Thoughts on replacing 911 SC ECU with Raspberry solution

[AndrewCologne]

Originally Posted by AndrewCologne

The only thing is, sometimes after driving when the engine changes to idle, independent from the arduino, the LC-2 sensor controller outputs an error...

Fixed, ... as the control unit from innovate was connected to the relay switching circuit which provided a bit too low voltage/ampere.
Now its connected to where it should be, ... to the same circuit where the eco is connected (->Terminal 87).

Now fine-tuning of the duty cycle regulation/programming is on the to do list for this weekend.
The idle itself ist not oscillating but the sensor voltage reading of the arduino isn't that smooth like in case of a Digi MultiMeter which outputs an almost flat and smooth 2.4-2.5v signal (= 14,7:1 AFR) therefore the sensor voltage output is interpreted incorrectly by the arduino and results in a bit too lean mixture which drops the idle a bit compared to fixed 50% duty cycle resulting mixture/AFR. So the voltage reading paart in the code needs a smoothing/averaging routine ... which I should have taken into account before :-).

[AndrewCologne]

Idling fine without oszillating.



Perfectly at stoich around 2.45 volt +/- 0.1v of the sensor ( in case of the LC-2 at factory default 2.45v means 14.7:1)

Now ... when rapidly deacclerating from higher rpms (i.E highway) the idle drops below 700 rpm for about 1 Sec. This was also the case when using the original ECU.
As the Microswitch on the throttle housing tells the ECU when changing from idle to cruise or WOT and ...as well as the change back to idle, I could use this for a proportional enrichening by 1/2 sec when closing the throttle rapidly.

The 30% throttle switch works as expected. I recognised on a highway this 65% duty cycle state pushes the AFR up to between 12.8:1 and 13.4:1

[Coastr]

Interesting. Would it be possible to piggyback the arduino w/screen to check for effective ECU operation?

[AndrewCologne]

The whole board with the arduino electronic will be placed in an old original ECU case where the orig board will be removed so from the outer side all will be connected and looking pure OEM like.
The LCD-display could be integrated as panel in the center console, right below the A/C controls.

[Coastr]

Yeah I guess I was asking if I could replicate your setup condo that test for proper Ecu operation. The box is there, but is it working ? That's the question I wanted to figure out.

[AndrewCologne]

News on this ...

For a prototype with OEM look and OEM connections I took an old CIS ECU Box and removed the old circuit board. The new board came with the needed components and its working flawlessly, means no rewiring etc needed, just plug an play as its OEM like.







The DC regulator is still a stand alone device mounted on the board but will be changed to a solution which is integrated with single elements into the board layout.
The LCD display will be connected later via a disconnectable external plug solution

BTW: With this broadband lambda setup I recognized that when accelerating over 5 sec., the combusition tends to suffer from a too lean mixture!
A fixed factory default 65 % duty cycle should result in at least a 13.5:1 AFR but about 15.5:1 was the reading. So I guess independent from the ECU my CIS maybe suffers from a system or control pressure drop. I'll check that. Would be nice to know if this also happens to other lambda-CIS based engines build in '80-'83. At a first quick view the voltage at the fuel pump is 11v when circuit reaches the pump when key is "ignition on" and about 12v when engine is running, to me this looks ok.

[AndrewCologne]

Since 4 weeks testing, optimizing, testing and so on.

Much better and easy duty cycle calculation/behaviour than the original, results in much faster and more precise 14.7:1 AFR output at idle and cruising.
See the voltage reading on the upper left as its kept at stoich (around 2.45 volt).



Narrowband-Output of the LC-2 works perfect as well, means same easy duty cycle behavior and so on like at wideband output. This weekend Ill change the sensor with the real OEM narrowband O2 sensor. Target: Beside the option of working with a wideband sensor controller, also the original sensor setup must be supported by this CPU controlled ECU device.

A big plus of a wideband sensor/controller is also that the pre heating of the cold sensor makes the lambda system initialize/starting much more faster as with the OEM narrowband sensor. About 30 Seconds and the engine will be kept at stoich when engines temp is above 15C°.

