Rethinking the Analog EFI

[Captain Obvious]

You could also have a look up table method that plots MAP against RPM

 

Look up tables aren't analog. 

[240260280z]

I could make one with a resistor matrix and op-amps used as a comparators along with free-running ramp generator to read each array.

 

Digital is really just quantized analogUE.

I see chokes, you see core.

 

Edited July 8, 2015 by Blue

[FastWoman]

Ah, Captain, ye of no faith.  Pay close attention to the noninverting input of IC5, a tricky little detail that I admit is easy to overlook.  This "driver" IC maintains an output voltage that is a fixed increment (Vr) above the timing cap's voltage (Vc).  That is, the driving voltage ramps too.  The charging current is therefore a very constant Vr/R.  It doesn't diminish as the voltage ramps upwards, so long as the driving voltage (Vc+Vr) doesn't clip. Here's how the circuit behaves:

 

 

Vc = timing cap's voltage

Vr = voltage driving the ramp.  I've used 1V in this example

IC8 = output of comparator (arbitrarily used TTL'ish voltages)

 

When the timing cap charge is dumped, voltage goes to zero.  The driver voltage then drops to 0 + Vr = 1V.  Charging current is Vr/R.  The driver voltage tracks Vc until it clips near the supply rail.  I'm illustrating the clipping voltage at 5V in this example.  I've arbitrarily set a threshold at 2V.  When Vc reaches 2V, the comparator output goes low.  It's admittedly a weird circuit, but it should yield a freakishly linear ramp, using much/most of the available range in supply voltage.  Moreover, large changes in charging rate can be made without chewing up too much available supply voltage "overhead."  Mind you, I haven't breadboarded the thing yet.  (See why having this thing dance around my head -- at least ICs 5-7 -- left me sleepless last night?)

 

EDIT:  I did seem to neglect drawing the ground on the noninverting input of IC7 in my schematic.  Sorry for that omission.  So ICs5, 6, and 7 are a difference amp, a follower, and an inverting amp, respectively.  Also it should go without saying that the conditioned correction signals circuits are just shorthand.  I might also have to invert a signal or two, prior to the ramp generator.

Edited July 9, 2015 by FastWoman

[FastWoman]

Surely you must admit that's a cool little ramp generator, even if I never build a retrofit analog ECU! 

[rossiz]

you people frighten me.

 

i'm just a simple carb-man... your ways are strange. 

 

why does this thread make me feel so completely dim and yet i'm so strangely attracted to it - like road kill: can't look away!! i have nothing of value to add here, other than wide-eyed awe at the thinking going on.

 

i guess it's just the crazy creativity driven by a passion to problem-solve. love it!

Edited July 9, 2015 by rossiz

[FastWoman]

Goeff, it's carbs that scare me -- don't understand them -- don't trust them!  They never seem to work right for me, no matter how carefully I clean out all their intricate little passages, and no matter what sorts of ritualistic kabuki dances I do.  (Yes, I know SUs are special and different.)  I've think I've finally gotten to the skill level that I can keep the mowers operational! 

Edited July 9, 2015 by FastWoman

[FastWoman]

Thinking a bit further, one way to do the open-loop thing immediately after starting, before the wideband O2 sensor is hot and operational, is to have a timed relay that feeds IC5 with a pre-set Vr.  After maybe 20 or 30 sec of operation (?), the relay would click off, and closed-loop operation would begin.  Maybe the open-loop Vr would be based on the conditioned CTS input.

Edited July 9, 2015 by FastWoman

[Captain Obvious]

Took me a page and a half of scribbled op-amp circuits to catch up to you (and I think there are some surmountable resistor value related errors in the design), but I got it. Neat use of active positive feedback. I won't guarantee that it would actually be as linear as expected without bread boarding it, but it certainly looks promising. Your kung-fu is strong.

 

I reiterate however, that I think you're loony for even thinking about such ventures...

 

The original Bosch design is a highly characterized black box of analog voodoo that cost probably hundreds of thousands of 1970's dollars to develop, took thousands of man hours, and several design iterations before they even had anything that would start and run under all input conditions.

 

That said, I'll do whatever I can to help. Haha!

[FastWoman]

Resistor values?  I dun' need no stinkin' resistor values! 

 

I fully agree with you about the mysterious analog voodoo circuits and the $$$,$$$ of research behind the original ECU.  The closed-loop operation relieves us of some of those difficulties.  We only need to get pretty close to the mark, so that the WB sensor can do the rest.

 

And my overall strategy is different from Bosch's:  They use fixed increment corrections, and I take advantage of the multiplicative properties of my ramp circuit to apply corrections proportionately to the air intake, as metered by the AFM.  Unfortunately I don't know how much voodoo is built into the AFM.  It's possible I need non-linear responses in the circuit design to complement the nonlinear AFM output.  I have no idea without some sort of flow bench testing.  (Of course we could use an MAF sensor instead.)  But I do know this:  The CTS modification on my car does get me running pretty much in the right mixture ballpark under low and moderate load, but my engine leans out and starts missing under high load.  That suggests to me that the incremental correction for coolant temp, applied additively, should instead be  applied multiplicatively.   I don't know whether there were any good multiplier circuit designs back in the mid 70's.  We did have some good multiplier chips from Analog Devices back in the early 80's.  Somewhat related, I still own a 70's vintage Heathkit signal generator that uses a hot filament in a light bulb as a key component in its AGC -- clever, but not too impressive technologically -- and way, way too fussy for an automobile.

 

I think a lot of the reverse engineering can be done on a test bench, to determine the magnitudes of the various corrections, relative to the base pulse -- or at least relative to each other.  That would be a starting point towards figuring out each of the conditioning circuits that I've short-handed at the beginning of the schematic.  And again, it's the wide-band sensor that will save the design and make everything good.  The design objective should be to take as much of the burden off of O2 sensor compensation as possible.

 

Anyway, this is not something I hope to do soon.  I'm simply throwing out some ideas as a starting point.  Maybe others will be inspired to tinker with the design.  Who knows... I might too.  I do look forward to breadboarding that ramp generator!  But for now, I coughed up an idea, and there it is!  And I have a place where I can come back and find it later.  Done for now.... 

Edited July 9, 2015 by FastWoman

[FastWoman]

Oh, and you want insanity?  If I had the time and the right automobile -- a 1950's something or another -- I'd enjoy working out some sort of EFI using vacuum tubes, with the ECU glowing in all its glory, displayed proudly in the cabin like a "flux capacitor."  But alas I have neither.

[FastWoman]

Light bulb moment:  You can't simply take the Vout from the ECU and translate that to a meaningful injector pulse.  You have to work RPM into the mix!  The longer the air flows through the AFM at a given rate, the more there is to burn, and the more fuel is needed.  So it's air flow times the rotational period of the engine that matters.

 

Complications, complications...

 

So AFM sets a ramp rate, with the ramp reset with each spark, resulting in a sawtooth waveform.  A peak detector (diode feeding a cap, with a resistor across the cap allowing the signal to decay) monitors the peak voltage of the sawtooth, which essentially integrates airflow over time.  THAT voltage then sets the threshold for the circuit I laid out.

[Captain Obvious]

You are clearly certifiable. 

 

BTW - Superlen had characterized a couple AFM's on his flow bench...