M119 O2 Sensor Behavior Changes w/Replacement Cat
 

I have been dealing with a DTC P0153 (Bank 2 O2 Sensor 1 response or ageing) on my recently bought 1996 SL500. The conventional wisdom is that the fix is to replace the sensor. So, since the car has 112K, and three of the sensors were original, it got four new sensors. The P0153 recurred about two weeks later. Some internet searching showed exhaust leaks around the cat to be a possible cause. Sure, enough, there was a small leak at the rear of the generic replacement cat where it had been rather poorly welded into place.

Local advice was that the generic cat should be replaced and that it would not work welding the hole. In fact, the local Meineke guy refused to touch it. So as sort of a science experiment, I used some stuff call ThermoSteel to seal the hole.

Then I hooked up my PC-based scanner, which allows you to get a real time display of O2 sensor activity. An image is attached. What it shows, is less "crisp" change in S1 (upstream) sensor voltage on B2 (replacement cat) compared to B1, and a much lower voltage from the S2 (downstream) sensor.

The B1S2 is running a constant .7 volts, B2S2 about .1, but B2S2 ran at about .5 volts until the car was fully warmed up.

So, is this why the replacement cats don't work, or do I have a different problem?

What should the voltage on the downstream S2 sensor be?

TIA

Chuck,

It would be real helpful to see the adaptation numbers for both sides both loaded and at idle.

You have something going on on bank 2, for sure. The interesting thing is that O2 sensor control is attacking a rich condition but your B2S2 is showing a lean condition overall. could be a lying B2S2 I suppose, you might try disconnecting it and seeing how those trim numbers change. Unfortunately fuel trim numbers give a very poor indication of MBs adaptation system as LTFT has only one value and MB uses loaded and Idle control. O2 sensor control or lambda is pretty much the same as STFT.

I took one STFT reading. B1=2.3, B2=1.6.

What is a good way to produce the load numbers? I can easily start the trace at stop, and then drive up a moderate hill that's about 3/4 of a mile long. I will kill the trace at the top of the hill. Then I can return to the garage and do a second session at idle.

Does that work?

For the loaded number drive at a constant speed of say 40-50mph. Unfortuantely LTFT or part load adaptation doesn't change quickly. It probably will take 3-5 miles and if one graphs it it should basically look like a straight line.

I'm presuming that your previous numbers were all short term at idle. Both banks showed fuel control (up and down on the graph) but B2 showed it all happening with negative numbers whereas the B1 numbers were back and forth around zero, which is the way it should be. After a while O2 sensor control (lambda) will borrow from long term adaptation and zero itself, but it can take a while. Your O2 sensor code will usually set when either control is pegged without help from adaptation or when the sensor is slow to act. This can be caused by a poor sensor heater circuit.

This may be too much information ...
 

OK, I will do the load test tomorrow. I need to be at a meeting about 15 miles away, so I will take a route that should provide at least 30 minutes of driving.

So you want to see a snapshot after driving, at least 40 minutes, along with the STFT and LTFT numbers?

I will also swap the S2 sensors and see if anything changes.

One other piece of information, re: your mention of the sensor heaters.

Soon after I bought the car, the CEL came on with a bizarre collection of codes. Both cam timing solenoids, air pump relay, all four sensor heaters, purge valve current and a couple of others. This, I believe, was caused by the 20 amp fuse in the base module being blown. I replaced the fuse, and after about 100 miles the first P0153 occurred. I replaced the sensors, and after another 100 miles, the second P0153.

Thanks for your insights on this.

Thanks for your insights on this.

PS, two log files from the short load and idle runs are attached. If there is an idle moment, one can go to http://www.obddiagnostics.com/index.html to download the free software, and use it to play back the logs. This is actually kind of cool.

I have never done any evaluation of MBs using OBDII LTFT (I always use the part load and CTP adaptations MB uses). I have done it somewhat with BMW which is basically the same or should be. I just don't exactly know how it applies as MB uses two seapate (or three on later cars) addaptive zones. The concept with addatation is that as STFT or O2 sensor control (as the Germans call it) approaches its limit of correction an adaptive correction is made percentage wise to the base fuel calculation. This sets the O2 sensor control back towards the center of range.

If you note your B1 fuel trim it all takes place about zero. B2 never gets above zero all negative. This means the base calculation is giving too much fuel and O2 sensor control is working to the negative subtracting fuel.

