dmel

I have taken certain readings directly from the sensors connected in the B134 connector of the ECU to have a look (MY 99).

The lamda sensor reading when idling varies wildly from near zero up to 800mV or so. The voltmeter displays a couple of readings per second. Is this normal?

And if yes how does the ECU control emmissions? Does it take an average over a greater length of time perhaps?

JIM THEO

I am not expert on this (although I am technician), check if your voltmeter has enough resistance. If no, this is the reason for your readings and you probably destroy your lamda sensor! IMHO use a reliable digital voltmeter.
JIM

Cosie Convert

"the sensor heats up too much with prolonged boost and reads over lean"

John, Do you have any idea what temperature the sensor starts to read lean ?

I was looking at one of Steven Done's data logs of a full power dyno run where he logged the Lamda output, it remained constant at .92 / .94v throughout the run right up to 7k rpm

cc

Cosie Convert

Thanks for that, having monitored my EGTs under various conditions, on a motorway run at 80 mph the EGT settles between 520 - 580 degC a 'good thrash' takes it to 650 - 700.

With all the talk about possible leaning out after extended high speed running and connecting that to No3 big end failures etc Seems that in it's current position it gives up just at the point that we need it most ?

I wonder if it would be worth fitting the lamda probe further down the system to keep it cooler ?

The downside may be longer warm up time ? not sure if that's an issue with the heated sensor ?

I guess you could always fit a secondary probe just for WOT monitoring ?

john banks

How about a proper wideband sensor for that purpose. In the link manual it also describes moving out the lambda sensor a bit to keep it cooler with a washer or something - should be easy to do, not done it on mine yet.

Cosie Convert

Wideband sensor bit out my pricerange I'm afraid !

john banks

There was a posting from someone in North America in my DIY AFR meter thread who pointed out that you could get them off Honda from scrapyards - at least in North America - check it out - it was not expensive - and I'm sure we could come up with a meter to read it if that is an issue?

Adam M

I was under the impression that the standard ecu always oscillates about stoichiometry as it systematically purges the catalyst.

If you run consistently at one af ratio obvious the cat and NOx trap become saturated and no longger filter. For this reason, in closed loop zones the ecu will always give an average reading of stoich (or whatever is desired) but will by no means be constant on an af meter.

In fact modern cars have two lambda sensors one downstream of the cat, which senses this variation in the af ratio in order to determine whether the cat is functioning properly or not. The net result being the output of the downstream lambda sensor mirrors the wave output of the upstream one but with a much lower frequency as the cat acts as a buffer, the frequency of the down stream lambda sensor being determined by the size of the cat.

mutant_matt

RE: the Wideband sensor, I have a chat with Mark (R19KET) about this and he said the trouble with doing that is (apart from the costly WB sensors) is that they usually require a temp compensation to provide a meaninful signal.......

That certainly sounds like a can of worms to me.......

Matt

john banks

Can of worms

Did you read the link?

http://www.techedge.com.au/vehicle/wbo2/default.htm

Hackish posted this and said the sensors are about £100. Might be worth a look?

They are also talking about assembled units rather than kits and show this response slope:

[img]ftp://ftp.diy-efi.org/incoming/DIY_WBafrVolts.gif[/img]

[Edited by john banks - 2/4/2002 1:22:18 PM]

[Edited by john banks - 2/4/2002 1:29:09 PM]

dmel

Thanks for the reply John. By the way, in all circuits in which you use engine sensors, do you use the sensor GROUND which is found (if I remember correctly) in the same connector (B134) as the sensors?

Jim, yes the digital voltmeter has a high enough input impedance, in the region 1.5 MOhm.

john banks

For the lambda meter I did, for the others no - I am only reading TPS and MAP and the exact values are not crucial. I am reading frequency of the RPM line.

At low current high impedance I cannot see different grounds being a real issue to be honest unless you are reading to 10mV like you do with a lambda sensor.

mutant_matt

John,

RE: the ground question, how important is it to take the Ground from the ECU for the AFR gauge then? (as I've used the Ground off the back of the stereo (as it's near the display)).

RE: the Wideband, whilst it's interesting, after you've got the PCB, the display and the sensor, you're looking at about £200. Is it really worth it? Perhaps....

Matt

john banks

Don't know. I heard you can get offsets. This way I know I don't. It is the only sensor I can think of which you need to read accurately where 20mV would make a difference.

No IMHO as I stated in the original thread.

john banks

The ECU is causing this behaviour through closed loop feedback. The Lambda sensor exhibits this behaviour when at stoichiometric mixture which is best for warm idle and gentle cruise conditions for emissions, economy etc.

If you are reading at 3Hz you will not see enough of the peaks to be able to interpret what lambda you have on high boost - you need a faster sample or an LED display for this - the signal is not rock steady.

Find a response curve for a typical narrow band lambda sensor and you will see almost a vertical line around AFR 14.7:1. You can follow out the tails on the graph and before the sensor heats up too much with prolonged boost and reads over lean you can estimate you AFR on boost, which is really the only time most of us are interested.

UK cars typically see between 7 and 11% CO on full boost.

See my DIY AFR meter thread in drivetrain - it was at the top only within 48 hours for much more disussion on all this and values and how to read etc etc.

Agree with Jim also - but most DVMs have a resistance of >1Mohm. If you are sampling at 3Hz I am sure it is one of these. If you use an analogue one as pointed out you can easily blow the sensor (overload).

[Edited by john banks - 2/3/2002 10:30:14 AM]

john banks

I get 0.89 high up the revs, 0.87 lower down. If you hold the boost on my car for more than about 20 seconds it starts to read leaner - just a function of an overheated sensor - the Link ECU manual on the MRT site has details of how to drive when monitoring lambda or using the Link's self tuning setup - although most agree it doesn't work very well.

Basically if you drive it on boost for say 10-20 seconds and leave it a minute or two cruising and then read again it is useful. If you are thrashing it it will read leaner and leaner (artificially).

Pat previously was discussing temperature compensation for lambda readings, but I never saw any detailed correction data.

Lambda readings are an inexact science but still very useful along with everything else.

It is OK to run it through the revs say in 3rd once at WOT, but then let it cool for a few minutes before reading again in my fiddling about with it.

[Edited by john banks - 2/3/2002 3:09:38 PM]