Motronic 3.2 debugging idle
 

Hi,

Question for the Motronic ECU crowd !

I have a 1985 Motronic on the test bench (std chip) that allegedly has a 'rough idle' & 'shakes to a stall' problem. I will only be able to test it in the car next week.
Apparently the car is driving around 'no problem' with another Motronic DME.

Doing the checks, the DME seems to respond to all inputs in a normal way (speed signal, reference signal, CHT, IAT, ...).
At 800 rpm, it holds a very stable pulse towards the coil at the required frequency and a nice injection pulse.

However, when I vary the rpms a bit (speed signal), there is almost no reaction to be seen on the ICV output. The duty cycles slowly changes a bit (max 10%).
I can get an 'immediate' response if I change the CHT input from cold to warm for instance so the driver circuit works.

So is the assumption correct that the ICV is not used to regulate idle ?
Or should it react promptly to changes is rpm (700 - 900 variations) ?

I would appreciate any pointers on how the idle regulation circuitry actually works on these Motronics.

Kind regards,

Frank

https://uploads.tapatalk-cdn.com/202...56e3c43894.jpg
Photo of my Motronic ECU tester for attention ;-)

Can anybody explain how the idle gets regulated by the motronic ? (1985 std chip)

Frank


Sent from my iPhone using Tapatalk

The ICV is used to regulate idle, that's its job. The ECU feeds a pulse-width-modulated signal to the ICV to control the idle speed. You might need to adjust the AFM gate spring position in the airbox thingy. By the way, it might be more informative to measure the Air Fuel ratio using an AFR meter.

I can see that the ICV is a block wave signal. I was expecting it to vary quickly when I deviate from the idle rpm (800) to compensate. However, I can only see very small/slow changes in the duty cycle (idle switch connected to GND).

I was wondering if I can check anything in the dme itself ?


Sent from my iPhone using Tapatalk

Make sure your closed throttle contact switch is making contact. It’s the two wire connector.

If the idle contact is open for any reason, the ECU will not control the ICV because it thinks the throttle is open.

The ecu is off the car and I simulate the switch by putting pin 2 to ground. It does react immediately to the idle switch on / off but only seems to regulate the ICV on very large devitions of the rpm.
I traced the signal back on the board to the processor and this is indeed how the processor outputs the signal (in the 5v domain).

It seems to act as a slow PID controller, I was expecting a fast-following controller ?
I can vary the output more by simulating a cold cyl head temp than it reacts to the rpm changes (600 -1100 rpm)

Frank


Sent from my iPhone using Tapatalk

When I do this on my DME simulator I look for are a couple of things. First, with the idle switch engaged and WOT not engaged verify that fuel pulses cut out above something like 3000 RPM. This verifies the idle switch is logically correctly recognized by the DME software. Idle and WOT switch inputs are internally wired into a resistor network to one of the 8 A/D converter inputs. Theoretically, if the A/D converter has an issue or the resistor network is wrong the DME might misinterpret your input on pin 2 as another condition.

Next, if you watch an ICV while setting the simulator to around 880 RPM and AFM input near lowest reading you will see that the ICV reacts fairly quickly to slight variations of the RPM thereafter. However, this only happens in a small band around the setpoint RPM. Once you increase RPM it closes and when you exceed a certain RPM the ICV snaps back to a different more open position. I assume that is so it can "catch" RPM dips when you close the throttle body by letting of the accelerator.

And as you correctly identified the output is a differential PWM signal with about 100 Hz. So in other words one output is the inverse of the other. That signal is also used as watchdog input. With this heartbeat absent the MCU resets.

Ingo

Ingo,

many thanks for the pointers !

I repeated some tests :

1. The IDLE switch is recognised and works as described by you. Under 'idle' the injectors cut out at about 1600-1700 rpm

2. I checked all the inputs to the A/D converter (Idle, WOT, IAT, CHT, AFM, Volts, ...) - pins 1-4 and 26 - 28 on the A/D and they all have correct and stable values at the 'input' and react to the simulated inputs.
Except the altitude box - not connected / not simulated for the moment.

3. When I run through the RPMs 550-1300, Idle switch on (GND) I get this response (ie nothing) : see short video : https://youtu.be/cNqamH6q1a0

However, I can vary the pulse width significantly by changing the CHT (?)
(Since I don't have a spare ICV, this is by using a LED on the output - pin 33 in this case)

Flip the idle swicth off and it goes to a 'standard' duty cycle immediatly (exact value depends on CHT)


I don't quite understand your final comment :

"That signal is also used as watchdog input. With this heartbeat absent the MCU resets."

