Bosch LSU 4.9 sensors dropping like flies

dr_jones
Posts: 162
Joined: Wed Jun 08, 2016 3:24 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by dr_jones »

So - new LSU4.9s are here and I think I settled with a heater initial heating strategy using runTimeLR as I previously mentioned.

This strategy should meet the criteria detailed on the 4.9 data sheet which states 1.5v max during dewpoint phase (@ TimH - you mentioned 2v max here but UI've not been able to find any ref material suggesting the (slightly) higher number).

The data sheet also seem to indicate no ramp up is needed to the initial 1.5v so I've gone with that but then a jump to 8.5v and then a controlled (0.4v/sec) ramp up to operating voltage (13v) .

I've avoided the sudden jump to 8.5 volts and in stead ramp it up at 3.5V/sec (assuming 13.5v is 100% duty).

In fact the theoretical max in this cal is 80% duty so around a 10.8v.

Now as my output duty calcs are based on the logs of vbat (always aroung 13v and somespikes up to 13.5v) I may be a little out here - I'm not sure if the syvevs 12v outputs are capped at 12v or whether the follow vbat - I'd certainly be interested to know.

Below is the current table I'm using for this strategy - I guess time will tell if it's good or not!
Attachments
heater_control_runtimeLR.JPG
heater_control_runtimeLR.JPG (101.64 KiB) Viewed 32216 times
dr_jones
Posts: 162
Joined: Wed Jun 08, 2016 3:24 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by dr_jones »

No good! just killed another sensor after a 10 minute journey (all seemed well then a cool down for 45 mins or so - restart - sensor dead within a few mins.
Ryan.g
Syvecs Staff - Caretaker
Posts: 498
Joined: Fri May 30, 2008 4:05 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by Ryan.g »

Hi Ed

Sorry for not seeing this sooner on the Forum, its a area i rarely get time to check but just received your support email this morning which pointed me to this thread.

Have looked at the data logs this morning there is indeed looks to be either a faulty component installed on your S6Plus (CJ125 Chip) or a Issue in the wiring loom for lambda wires . While these boards get tested before shipping your issue seems to be intermittent as can see in the Logs you sent us. One log the LSUTemp is working correctly... The other is not and telling the Ecu that the Sensor is cold and hence the Heater drive is being driven fully. This will be killing the sensor for sure!

Please can you double check your wiring and crimps for the lambda wiring, Check resistance on each pins for lambda to S6Plus pin and if all good then i will ask our Sales team to send you an RMA and get the board changed.

Please update support via email on the wiring checks

Ryan
pat
Syvecs Staff - Cleaner
Posts: 356
Joined: Fri May 23, 2008 10:23 am
Location: Out there... somewhere
Contact:

Re: Bosch LSU 4.9 sensors dropping like flies

Post by pat »

Ed,

Apologies for the delay in looking at this - as previously explained I have a lot on my plate right now, and I am not part of the support group.

I have now looked through all the logs. The first lot aren't so useful since they don't show the LSU feedback, so we can't draw any concrete conclusions from those. Then there are some that show the LSU control working without a problem, which is good but doesn't identify the problem. Then the last few show a dead sensor. Unfortunately not a single log shows anything particularly helpful here, although there are two of the early logs that show an issue with the lambda reading, which Ryan has already pointed out. It seems to burst into oscillation, mostly when lean / overrun, but also sometimes when nearer to stoich.

Things to note are :

1) Internal resistance is measured by applying a bias current to the Nernst cell. There is a "chopper" that generates a 3kHz bias current waveform - this is sampled in both directions and compared, applying Ohm's law will give the internal resistance of the Nernst cell. Given that the internal resistance reading goes high it suggests that there is an issue with the Nernst cell. Given that two of the three sensors did not have heater failures, there is no reason to believe that the Nernst cell resistance was high due to cold temperature, ergo a damaged cell is most likely.

2) The lambda reading itself is a voltage that is proportional to the pumping current. Given that the reading goes lean when the sensor fails, it suggests that there is something wrong with the Nernst cell. If there was an issue with the Ion Pump then there would be no error at stoichiometric oxygen partial pressure - the Nernst cell would output the normal 450mV and there would be no need to pump ions since the Nernst cell is "on target" at 450mV. Only if the Nernst cell was not outputting 450mV would the ion pump amplifier start to drive current through the ion pump cell to try to get the Nernst voltage back to 450mV, and hence we would see a non-stoichiometric reading even though we were at stoichiometric.

