March 28, 2024, 03:20:36 AM

News:

For advertising inquiries or help with registration or other issues, you may contact us by email at help@harleytechtalk.com


M8 - Set Lambda during a CL to OL transient

Started by hdmanillac, January 02, 2022, 02:34:00 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

hdmanillac



Set Lambda (white) is the signal from the ECM
SL (red) is calculated Set Lambda based on AFR table
LF (green) is calculated Lambda for the front cylinder
Ping F (yellow) is calculated based on Spark Knock Front signal.

My question is the following: Why on M8 motors Set Lambda seems to be freezed during Closed Loop to Open Loop transcients ? It was not the case on TC motors.
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

What is the difference between set lambda (white) and calculated set lambda based of AFR table (red)? 
Semper Fi

hdmanillac

The white trace corresponds to the ECM signal and comes from the DPV log.

The red trace is a calculated field in MegaLogViewer that is supposed to follow the AFR table (SL=AFR(MAP;RPM)).

I don't understand why the ECM signal Set Lamda is not following the AFR table during transcients betwin CL and OL.
2017 FLHR + 2019 FXLR + 2007 XL1200R

rigidthumper

Correlated with the knock retard activity? The ECM has a subroutine to enrich the AFR outside of the map during knock events.
Ignorance is bliss, and accuracy expensive. How much of either can you afford?

hdmanillac

No, the knock retard is not my first concern. But maybe there is a link with this observed phenomena.

This phenomena exists even without knock retard.
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

In some cases that take a bike out of closed loop.  It isn't based on target anymore.  Acceleration enrichment table for example is a pulse width adder based on VE table.  PE mode over rides main AFR target if it is set richer.  During these times.  The ECM doesn't care about AFR main target table and since it isn't using it.  Might not show it accurate in the data.

Does the data allow you to see when acell is active or when PE mode is active?
Semper Fi

hdmanillac

The DPV no longer records AE DE signals in the logs for M8. And for PE the DPV never recorded it neither on M8 or TC.

Even if the ECM doesn't care about AFR table during some transcients, Set Lambda (the signal) is normally not affected by AE DE and PE. Set Lambda is normally taken directly from the AFR table.
2017 FLHR + 2019 FXLR + 2007 XL1200R

Jamie Long

January 05, 2022, 12:52:07 PM #7 Last Edit: January 05, 2022, 01:31:05 PM by Coyote

Quote from: hdmanillac on January 05, 2022, 09:23:03 AMThe DPV no longer records AE DE signals in the logs for M8. And for PE the DPV never recorded it neither on M8 or TC.

AE/DE are not native channels on the Harley M8 databus


Quote from: hdmanillac on January 05, 2022, 09:23:03 AMSet Lambda is normally taken directly from the AFR table.

Set Lambda is derived from the value in the AFR/Lambda table + other lookup tables & multipliers (PE, component protection, warmup enrichment, etc..)

hrdtail78

My bad.  I am use to TTS, and it shows accel and decel with the m8's.
Semper Fi

Jamie Long

January 05, 2022, 03:46:08 PM #9 Last Edit: January 05, 2022, 03:53:45 PM by Jamie Long
Quote from: hrdtail78 on January 05, 2022, 01:54:36 PMMy bad.  I am use to TTS, and it shows accel and decel with the m8's.

As far as I recall AE/DE are not active PID's in Centurion, DTT, ect.. on M8 models either (at least on 2017-2020 bikes).
Just curious how its expressed in the software you're using, are you able to view these as channels or on/off bit? 

hrdtail78

You cannot see attachments on this board.In histogram it has also shown as AC DE not being green when activated.  With TC's it was bit in the upper right corn of main screen of Data Master. Same as it showed kick stand placement.  In later m8's.  It will show as a trace at the bottom in bits, and in the list but it needs to be pulled up in one of the blank data channel spots.  Both will come under the same on both places.   
Semper Fi

Jamie Long

Quote from: hrdtail78 on January 05, 2022, 05:51:23 PMIn histogram it has also shown as AC DE not being green when activated.  With TC's it was bit in the upper right corn of main screen of Data Master. Same as it showed kick stand placement.  In later m8's.  It will show as a trace at the bottom in bits, and in the list but it needs to be pulled up in one of the blank data channel spots.  Both will come under the same on both places.   

Understood, thx

Jamie Long

Quote from: hdmanillac on January 04, 2022, 04:47:19 AMThe white trace corresponds to the ECM signal and comes from the DPV log.