The only input to the ICV circuitry that is rather isolated is the ADV11 output from the uP ?

Kind reagards and thanks for sharing these insights already !

Frank

UPDATE : I see what you mean, the Motronic checks the presence of the ICV output signal internally through IC S890 and resets the CPU if required.

Correct! The early Intel processors lacked the logic that could monitor whether the processor was reading its instructions and was operational after a power-on.
If not, it was continually reset until "up & running". Most all processors now days have a internal watch dog timer which does that function.

Thanks,

Learning so much here.
The icv signal is present, I monitored the reset circuitry and it is not resetting the cpu as far as I can tell (unless the reset are very short pulses).

I do see an immediate (moderate) reaction in the duty cycle and as Ingo said, once idle gets high it goes to a fixed position to anticipate throttle closure. So far so good.

What I do notice is that the pulses towards the injectors are ‘reset’ once every 2 seconds to an extremely long injection pulse (as if afm is full open/wot) and then they decrease step by step until it hits a certain length again and it ‘resets’.

Also clearly audible in the injectors.

So if it is providing a varying injection pulse with stable inputs, that would probably be the cause of the ‘unstable idle’ and it is partly masked at higher rpm?

I am going to see the car now and test all the sensors ‘live’.
Pretty sure it will be rough with this unstable behaviour, I will post a video of this behaviour afterwards.
Also noticed someone has been ‘in here’ before as one of the transistors in the injection circuit is non-std.

Frank


Sent from my iPhone using Tapatalk

Are you referring to T402? This Darlington transistor can have a hard time driving the six low-impedance injectors wired in parallel. It may overheat. I think the original RBDT65A is NLA.

Following this very interesting thread, trying to learn and understand how the IDE works. However I only have the schematics for the 944/951 version of the Motronic IDE.
Would anybody here have a decent readable jpg or pdf copy of the 911 version that they would be willing to share via MP?

Motronic 3.2 debugging idle
 

Send me a PM for the schematics

No problem finding most of the original transistors here.


Sent from my iPhone using Tapatalk

PM sent. Thanks.

Not really, unless the injectors are over-voltaged, e.g. bad alternator.

An actual IDE; https://www.silabs.com/developers/8-bit-8051-microcontroller-software-studio

Silicon Labs IDE

"The Silicon Labs Integrated Development Environment (IDE) is a complete, stand-alone software program that includes a project manager,
source editor, source-level debugger and other utilities. The IDE interfaces to third party development toolchains to provide system designers
with a complete embedded software development environment."

Very, very minor differences!

Actually, I do have a copy of the 911 Motronic schematic, but its low resolution does not allow to compare component values with that of the 951/944 version. So whatever minor differences there are, I'd like to identify them in case I need to stick my nose in my 911 88 Motronic.
So if you have a good copy of the 911 version and are willing to share it, please don't hesitate. :)

Think about what key function changes between a four and a six cylinder engine ECU.
You can always buy a 944 DME ECU on eBay versus relying on a marginally accurate schematic. They're cheap!

That would have to be the S400 (0127) circuitry, including the Darlington. I have already checked this part. As far as my eyes can see it (as some component values on the 911 version are unreadable), I can't find any difference between the two.

Or don't have any values indicated.

OK, back to the original problem ;-)

First we ran the car with another (known good) DME - no issues, clean idle, good injector waveforms, nice acceleration.

With the problem DME, idle was indeed rough but the ICV was regulating as Ingo said in a relatively small band around 800 rpm. However, the idle quickly got 'out of control' sweeping to 2000 rpm, dropping to 500 rpm and always eventually stalling. (Way too wide for the ICV to control anything)

Checked and known working : Idle switch, ICV, stable Vcc on the board (both 13.7 Vbat and 5.05V on the 5V level), O2 disconnected, Vref stable on the ADC, all inputs stable on the ADC, ref & speed pulses stable and correct on the uP.

When looking at the injection pulse (pin 1 uP) and the resulting 'peak & hold' output with the whole injector chain connected (I couldn't replicate that at home) the problem became clear : the injection pulse jumps up and then shortens, then jumps up again.