I also added support of monitoring the engines temperature und the fuel control pressure on the right bottom. The display shows a value of 3.2 control pressure but its actually around 3.4 – will have to optimize the pressure reading in this part of the code.

This CPU controlled setup/version of the ECU shows that a much wieder range of duty cycle values is the result for keeping an optimal AFR during als revs when driving, accelerating.
Example: Engines pressures are ok (System and Control pressure), timing is just 5° advanced as of the premium/Super Plus 98 Octane fuel we have here in germany. SSIs installed.
Basic CO setting is exactly at 0.5%. I added the support of a simple switch where the system is forced to an open loop cycle of 50% and the lower right side of the LCD shows the CO being read from the wideband sensor. This makes it possible to adjust the basic CO much more precise as if using a more spartanic Gunson CO tester or similar.
With this basic setup at idle the calculated duty cycle is at about almost flat keeping 48%. When driving faster (still not accelerating with over 30% throttle ) I realize that the cycle will drop to almost 25%! means without this lambda control the CO would be very hight, even if not needed. When Accelerating the cycle rises immediately to static 65%. With the switch which makes the LCD showing the CO output at such an acceleration the AFR drops to almost 11.5:1 !
12.5 would be here the optimum. This shows me that 65% in some cases would be already too much.

So I did a test and even lowered the basic CO at idle to approx. 0.1%!
This (guess what) rised the duty cycle at idle to about 52% which is logical, resulting idle duty cycle also here still flat and smooth. Now, when accelerating on the highway and the ECU switches to 65% open Loop, the resulted CO is about 12.3 which is almost ideal. This week and with this setting I had the chance to test the cold start at approx 10 C° which made the ECU running at initial 65% till the temp rises and make the system go into closed loop. Here with basic 0.1% CO, the car also started up directly and the cold idle was calm and smooth. Till 35C° and with sensor still cold/initializing the microswitch makes the cycle rising to 75% for 2 Sec which makes changes from idle to cruising when driving very comfortable – like factory default.

[joostfra]

I read your story about the 930/16 engine on elfertreff.de and this was very interesting for me.
On elfertreff.de it was not possible for me to register.

Some time later I see you are posting on this place about the same project.
The arduino lambda regulator for k-jetronic.

This year my Porsche was at the stadium I needed (wanted) a regulator for the k-jetronic.
Of all the threads I read about the subject, yours on elfertreff en here were those which put me to action.
I had a arduino, so i build a first setup with a breadboard.



After the result was satisfying, soldered it all to a PCB.



The LCD I had for a Raspberry, worked perfect in this setup, with a 3meter UTP-cable.



I am using the original narrowband oxygen sensor.
For now my first discovery was, the mixture was much too lean above 3000 rpm.
Now with my simple regulator, acceleration is much faster.
And the oil temp stays lower.

[AndrewCologne]

Hi Joostrfra,

welcome in the crazy lambda world.

In my case and also in case of some other testers the ECU discovered that the 930.16 mixture enrichens significantly at mid revs. So for hitting Lambda 1 a Duty Cycle of 10%! is not unusual.

So, such a box beside providing a perfect combustion via lambda control, ist also shows directly if something in the actual K-jet setup is odd. In your case it can be that the sensor plat is out of adjustment. This also explains why in regular mode (without your connected DIY ECU) the accel. is worse.

[Bigtoe32067]

I’m amazed at all the knowledge on this board from all over the world. Keep up the good work and we’ll all follow along as you work through your issues.
Thanks for posting this here and please continue to keep us abreast of your progress.
Tony

[Walt Fricke]

I'm working on a somewhat simpler system: create a way to adjust the WOT (>30* throttle) duty cycle manually (set it, not have AFR values change it) on a system without an O2 sensor. The plan is to disconnect the WOT terminal of the throttle switch (pin3) from the ECU. This should not change idle or warmup values if I read the schematic correctly.

a) Is the 15*C switch closed (to ground) below 15*? Schematic shows that, but is that correct?
b) More as a curiosity, is the 35*C switch open below 35*, and closed above that?