The red trace is a calculated field in MegaLogViewer that is supposed to follow the AFR table (SL=AFR(MAP;RPM)).

I don't understand why the ECM signal Set Lamda is not following the AFR table during transcients betwin CL and OL.

If you created a calculated field in MLV software and it does not follow Set Lambda the ECM is commanding, your calculations are leaving out a piece of the equation and missing one or more factors mentioned in my earlier post

hdmanillac

Ok. I suspected there were some hidden parameters influencing Set Lambda.

But then how to explain that we did not observe that on a TC?
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Quote from: Jamie Long on January 05, 2022, 12:52:07 PM
Quote from: hdmanillac on January 05, 2022, 09:23:03 AMSet Lambda is derived from the value in the AFR/Lambda table + other lookup tables & multipliers (PE, component protection, warmup enrichment, etc..)


This is what I thought also.  The data coming out of the ECM is already calculated with the other tables. Set lambda, desired, targeted lambda.  How ever it is labeled. 

With this.  Sometimes programs will just repeat last known data if it hasn't gotten new data to refresh it.  This can clearly be seen by opening logs in basic excel file.  Not saying this is bad.  Just what it is.  Do we really need to update eng temp 25 times a second?  If you look at o2 readings.   With a controller that records at a quarter speed of the data recording device.  The options are to repeat data or show no data. 
Semper Fi

Jamie Long

Quote from: hdmanillac on January 06, 2022, 03:02:46 AMOk. I suspected there were some hidden parameters influencing Set Lambda.

But then how to explain that we did not observe that on a TC?

You answered your own question; if the calculated fields you created in MLV software matched the ECM's Set AFR/Lambda in one example but not another the factors were correct in one equation but not the other. Without having the data in front of us I don't know specifically what we are working with, however the lookup tables vary (many times greatly) by calibration level and vehicle model. For example an M8 bike with a 824/921 calibration has very different lookup tables than an early AFR based 176 calibration. 

The Set AFR/Lambda channel on the data bus is reporting the net result, post process; it goes like this:
What should the ECM command? = programming (calibration lookup tables)
What did the ECM command? = ECM data stream - Set AFR/Lambda
 

hdmanillac

An other input:

if the bike is equipped with a target tune kit, it works in a closed loop practically everywhere on the AFR table and we do not observe this "Set Lambda" gel phenome during the cruising / acceleration transients.
2017 FLHR + 2019 FXLR + 2007 XL1200R

Jamie Long

Quote from: hdmanillac on January 06, 2022, 08:42:51 AMAn other input:

if the bike is equipped with a target tune kit, it works in a closed loop practically everywhere on the AFR table and we do not observe this "Set Lambda" gel phenome during the cruising / acceleration transients.

Same rules as noted earlier apply with Target Tune

hrdtail78

Quote from: hdmanillac on January 06, 2022, 08:42:51 AMAn other input:

if the bike is equipped with a target tune kit, it works in a closed loop practically everywhere on the AFR table and we do not observe this "Set Lambda" gel phenome during the cruising / acceleration transients.



Something that has never been answered to my satisfaction.
Quote from: Jamie Long on January 06, 2022, 08:53:49 AM
Quote from: hdmanillac on January 06, 2022, 08:42:51 AMAn other input:

if the bike is equipped with a target tune kit, it works in a closed loop practically everywhere on the AFR table and we do not observe this "Set Lambda" gel phenome during the cruising / acceleration transients.

Same rules as noted earlier apply with Target Tune

This is something that has never been laid out and answered to my satisfaction.  It's advertised as 100% closed loop for 100% of the table.  Yet, same rules apply to TT as apply to limited narrow band closed loop areas.  There are adder tables that kick you out of closed loop.  PE mode being one of them.  With PE mode in later cals controlling more and more of the riding area.  How does it all play along.  My best understanding is TT has influence over the fuel trim tables.  These trim tables have influence over VE tables.  Having these trim tables active while adding the PE enrichment into the mix.  Would only skew the VE tables away from the target table.  So, how can the ECM stay in closed loop mode and activate PE mode at same time?

Seems it would be best to turn off the adder tables that we can, and shut off PE mode while using TT feedback and just worry about mapping VE's to airflow and then reactivate these tables once we are satisfied with VE mapping.  I mentioned this to DJ's tech line and was told it figures it out.  I asked how and was told it just does.