With 3 injectors connected now I can replicate the problem on the bench .
Video of injector pulse (pin 14) here : https://youtu.be/bQrVtaFdOKA
Video of output on pin 1 of the uP here : https://youtu.be/g_ePbtH7nv4

The uP pulse comes through correctly on the R421, although through ADV12 on this board (older board than the schematic). However T405 has been replaced by a 2N3019.

http://forums.pelicanparts.com/uploa...1641488531.JPG

Another thing I noticed, even with 3 injectors running, I do not see the positive peak on the scope at closing (first pic below) as I do on my good running car (last pic below).

http://forums.pelicanparts.com/uploa...1641488849.JPG
http://forums.pelicanparts.com/uploa...1641488849.JPG

Any help in this area is greatly appreciated !

Probably gonna check my stash of electronics and see if I can find a BC639C to make
it closer to the original again.


Kind regards and thanks for helping me along this quest...

Frank

Given the signal from the fuel output of the processor, your problem is not in the injectors' drive circuitry.

This is because you only have three injector connected, i.e. not enough output current.

Arbitrary parts replacement, not good troubleshooting!

Frank,

after watching the videos it appears you assume the variation in injector duration (pin 1 uP) is root cause of your issue. Have you tried the known good DME on the simulator? Also, do both DME have the same software (chip)? This behavior might actually just be due to how your simulator drives the DME. If not it would be the A/D converter producing erratic values for load or CHT or all inputs.

BTW, the spike at the end is when the injectors turn off and their magnetic field breaks down. The active snubber T403, T404, T405 absorbs most of that energy when T404 turns on.

Ingo and others,

Thx for the feedback.

I will try my good DME on the sim, but the behaviour on the car was the same, only worse since the rpm input was not stable.
So I doubt it is a sim issue.

And indeed, it could be that the ADC is producing unstable output. I didn’t think of that.
I can only ‘see’ its inputs and as said, they are clean and stable (also both VRef’s). I cannot observe or decode the output.

Given that someone has changed at least one transistor in the injector driver (I cannot see those under the cooling fin - is there a good way to get that off, seems rivetted). And some points in that circuit have been resoldered, I immediately suspected that subsystem.

Now, even with 3 injectors there should be some ‘recoil’ ? Now there is nothing, less than 1 V !
If not those transistors, is the large diode maybe damaged ? (OF622)

Frank


Sent from my iPhone using Tapatalk

I agree but this is puzzling me !


Sent from my iPhone using Tapatalk

I only say that because my simulator's RPM signal is not very frequency-stable and I sometimes see similar funny patterns where the injector duration follows that. I would check for that just to be sure.

I have a reference DME with all chips socketed and the quickest way to test is to de-solder the IC in question and test it in my reference DME. Doesn't take more than a couple of minutes per IC and gives definitive answers. Decoding the A/D converter isn't straight forward.

The BC639 is a pretty generic small-signal version and I highly doubt it's the reason you're having issues with the DME. And the riveted transistors are most likely stock because else you'd see someone messed with the rivets.[/QUOTE]

Your previous messages stated you see a large spike in the car but not very much on the bench. If that is the case I don't see the diode as the issue unless it is short. You can easily measure that.

The spike at the end of the injector duration depends on what injectors you're using on the bench. In the end if the load is different the generated PWM pattern from the 0127 will be different and the amount of stored energy in the injectors is different. So unless you're having the exact same conditions on the injector output I wouldn't worry about this. I just use a big inductive load similar to injectors on my simulator so I get PWM patterns and reactive end spikes.

Nice setup if you have a tester-dme like that. Indeed that helps.

To be clear : the large spike on my good car is 6 injectors AND the good dme.

The ‘no spike’ is observed with 3 identical injectors (I’ll try 6 tomorrow) to my good car and the bad DME. Same was observed (no photo) on the car with 6 injectors.

I put a copy of my 1985 original eprom in the bad dme to make sure it is the same as there was an unknow chip in there.

I’ll have to read up on how this ADC works / is controlled by the cpu (OE, ALE ?) and check all the paths to the cpu.
Same for the latch and eprom paths.


Sent from my iPhone using Tapatalk

If you have access to a 2-channel o-scope you can verify the output of 0127 on pin 5 to see how the snubber turns on at the end of the injector cycle. Even with a 1-channel you should be able to check that signal from pin 5 through T405 and R413 and verify that T404 turns on.

Keep changing those parts, surely you'll eventually find the bad one, right? Lucky for you that surface mount parts weren't used in the '80s when the Bosch DME ECU was designed.
Good luck, and have fun!

I am not changing any parts yet as you & Ingo pointed out there could be other reasons for the disturbance.