At WOT throttle angles, a push button adjustable PWM board (grounded through pin 3 of the throttle switch) will control the FV ground, and thus the duty cycle. Below that (idle, cruise) the ECU will control the FV ground/duty cycle as usual. A simulated SPDP switch (also controlled by throttle switch pin 3) will switch the FV ground between the PWM system and the ECU system, to avoid interference. (The only power MOSFET SPDP switch I can find is only available from the Czech Republic - or at least not from a USA reseller, so I have to make my own).

The idea here is to optimize for power the WOT AFR values, without trying to adjust the WUR WCP physically for changes in atmospheric pressure. Elevation (of various race tracks) would be the principal barometric pressure difference. I suppose this system could also work in conjunction with the O2 sensor inputs, which don't affect WOT.

Microprocessors and accessories can do all this, and more, to be sure. I'm aiming at a single purpose using the more analog tools I am more comfortable with. Andrew's system would be much better for a street driven car.

[AndrewCologne]

Originally Posted by Walt

Andrew's system would be much better for a street driven car

The needed Lambda 0.85 is hit on WOT, I think thats what you want. As it also supprots wide band sensors, you can even set the AFR for normal driving lower if needed. It does exactly whats needed even on a track.

a) Is the 15*C switch closed (to ground) below 15*? Schematic shows that, but is that correct?
b) More as a curiosity, is the 35*C switch open below 35*, and closed above that?

The 15°C switch leads to ground if engine is cold (Bentley is wrong here). And in case of the 35°C Switch its the opposite – leads to ground when engine is warm.

You can simply disconnect the 35°C Switch and lead the connector of the 15c switch to ground and force a 65% flat duty cycle over all engine states. With that, just set the CO 1.5% and try what happens at WOT. In your case you should almost hit the Lambda 0.85 neeeded for an optimal combustion at WOT quite close. Just try.

At WOT throttle angles, a push button adjustable PWM board (grounded through pin 3 of the throttle switch) will control the FV ground, and thus the duty cycle.

I your case a changover relay would be driven by the throttle switch, where this relay switches between the PWM signal of the ECU and the signal of the external PWM signal generator. But be aware that here the PWM affects the ground signal of the FV. But honestly ... in your case of using the CIS based engine for race I would switch to a lambda controlled wideband sensor solution where you can set/adjust everything. On special request its even possible to support the reading of the revcount which would give you the option of a rpm based fueling. Even support of reading the manifold vaccum for checking the actual engines load is possible.

As you said you use your engine for race, ... what dizzy are you using and what type of vaccum unit? Both the original ones assembled with the engine when it left the factory?

[Walt Fricke]

Andrew
All relevant parts of this 1982 US SC engine are stock except the headers and exhausts, and deletion of the O2 sensor.

Interesting how much information one can get from the factory wiring diagrams (along with places like Pelican if you sift through them). Like the fact that it is the ECU which grounds the FV.

Because it is simple, I'll try your suggestion of using the two temp sensors to produce a constant 65% duty cycle. My current thinking is that the stock 65% at WOT is too rich, but I have the LM2 and this is easy enough to do. I can see the LM2 display on the track, and compare the traces recorded later.

One thing I don't think I have is a fine grained chart showing, in three columns or the like, lambda, AFR, and CO. Lambda 0.85 is what - 13.2 AFR? 12.5? CO 1.5 I can see is rich, but the same AFR?
No handy application out there like the metric to imperial units conversions, which are so helpful.

[Walt Fricke]

Never mind - Google is our friend - plenty of AFR/Lambda charts out there.

[AndrewCologne]

Andrew
All relevant parts of this 1982 US SC engine are stock except the headers and exhausts, and deletion of the O2 sensor.

You should think about purchasing a 123 ignition distributor (the one with 6pin CDI support) where you can set the advance curve and also the vacuum based advance to your needs as here regarding to performance the goal is significantly bigger than just lowering the AFR.