3) The 18:32 log from 30/1/17 shows an opposite effect to above. Note that the lambda reading goes in the opposite direction. The sensor starts off cold, as can be seen by the fact it is reading lambda 1.0. There are external resistors that allow a tiny amount of current from the ion pump output back to the Nernst comparator - if the sensor is not present or has extremely high internal resistance then the ion pump can easily bias the non-existant Nernst output back to 450mV and hence we get a lambda reading of 1. As the Nernst cell heats up we can see that the lambda reading drops and it pretty much pegs at the rich end of the scale. The fact that the internal resistance measurement stays pegged at max, even though the sensor is warming up and doing something is interesting. It suggests that, even though the Nernst cell must be outputting some voltage, its internal resistance is still high (or that there is a fault with the internal resistance measurement system - but the fact the sensor does not work on a different controller suggests the resistance measurement is right).

4) The 18:33 log from 30/1/17 shows the lambda reading swing to lean - it is a continuation of the sensor warm-up from the 18:32 log. It should be noted that an NTK will go from lambda 1 to rich, before finally going lean (in fresh air) as it is warmed up so it is possible that the above effect is similar, but it does seem more extreme. The log on 12/1/18 at 09:57 would contradict that though - this shows a general lean-ward trend during a fresh air warmup (with a tiny rich-ward dip just after it starts to read). What is then interesting is that the sensor seems to be trying to settle to around 3.5 volts, but the log ends and a new log starts just 0.4 seconds later yet the voltage has jumped. It does this again at 2.3 seconds into the second log, and it jumps all the way to 5V, which is somewhat unusual (it even flags a sensor error as a result). I'm not sure I've ever seen a CJ125 output that much! Indeed none of the other logs show this extreme.

5) Even "dead" sensors still generate a reading, whereas an open circuit would cause a reading of lambda 1.0, so there must still be some Nernst voltage being generated - it is just wrong and this could correlate with the internal resistance being high (not open circuit). It is curious that even when the heater drive strategy has been modified to be a better match with the datasheet, sensors are still failing. I would ask the question about the orientation of the sensor - is it in line with the datasheet recommendation (section 4.4) ?

Given that there are some issues with the lambda reading in the earlier logs, and that there is this strange excursion to 5V on the lambda input, it seems extremely unlikely that the problem is being caused by some problem with the heater strategy but instead that the sensors are being damaged by some kind of wiring / interference issue. I would suggest that you try making a complete replacement for the lambda subloom (all the pins are in a row on the same plug) and route it into the engine bay, avoiding high voltage wires if at all possible, and if not then to cross them at right angles.

Best regards,

Pat.
dr_jones
Posts: 162
Joined: Wed Jun 08, 2016 3:24 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by dr_jones »

Thanks Pat - thanks for taking the time on that - very informative & helpful as always.

On 5) "orientation of the sensor - is it in line with the datasheet recommendation (section 4.4)" - around 11 oclock protruding very slightly into the gas flow - also it *is* further than might be optimal from the turbine housing (approx 90cm) - so in large it meets the recommendation (I can't see a section 4.4 - do you have a more detailed data-sheet - I'm using the recommendation under installation notes - http://www.bosch-motorsport.com/media/c ... 659pdf.pdf) - where its arguably not meeting the recommendation (Install at a point where the gas is as hot as possible.) is the distance from the turbine housing - it would be possible to mount directly after the turbine housing or in the headers themselves - but I don't *think* this has been the issue - its the same place the first sensor has been since last winter (when it previously being controlled by an LC-2 feeding an Autronic SM4).

On the loom/routing topic - current route itself avoids all high voltage / high load wires - in fact it drops out of the firewall (passenger side footwell) then runs directly back to the sensor along the side of the gearbox - so I think I'm good there. Could be worth while me double checking that loom though an potentially soldering as well as crimping the terminals - although it did check out fine this morning and the suspicion is an issue on the controlling CJ125 the problem does seem to be erratic / intermittent and to that end its always worth increasing the robustness of each potential failure point.

-Ed.
pat
Syvecs Staff - Cleaner
Posts: 356
Joined: Fri May 23, 2008 10:23 am
Location: Out there... somewhere
Contact:

Re: Bosch LSU 4.9 sensors dropping like flies

Post by pat »

Ed,

the more I look at the logs the more I am concerned about sensors being damaged during or just after start. As mentioned to Ryan on the phone, a Subaru manifold is just a few centimeters short of the distance between London and Glasgow - that is before the gas gets to the turbine, which will cool it down, and then it has to travel further to the sensor.

The datasheet you linked to is not the one I have here, it seems to be an abridged version thereof, but what I find interesting in it is that the highest output voltage listed for an LSU 4.9 is 3.868 volts at a lambda of 10, ie pretty much fresh air, yet you have in one of your logs an output of 5V. That would correspond to a pump current of 3.3mA, and whilst that would still be allowable (max 6mA) it does beg the question WHY ?