I will first compare again to my good dme on the sim (same conditions).
Then figure out a way to ‘see’ if the adc is generating a good output to the cpu. I assumed the digital board was indestructible ;-)

I will also hook up the full six injectors ( first check if my bench power supply can take that load).

When it eventually comes to it, no prob on redoing some components in the injector driver. I can order most of those for around 20 eur all together.
Even the adc is only a 6,5 eur part.

But as you stated, I’d better do some deeper analysis first.


Sent from my iPhone using Tapatalk

Frank, you seen my answer to your PM last night? (Thank you again for the schematics.)

As I told you, I think the missing 50V pulse is intentional in the know good DME. Check this page with typical waveforms. Though different technologies, they show that the closing "high voltage" pulse is there to accelerate the closing of injectors. In the schematics, it appears that this function is provided by the ZY47 (D401) zener diode. This is a 47V diode. May be this diode should be checked?

OK,

so with the hints from yesterday, work continued.

:o I should have checked everything again when attaching the chain of injectors yesterday ! I was having a bad case of noise in my AFM feedback from the simulator to the Motronic. That was probably causing the jumping effect.
I seem not to be able to get rid of the noise, so I wired up my spare AFM and it is much more stable, on the 'good' and the 'bad' Motronic box.

http://forums.pelicanparts.com/uploa...1641566150.JPG
http://forums.pelicanparts.com/uploa...1641566150.JPG

Only significant difference is (same conditions, same inputs, same injectors) :
- good box : at closing a large positive peak is seen (as can be expected)
- bad box : no positive peak

@Ingo : I got my old analog scope back from the attic, here is a trace of pin 1 (drive of the main darlinton) vs pin 5 ('snubber') output of the Bosch 0127 IC. It seems that the snubber circuit is turned on much at the same time than the main transistor or am I reading this wrong ?

https://youtu.be/uYn7Jre2OiI

More investigation to do,

Frank

Sorry, wrong pic posted for the 'bad' ECU, this is the one with the same conditions.
http://forums.pelicanparts.com/uploa...1641567003.JPG

Frank,

I noticed the funny pattern in the hold section of the injector duration in your “bad” DME. This doesn’t look right. It’s almost as if the current feedback is oscillating.

Look at your good DME where this pattern isn’t present. I have seen marginal 0127 before. They do work but have sustained some kind of degradation or damage. That’d be my guess.


https://uploads.tapatalk-cdn.com/202...980e7058cf.jpg

I wasn’t able to see much on your analog scope video but if I have time I’ll record and post some scope traces later today.

Sharp eye !

I have checked T406 : seems to be working, it opens during the whole cycle.
- bottom channel = T402 injection signal
- top = Collector - emitter voltage T406
http://forums.pelicanparts.com/uploa...1641571111.JPG

Then after R413 i get this and don't understand how this is supposed to work :
- bottom = same T402 main injection
- top = base T403 (PNP)
http://forums.pelicanparts.com/uploa...1641571323.JPG

More or less the same signal then at the T404
http://forums.pelicanparts.com/uploa...1641571414.JPG

Frank

@Ingo : the bottom irregularities are due to the timebase on the el-cheapo scope (35 EUR). On the analog scope they line up quite nicely at the bottom.

As I didn't understand the previous waveforms, there had to be something wrong in the snubber circuit and I believe I found it finally. The diode over the T404 transistor appeared not to be doing it's job. So I desoldered it and
(a) it was the wrong type in there (possibly due to my schematics being of a later version that this board) and
(b) passing current happily in both directions ?!

Replaced by one that I had (1N4007 instead of a 1N4002) and I can now see a correct waveform.

http://forums.pelicanparts.com/uploa...1641579482.JPG

The board is happily running now with 6 injectors. Heat after 30 min of running is 44°C or 113F on the heatsink, so well under control.

http://forums.pelicanparts.com/uploa...1641579565.JPG

Too late to put it into the car and test it, that will be for tomorrow.

That's a funny way of saying the diode was shorted and prevented the output to ever rise beyond the +12V level - just like your scope trace showed.

Ingo

Indeed !

Spcial thanks to Ingo for all the help and asking the right questions. I runs super smooth in the car now ! It seems even better/smoother than my 'good ECU'.

And incidentally, once this was fixed, the heavy noise in the power supply and ADC of the arduino in the simulator are gone. it did add some caps to better control it though. But 2 projects finished at once : the bad ECU repaired and a good simulator for the future.

And I learned a new area of my Porsche that I didn't really know before.


Thanks again,

Frank