Interesting how much information one can get from the factory wiring diagrams (along with places like Pelican if you sift through them). Like the fact that it is the ECU which grounds the FV.

Yes, thats what I meant above.

Because it is simple, I'll try your suggestion of using the two temp sensors to produce a constant 65% duty cycle. My current thinking is that the stock 65% at WOT is too rich, but I have the LM2 and this is easy enough to do. I can see the LM2 display on the track, and compare the traces recorded later.

The trick here is that a 65% at idle when adjusting the idle CO makes the initial CO per screw lower. So this will affect the 65% at WOT directly. Just a thought, so just try and let me know.

One thing I don't think I have is a fine grained chart showing, in three columns or the like, lambda, AFR, and CO. Lambda 0.85 is what - 13.2 AFR? 12.5? CO 1.5 I can see is rich, but the same AFR?
No handy application out there like the metric to imperial units conversions, which are so helpful.

Lambda = given AFR / stoich AFR
So no chart here needed, its that simple calculation

So 12.5 / 14.7 = Lambda 0.85

But here it depends on the ethanol content of the fuel where the stoich AFR is. Here we use Super Plus 98 Octan which comes with 5% Ethanol. 102 Octan Premium also sold here comes with 0% Ethanol.

Stoich AFR with 0% Ethanol ist 14.7:1, but with 5% Ethanol its 14.4:1

So with 5% Ethanol gas you will need an AFR of 12.24:1 to match the optimal Lambda 0.85 at WOT.

The charts available in the internet showing the relation of AFR or lambda compared with CO are nonsense. As Lambda 1 or AFR 14.7:1 never equals CO 0.0! For shure not before the catalysator, means direct from the exhaust port ... and also not behind a catalysator after its conversion of the emissions.

Lambda 1 of the 930.16 and 930.10 engine with factory exhaust system equals más on menos 1.2 - 1.8 % Co vol. (before cat at the catalysator's CO test port or without cat installed then measured at the tail pipe), wich means also matching an AFR of 14.7:1.

[joostfra]

Reading this:

On special request its even possible to support the reading of the revcount which would give you the option of a rpm based fueling.

I am thinking how do you get the RPM-signal to the Arduino ?
Is is the black/purple line from the HKZ-unit to the RPM-meter in the dashboard ?

In my case when i am cruising 55mph and lift the pedal to idle, the lambda gives the signal "lean mixture".
So my regulator increases the duty cycle.
Now was my idea to check: RPM > 2000 && Idle == 1 set DutyCycle to 0%

[AndrewCologne]

In my case when i am cruising 55mph and lift the pedal to idle, the lambda gives the signal "lean mixture". So my regulator increases the duty cycle.

At deacceleration its normal that the sensor reports a lean mixture, ... then the ECU increases the Duty Cycle till it matches Lambda 1 again. No further influence by the ECU is needed here as in regular you just deaccel. for a few seconds. Then the correct mixture will be provided by the FV/ECU again directly.

[Walt Fricke]

That (black/purple) is the tach wire. Nice 12 or so volt square wave generated by the Capacitative Discharge box's circuitry in the Lambda SCs.




Arduino can handle it nicely. I use one as part of a system to check gear ratios - compare engine turns to wheel turns, with the engine signal coming from this tach wire.

[joostfra]

That looks nice.
I dont have the equipment to make electrical signals visible.

The speed signal you get from the sensor on the gearbox.
With 2 wires connected to the speedometer ? ( is there 1 ground and 1 signal ? )

For the engine temp, you use the green/black wire from the oil temp sensor to the gauge ?
This is a resistor with a tempterature dependency. Dont you mess up the original when you attach the arduino through a voltage divider (2 resistors).

[AndrewCologne]

For the engine temp, you use the green/black wire from the oil temp sensor to the gauge ?
This is a resistor with a tempterature dependency. Dont you mess up the original when you attach the arduino through a voltage divider (2 resistors).

You can use a temp "sensor" instead of the 15c temp "switch" or a simple pt100 temp sensor elsewhere.