If you cancel LTFT and start driving within five miles you will see a stepped adjustment to a stable position. It could take a mile or two for each step so a graph will look like a straight line unless it has very long time frame capablilites.

Load. RPM or throttle position would be a good tracking point for comparison.


I would suggest only looking at just STFT and LTFT (and one of the above comparators) as O2 sensor voltage is redundant. STFT is the controllers interpretation of O2 sensor anyway. Plus it puts numbers to the relative correction taking place. In other words the amount of time the sensor voltage stays high or low is an expression, harder to see, of how much the system is pulling mixture rich or lean.

Results of freeway load test
 

So I took a trip on the Dulles Toll Road with my laptop hooked up. I started logging after I came up to speed leaving the starting toll plaza, and shut the logging down as I was exiting to the end toll plaza, somewhere around 11 miles. Then I took on additional reading when I reached the parking lot.

I also took some STFT readings while driving. They were mostly slightly negative in the -.8 to -1.6% range, with Bank 2 generally lower.

So here are the results from 3,617 data points:

Sensor /Average Voltage
B1S1 /0.4327
B1S2 /0.5637
B2S1 /0.5908
B2S2 /0.2299

Condition/ LTFT B1/ LTFT B2

Start of freeway drive/ 0%/ 1.6%
End of freeway drive/ 0%/ .8%
After 3 minutes stop/go/ 0%/ 1.6%

The spec for the S2's is >.485 volts.

Assuming I am reading this data right, I am puzzled why B2 would be running slightly rich according to the sensor, but have a positive LTFT ???

That is probably a good thing to try, just to rule out a bad sensor.

More data
 

When I fixed the exhaust leak, part of my science experiment was to clear the codes. Yesterday the CEL light came back on, same P0153.

The computer says this was what was going on when the code came up:

RPM 1550
Load 38.4%
Temp 185
Intake manifold pressure 16.5 in. Hg
B1 STFT -1.62
B2 STFT .0
B1 LTFT 3.1
B2 LTFT 1.6

The science experiment continues
 

So tonight I got under the car, swapped downstream O2 sensors from side to side, and put some more ThermoSteel on the welds including the front weld (for the first time). Hooked up the PC and took a test drive.

This time B2S2 behaved much differently. It looked a lot like B1S2, staying around .5V instead of dropping down to .1V like it did before. Back in the driveway, I let it idle for 10 minutes or so. As the engine temp came up, B2S2 started to fluctuate slowly around .5V, while B1S2 stayed at about .6V.

I think the car ran a bit more smoothly.

LTFT B1 was .8, and B2 was 0.

Given that both sensors are new OEM Bosch, I don't think the swap caused the change. I now believe that even the tiniest exhaust leak between the sensors can cause these problems. So I am leaning toward replacing the cat with the OEM version which comes as a single piece that bolts to the exhaust manifold and the muffler. I am working up my courage to ask the price of the Mercedes version.

Even more science experiment
 

Well, after getting no definitive advice from anyone except the Meineke guy, I decided to replace the left cat. Wanting the problem to go away forever, I popped for the genuine Mercedes article 129 490 68 19. Cost was $860 vs $710 for a 50-state "replacement."

Last night I dropped the old cat out and was able to get a better look at the problem.

- there is an unfilled hole on the top of the rear weld that I missed with the Thermo Steel (TS)

- the TS had come off in one place, probably causing a leak

- the exhaust blew several small holes in the TS in various places

See pictures for details. So I am clinging to my theory that even small air leaks or maybe the replacement cat caused the ECU to believe that the upstream sensor was not working.

And a definite lesson is that Thermo Steel is not all it's cracked up to be.

Hopefully, UPS shows up today and this issue gets closed out.

Any leak can skew the O2 readings.
It appears you can access the O2 sensor values in real time. Both sensors should dither, the front sensor between .1 and.9 volt, the rear sensor should be mostly between .1 and .45 volt. The ratio, front:rear, of crossings over the .45volt threshold should be greater than 2:1. If less than 2:1 the cat is bad.

The new left cat was waiting for me when I got home tonight and got installed without incident. I took the car for a 15 minute drive with the sensor plot on. This is attached, and can be compared to the plot in my initial post. The B2S1 "dithers" are now crisper, and B2S2 has stabilized at around .6 volts. Overall, B2 readings are almost identical to B1. So it looks like the problem is resolved. Time will tell.

Also attached, the new MB cat compared to the old "replacement" unit.

Great thread Chuck, I think you have it beat this time.