Rethinking the Analog EFI

[Captain Obvious]

There are just so many places where things can shift over time. Combined with the open loop nature of the whole system? It's inevitable that it will go out of spec. I mean, c'mon!! It's forty years!! Springs to caps to silicon to solenoid valves to RTD's? Connectors all over the place? Who knows... Maybe the ECU is unchanged, but the dropping resistors changed over time? Inevitable I tell ya!

 

I recently tweaked my AFM one tooth leaner. I don't have a wide-band, but my nose and my eyes were clearly telling me that I was running rich. I moved my AFM one tooth and it got a lot better. I'm actually surprised how dramatic of an effect I got with just one tooth.

 

Fastwoman, Love the Magic Eye tubes. Haven't seen one of those in quite some time!

[hr369]

Personally I think it's failure of the semiconductors -- most likely the mysterious black-box linear chips in the little "spider" cans.  Given enough heat, usage, and time, semiconductors do go sour.  Here's a little article about it that's a bit over my pay grade:

 

http://spectrum.ieee.org/semiconductors/processors/transistor-aging/0

 

The most common failure mode for Hitachi ECUs ('77/'78) seems to be a shortening of the injector pulse, hence a leaning of the mixture.  It would be difficult to trace down how/why this happens, because we know nothing of the proprietary circuits inside the mysterious little black boxes.

 

Pirating ic's is a worldwide problem. What's so mysterious about ic architecture?  Can't the logic be deciphered by an off the shelf logic sniffer?

[Zed Head]

Gasoline quality has drifted also.  The gas sold today, even without ethanol, is probably different from what was available in the 70's.  As engine management advances to handle more fuel variation, which makes a more dependable car, the fuel suppliers can get looser with their specs., which saves them money.

[FastWoman]

HR, I don't know whether a logic sniffer would be much use on an analog ECU, but this, too, is above my pay grade.  Perhaps there are analog sniffers off the shelf?  A more likely approach would be to open the cans and cut/grind down the chips under a microscope to trace out all the transistors and resistors.  There are 20 transistors and 11 resistors in a 741, so it would probably be necessary to trace hundreds of components -- not fun, but it could be done.  Again, above my pay grade.

 

--------------------------------

 

Chickenman, that magic eye with the eyelid is perfect!  It does have an evil look to it, and the deco touch is nice too.  My first magic eye was on a reel-to-reel Webcor tape recorder I found at a garage sale when I was a kid.  Fond memories.

Edited July 14, 2015 by FastWoman

[Captain Obvious]

Reverse engineer what's inside the cans using some sort of sniffer? Not a chance.

 

In the past, I have on a few occasions, ground down through component encapsulation to expose the silicon inside and do a little reverse engineering and failure analysis. On something extremely simple (like a FET) you can make out what's what. But on something more complicated than that, you really need a roadmap. The reverse engineering I did on stuff more complicated than that included design documentation from an ASIC house that had designed a chip for the company I was working for. And even with that documentation in hand, I couldn't get too far from the pins before I got lost in the topography.

 

With 1970's technology, the chips should be less complex than today's higher density CMOS stuff, but I'm with Fastwoman... Beyond my pay grade. Maybe someone who designs silicon all day could reverse engineer under a microscope, but that's not me.

 

I believe that with scopes and meters in hand and enough reverse engineering at the board level, I think it might be conceivable to determine what the cans DO (black box) and infer what's inside. I'd be happy to take a shot at that if someone wants to send me the schematic for the ECU.