Idle hunt

[Bitsyncmaster]

I thought of a change that can speed up my rate of change but I'm still not sure why you would need faster rate of change to mixture control. What makes me think about it is why did the Bosch designers make their rate so fast. Of course I have more options since I can do these things in software and they had to do it in hardware. If you can think of a reason for faster rate of change, let me know what it is.

Anyway I tried incrementing and decrementing the PWM by two counts instead of one. That still lets me wait a half second for the O2 reading to respond but doubles my rate of change. I may do more testing increasing that count change. With a narrow band sensor you can't really read when mixture is further off which I could use to make the count variable. Wide band I could do that.

I also will play with my idle ECU after I finish the lambda ECU to see how rock steady I can make the idle RPM. Without fighting hunting, I may be able to tailor the idle ECU different.

[Bitsyncmaster]

Still waiting on my first boards to arrive. UPS says today but tracking has not shown it even arrived in Maryland. Anyway was trying to find a way to plot my RPM very accurately with fast updates. Can't do it via ignition pulses since at 800 RPM those just have a count of 40 each second. Was thinking about using the RPM pulses from the stock diagnostic plug but I've removed my sensor and plug. Not sure I want to put that back in but that would give me much more resolution. Measuring the time between each ignition pulse may work but would have to account for the uneven fire.

[Bitsyncmaster]

Finally got my first three boards today. Here is the first one which ended up fitting very well. I am going to modify the board so it can detect an open sensor better than what I was using.

[Bitsyncmaster]

Design testing:

This may just interest a few people that have knowledge of the narrow band sensor and the lambda system.

The narrow band O2 sensor will output 0.1 volts with a lean running engine and 0.9 volts with a rich running engine. The range of those values is very sharp centering around 14.7 AFR which is when no oxygen is in the exhaust.

Our cars recommend the dwell setting of 40 deg when the engine is warmed up and everything running as designed. When the O2 sensor is warming up, the lambda ECU drives a fixed 45 deg. dwell which would result in a little richer mixture than 14.7 AFR. So I set my ECU to hold 45 deg. and measured the O2 voltage as it warmed up. My ECU has a test point which is the O2 voltage buffered with an op-amp so your meter will not affect the sensor values.

This is the result on my cold engine start.

0:30 .0016 volts
1:00 .0047 volts
1:30 .039 volts
2:00 .17 volts
2:30 .38 volts
3:00 .55 volts
3:30 .68 volts
4:00 .74 volts
4:30 .79 volts
5:00 .81 volts
5:30 .83 volts
6:00 .84 volts

This result was interesting with how good things work with design. I have not adjusted my mixture so it looks like it is still set correctly at 40 deg. dwell.

Now I'm going to set my ECU to hold 35 deg. dwell and do the same test reading the O2 voltage.

[Bitsyncmaster]

Ran with 35 deg fixed dwell. O2 voltage still climbed to .53 volts at 2 minuets time and then started dropping to .31 volts at 5 minuets. So I'm impressed that the lambda systems works so good as designed (except for the idle hunt problem with stock lambda ECU). So the WUR is still keeping mixture rich till about two minuets of running.

Idle is notably not as good sounding as the normal 14.7 AFR or the richer mixture. So as expected a rich mixture gives a better idle.

I have more testing to do. I keep thinking of ways to perfect my ECU design. I found out the O2 sensor is not good at sinking voltage (pulling the signal up to 3.3 volts) but seems pretty good at sourcing voltage (driving a load resistor).

[Bitsyncmaster]

FYI: If you do a lot idling without revving the engine it will mess up the narrow band O2 sensor. Some high RPMs for a few minuets seems to clear it up.

I am doing a lot of idling testing my software changes and suddenly had unexpected results from my lambda ECU. This drove me nuts for a while till I measured the O2 voltage and it was not responding.

Turns out also the O2 sensor cools off fast even though the engine is warm. A quick punch of the throttle bring it back up quick.

Also the sensor voltage does jump around during normal idle. One sample every half second still works (what I started with) but I'm thinking of doing a bunch of samples and averaging to see if it helps working at a faster rate.

[cis6409]

Great work, Very interesting results, Thanks for posting them.
Very enlightening

Shane

[David T]

More modern cars have a heater in the O2 sensor to (a) warm it up faster to get into closed loop quicker, and (b) to avoid the sensor from cooling because of prolonged idle. Maybe changing over to a heated sensor is the way to go.

[Bitsyncmaster]

Well my O2 sensor died. Not sure if it was all the idling or when I was doing early testing with a pull resistor on the sensor. I will need to replace it to continue testing. Think I have an old one somewhere, if not will order one.

[Bitsyncmaster]

Got a new O2 sensor and installed it and still showed same problem. Turns out I was turning closed loop on to soon so swung the duty cycle to much and caused the sensor voltage to max.

Anyway the new sensor looks like it produces a cleaner voltage with less noise.