Rethinking the Analog EFI

[FastWoman]

I'm just thinking out loud:  I had a notion for how to build an interesting linear circuit, with a possible application towards replacing our crumbling ECUs.  We'd have off-the-shelf parts and open-source design!  I don't know whether I'd want to get into a project like this, but I'm thinking about it.  Sure, it wouldn't have the coolness and sophistication of a digital design, but then again, isn't there a certain coolness about maintaining analog control?

 

My thought is this:  It should be possible to take two voltages and generate an output pulse whose width is a scaled product of those two voltages.  This schema would use a typical voltage ramp (timing capacitor) with a threshold trigger.  One voltage, transformed/scaled from the AFM output, would create the threshold.  The other would essentially set the mixture, as a weighted average of some base amount, with compensations built in for other factors, such as CTS, ATS, TPS, etc. (now including an O2 sensor!  woohoo!).  This "mixture" voltage would set the rate of the ramp.  What would make this circuit unusual is that the mixture voltage would be summated with the ramp voltage to create a charging voltage (i.e. that feeds the timing capacitor via a fixed resistor), so that the current feeding the timing capacitor is scaled in proportion with the mixture voltage.  Designed this way, the ramp would be linear, rather than exponential (approaching an asymptote of the charging voltage).

 

This strategy would depart from Bosch's incremental compensations and would adjust mixture uniformly at all rates of airflow.  To me that makes more sense.

 

So I'm thinking this thing really shouldn't be all that hard to create.  The hard part would be reverse-engineering the weighting of the various compensations to the "base pulse."

 

Again, I'm just thinking out loud and putting my thoughts down here as a note to myself in the future.  Not enough time right now to sketch out a circuit.

 

Does anyone else have these sorts of troubled thoughts?

Edited July 7, 2015 by FastWoman

[Captain Obvious]

Does anyone else have these sorts of troubled thoughts? Well I'm sure you're not surprised to hear that I do, but my advice to you would be... Think about something else. Maybe boating.

 

What you described is reminiscent to an integrating A/D. You ever mess with one of them? It's similar except you're proposing a variable threshold instead of a fixed reference voltage.

 

Something else... Maybe sports? 

[Captain Obvious]

Not my fault....

 

 

 

[Zed Head]

I ponder how to make a 90's era system work on my car.  Waiting for my Pathfinder to break so I can cannibalize it.

[hr369]

The LE jetronic crowd thinks similar thoughts.

 

"I propose using a MAP sensor pinched from the Megasquirt apps and then feed this into the dev system board.  then drive a digital pot (via I2c or SPI) and simulate the AFM and feed back into the LE jetronic so it doesn't know any different. This way I stay with the analog beautiful FI heart"

 

 

 

"What I am proposing is an active circuit that can, not only, invert the relationship between air volume and sensor output voltage, but also potentially alter the response linearity (linear versus exponential). This could be accomplished with an Arduino as the base of the converter. The programability of the device allows it to adapt a variety of modern air volume determination methods to the analog expectations of the L-jet. It will also be able to program delays into the signal processing to simulate inertial sluggishness in response, and a variety of other "real world" anomalies. It is an interesting concept and possibly an exciting project to complete."

 

 

This electronics guru posted this way back in 2013.  He's yet to finish it however.  It's alot of work!

Here he describes the interface (language translation box) between a newer "hot wire afm"  <insert any ecu piggyback here > and the analog ecu.

 

 

 

What is the voltage change for each type :- Don't care it's just a reading at an air flow rate
What is the change in voltage with respect to air mass, linear, log or expo :- Again dont really care it's just a reading just like all other sensors I use at work. Is it the same for both types :- Again don't really care just need to know what it is.
Compensation for over swing :- Please expand but worse case it's just another input to measure, remember I am feeding my conditioner just into the Ljet AFM sensor input the Ljet still makes it's own decisions based on all it's sensor readings, it knows the throttle position already. I am just making a sensor black box conditioning circuit to trick the LJet to think it's still connected to the flap type AFM, not a EFI unit.
Works like this Plug little black box into the original AFM and the new hot wire type (both are in series) on the bike
Decode the revs and run the black box computer to give very fast commer delimited data to a files on the EXACT reading of each sensor at each rpm readings. Take data and import into windows excel, plot RPM vs Voltage reading for both AFM's compute the formula from the results and you have your algorithm for your lookup equilivelant for the black box computer Then plug the black box in between the hot wire MAF and the LJet FI unit and when I rev the bike, all I have to do is Supply the original AFM value to the Ljet input for the AFM I do this sort of stuff all the time at work with sensors we use in our monitor equipment designs.