Unfortunately the log at 18:32:11, the one that goes the "wrong way", was taken with an engine being started, so it is not possible to tell what would have happened in a free air context - it *seems* like the fact the engine has started has somehow damaged the sensor. It does not make sense that the log leading up to this, ending 17:43:05 should end with a perfectly happy sensor, there be no additional logs in the meantime, and then the next log, using the modified heater strategy that does respect the datasheet and does not go full power on the sensor immediately (it takes 8 seconds to start getting a reading at full power and this log takes longer), pretty much kills the sensor as soon as the engine is started. Whilst "immediately" is "within 10 minutes", it did not take 10 minutes for the sensor to die - at least according to the datalogs I have here.

With regard to lack of sensor issues on a different controller - could you please confirm that you used that original sensor with the S6GPPlus and that it failed on the Syvecs ? Or was it a different sensor (physically a different unit, even if it is the same model) ?

Best regards,

Pat.
dr_jones
Posts: 162
Joined: Wed Jun 08, 2016 3:24 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by dr_jones »

Hi Pat - yes the distance from exhaust port to sensor is somewhat 'less than ideal' - a move to directly after the turbine (or within a few inches) may be a wise choice - however certainly worth noting the original sensor seemed happy enough in the current position for > 12 months would have taken in at least a few very cold morning starts in its pre-syvecs (Innovate LC2 controlled days). Unfortunately their sensor heater warmuup strategy is unknown (at least to me!). This sensor was the first sensor I saw this behavior on - but there have now been 3 since that have dies in a similar way.

Yes I see what you mean by the 3.868v it would suggest that a 'good' whatever would be very unlikely to exceed that voltage.

In respect of the logs - in fact all the runs with the new sensor to it failing are using the same modified heater strategy that respects the datasheet - I previously only included those I thought would be helpful - I've been back through these and included additional logs including the first time the sensor was powered up without the engine running) and the interim non running logs and added clearer commentary but more importantly spotted something very interesting indeed.

log lrs 2017-01-30 17-31-48 plus 0h01m08s S6#1628.SD - (and its non-streamed counterpart) was the last run where the sensor worked

I did not capture the logs immediately but returned to the car just under 10 minutes later and powered the ecu back up to pull off the logs.

log 2017-01-30 17-43-30 plus 0h00m00s S6#1628.SD is a log captured whilst I did this - it suggest something is now wrong - LSU Feedback_U03V is at 5v and LSU Feedback_U03 is at 100% - this was without startup - you can see in all subsequent logs these remain the pegged right up to where the sensor dies. I'm doubtful this could have been condensate related - the engine was up to temp when it was switched off and not started before the odd LSU Feedback_U03V behavior was observed.

Very odd indeed!

Note - I'm emailing the logs direct - they are too large to attach here.

-Ed.
pat
Syvecs Staff - Cleaner
Posts: 356
Joined: Fri May 23, 2008 10:23 am
Location: Out there... somewhere
Contact:

Re: Bosch LSU 4.9 sensors dropping like flies

Post by pat »

Ed,

The plot thickens.

Looking at the datalog from 16:46:39 we see that the LSU feedback is pegged at ZERO. That should mean there is NO drive to the heater since the cell temp is already way too high. That should be impossible since the sensor was clearly cold since it was not reading at the start of the file. The sensor DOES heat up though, you can see it starting the read. There is still ZERO reading from the LSU temp though. So the sensor is being heated up despite there not being any drive to it! You're not logging the heater drive so it is not possible to say whether it is actually being requested or not. The next log (which is streamed and also a native log) does NOT show the start, but it does show a working LSU, both in terms of the reading and now, magically, we have temp reading again! Something very wrong here! I think you need to log the LSU heater drive and also perhaps check with a true RMS multimeter that the heater drive really does match what is requested!

Cheers,

Pat.
dr_jones
Posts: 162
Joined: Wed Jun 08, 2016 3:24 pm

Re: Bosch LSU 4.9 sensors dropping like flies

Post by dr_jones »

Oh yes - so it does! - I'd not noticed that - that's 'bp01Duty_BoschLSU' is it not?

-Ed
pat
Syvecs Staff - Cleaner
Posts: 356
Joined: Fri May 23, 2008 10:23 am
Location: Out there... somewhere
Contact:

Re: Bosch LSU 4.9 sensors dropping like flies

Post by pat »

dr_jones wrote:that's 'bp01Duty_BoschLSU' is it not?
That may well be but I can't see it in the logs that are here...

Cheers,

Pat.
Post Reply