a monstrosity! I use 630v polyester cap between the two series
>resistors (to gnd), these caps have better reliability than just about
>anything in that size and price range. With a 5v chip running from
>12v auto, I drop about 4v across the resistors at normal running
>current and the other couple of volts with a low voltage drop out
>3 pin regulator. I also use a 5w 24v zener to absorb transients
>that get through the two resistor ladder network. These big zeners
>last forever when spike buffered by a couple hundred ohms, mainly
>since it has zero temp as it doesn't conduct during normal
>operation. Yes I could use a 1w zener, but the 5w has a much larger
>silicon die and larger legs, and hence stronger mechanical coupling
>during heat expansion etc etc. Always us a BIG part if you want it
>to last forever, what the heck, costs 20c more per product!
>
>With a very low power app I would just use the two series resistors
>to a 5.1v zener, with a 630v poly cap and 24v zener at the spot
>between the two resistors, and obviously a large 22uf or more tantalum
>(did I mention I hate electros?) and this should suffice. When I was
>a kid I used to buy the ex-military and ex-commercial circuit boards
>from the local scrap dealer, and strip the parts for my hobby use.
>Surprise I found many of the electros were shot but NEVER found a
>dead tantalum. Also noticed the military stuff always used tantalums.
>How much price difference? How much do you want a little can of
>corrosive sandwiched between your heasinks? Yes I will sometimes
>use electros but only if no other choice, and I will use 105 degreeC
>types which have better rubber plugs and different corrosive type.
>Then I just prey that I mounted them far enough from any possible
>heat source. :o)
>
>With input/output buffering, again use the largest ohms value
>series resistor you can use, always use a buffer transistor, something
>like a little BC337 (800mA 50v) is tiny and cheap, but has a HUGE
>silicon die when compared to the in/out pins on a pic chip! If you
>buffer this with a decent series resistor and cap, it will be pretty
>bulletproof. If you need to run any decent output current from
>the PIC use a transistor with about 10x the current/pwr needed,
>again to use the maximum size silicon die and metal backing plate
>(never use a transistor you haven't pulled apart) etc. I like
>BD139/BD140, these have a huge backing plate and chunky die and
>are dirt cheap if you can fit the physical size. :o)
>
>Remember that MOV varistors and the like sound nice, but many types
>use an unstable metal "dust" compressed into a solid, and I have
>replaced so many of these things over the last few years I reject
>that as a solution. Nothing kills spikes as reliably as a RC network,
>or a RZ network and with good choice of R and C (or Z) it will do
>so forever. Many of the MOV type solutions are a trade off of "MOV
>kills spikes, which at the same time slowly kill MOV" that suits
>many TV/telecom  manufacturers, but not suitable for a product I'm
>putting my name on!
>
>I apologise to the many professional engineers here for my "bucket
>science" solutions! I am not an engineer, I'm maybe a "re-engineer"??
>;o)
>-Roman
>

--
http://www.piclist.com#nomail Going offline? Don't AutoReply us!
use mailto:listserv@mitvma.mit.edu?body=SET%20PICList%20DIGEST