 

Edited July 8, 2015 by hr369

[FastWoman]

I hate it when these things bug me and keep me awake!  The only solution is to jot down a circuit and get it out of my head.  So here's what I had in mind -- a partial analog ECU:

 

IC1 adds various conditioned sensor inputs, together with a target "stoichiometric" voltage for the wideband O2 sensor, to generate an output voltage corresponding to the desired air/fuel ratio.  The pots adjust relative influence of each conditioned signal on the corrected mixture.  This is compared against the output of a wideband O2 sensor (IC2), and the comparator output is smoothed to yield a correction voltage, which is followed by IC3.  IC4 then subtracts this correction voltage from a baseline voltage, to yield a corrected voltage, Vr, used to set the ramp rate of the timing capacitor.  The voltage ramp is driven by IC5, which outputs  the sum of the timing cap voltage, Vc, and the ramp voltage, Vr.  Thus the cap's charging rate is constant with Vr, and thus the ramp is perfectly linear.  IC6 generates the voltage ramp output of the circuit.  IC7 inverts this output to feed back into IC5.  Q1 is used to dump the cap and re-initialize the voltage ramp.  Finally, IC8 compares the ramp against a conditioned AFM signal, to create a square wave injector pulse.

 

I didn't worry too much about signal polarity.  Four quad op amps would do it.  It's a bit Rube-Goldberg'ish, but so is the original ECU.

 

BTW, this circuit would run closed-loop from the moment of startup and would use the broadband sensor to calibrate to the initial richer mixture.  Is this even possible with a stone-cold sensor?  I suppose the output of IC2 could be limited to some range that would "work."  Hopefully the O2 sensor registers a lean condition when it's stone-cold.  Any thoughts about that?  (I understand it's a bad idea to pre-heat them before cranking the engine.)

 

There!  Now I can go to bed. 

Edited July 8, 2015 by FastWoman

[FastWoman]

HR, my '78's beautiful FI heart is the organ most in need of a transplant.  The AFM is fine.  The '78 ECUs (and also the '77 ECUs -- Hitachi redesigns of the original Bosch) seem to lean out over time.

 

But still, clever thoughts about substituting an MAF sensor!

Edited July 8, 2015 by FastWoman

[FastWoman]

Zed, I've also pondered transplanting a more modern EFI.  I would think something using a MAF sensor would work, especially a system designed for a 6 cylinder engine.  To your knowledge, has anyone done such a thing?  I have to think correct mixtures are correct in any engine, but that may be a naiive notion.

[FastWoman]

Captain, I hadn't heard of an integrating ADC.  I'm not sure it helps, but it's interesting.  Anyway, now I can go to bed.  I hope I won't be thinking about boating now, as I have to restring a couple of rotten halyards, and that's making me nervous.  Maybe I can think about football.  That would definitely put me to sleep! 

[djwarner]

I can't add directly to this thread as I'm still deep into understanding the myth and methodology of my Webers.

However has anyone considered a path of adopting the older GM TBI systems like Patton did?

http://www.pattonmachine.com/TBI-Main.htm

[240260280z]

You could also have a look up table method that plots MAP against RPM and gives the injector pulse width at that point.

 

There would be a stack of these tables that match the engine temperature.

 

Tweaks could be added but it should get the system roughly in the correct working area .

 

The coil would provide the input signal for rpm counting and triggering. 

Edited July 8, 2015 by Blue

[Captain Obvious]

The reason I brought up the integrating ADC is because it uses the concept of a linear ramp to a threshold and also uses analog switching to change around the input circuitry. I wasn't saying it would be a drop in for anything here, but a lot of what you suggested is "reminiscent". I didn't mean it was a solution, I just meant that it has a lot of the same trappings inside.

 

As for your sketch... I don't see anything on there anywhere that would result in anything except steady state DC voltages. I think your little graph in the lower right that shows a linear ramping voltage is creative license that I don't think would occur. If you want a linear ramp generated from a DC voltage, you have to integrate the area under the DC line over time. You need an integrator and I don't see one anywhere. No bad vibes intended, but I think the output of every op-amp in that sketch would be flat-line.

 

 

Yes, I went there.

 

Boating, Miatas, football, Sandy Bottom Nature Park, Plum Tree Island Reserve.