This industry is lucky in the fact that we have some good tools that already filter a lot of this data out for us.  This is how MyTunes, Vtune, Smart Tune, auto tune basic and pro all work.  I think this is what MLV and programs like it exist.  (not necessary for this industry) They take raw data and allow you to filter out any other tables and only worry about steady state while mapping VE's.  BUT since we do not have full access to all tables and all data fields with all the programs.  Sometimes it is easier to just use the programs created for what we have.

Semper Fi

Jamie Long

Target Tune calibration operate differently than standard tunes from several different standpoints; they operate 0-5v O2 inputs (rather than 0-1V), the O2's are interpreted differently related to O2 control such learn & freeze rate, and some definitions are used (or ignored altogether) differently between the different sets of O2's & hardware. The overall closed loop scheme with Target Tune is really basically the same as with the OEM narrowbands with some key differences; TT uses closed loop over the entire AF/Lambda table regardless of Lambda value; closed loop will follow any commanded AF value in the calibration as long as the command is within the range the sensor can measure (.70-1.15 Lambda/ 10.5-17.0 AFR). Closed loop is active during PE and even under DE, we filter decel out using min MAP threshold so it doesn't try to follow & learn what's happening during transients. You can also use CL during warmup enrichment once the sensors are heated (25 seconds or so). TT uses CLI (short term) and AFF (long term – stored) fuel trims, what the ECM learns in short term is eventually stored in eprom and will resume at key on, the calibration also controls the amount of rich/lean fuel correction that can be learned, as much a +/- 40 %.

I couldn't agree more on the helpfulness of the various tuning programs & utilites. Like others my craft has evolved greatly over the years as the bikes, hardware, and toolsets evolve.

hdmanillac

January 07, 2022, 04:36:37 AM #20 Last Edit: January 07, 2022, 04:43:10 AM by hdmanillac
I would like to understand why Set Lambda is stuck at 0.981 in an area of the AFR table where it should be lower.

Here is my AFR table


Here is Set Lambda from a log


At 2250 rpm, PE mode is off. And even if you disable AE DE in the cal, you still have this strange Set Lambda freeze.

This suggests that the AFR table is probably locked in the map, even if we can change the values.
2017 FLHR + 2019 FXLR + 2007 XL1200R

Jamie Long

Quote from: hdmanillac on January 07, 2022, 04:36:37 AMI would like to understand why Set Lambda is stuck at 0.981 in an area of the AFR table where it should be lower.

At 2250 rpm, PE mode is off. And even if you disable AE DE in the cal, you still have this strange Set Lambda freeze.

This suggests that the AFR table is probably locked in the map, even if we can change the values.

Its difficult to speculate without seeing all of the data and the calibration. Lambda is not stuck or locked, what you're likely seeing is the histogram you created in MLV is rounding the data in the columns, and remember the ECM interpolates the lambda command from a cluster of cells not a single value; for example look in the 20Kpa column in your histogram, it's .969-.980 when the lambda table is .960   

hrdtail78

Quote from: hdmanillac on January 07, 2022, 04:36:37 AMI would like to understand why Set Lambda is stuck at 0.981 in an area of the AFR table where it should be lower.

Here is my AFR table


Here is Set Lambda from a log


At 2250 rpm, PE mode is off. And even if you disable AE DE in the cal, you still have this strange Set Lambda freeze.

This suggests that the AFR table is probably locked in the map, even if we can change the values.

That is something I haven't seen.  Looking at the cell at 2250 and 90 KPA.  The target desire is way leaner than any cell around it.  I would suspect to see something from a value of .920 to .880.  Most tables that will activate in this area would be an adder table that would add fuel.  I can't think of any table that would influence the target to be leaner in this area.  The ECM is going to always go to the richer target, but PE mode is turned off anyway.

If you are emailing cals:  cnsspeedshop@hotmail.com  I'd like to take a look.
Semper Fi

hdmanillac

January 07, 2022, 09:44:38 AM #23 Last Edit: January 07, 2022, 09:56:59 AM by hdmanillac
hrdtail78, you have put the finger on ! I sent you an email.

 :up:
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Quote from: hdmanillac on January 07, 2022, 09:44:38 AMhrdtail78, you have put the finger on ! I sent you an email.

 :up:

My tuning laptop with MLV is at the house.  With the histogram feature I can get a better time stamp of when you were in that area and correlate that with the raw data in the excel file.

I'd make sure what type of filters you have going in MLV and I don't know if you are street or dyno tuning and weather is like.  BUT one thing I would try.  Set your entire target to 13.2.  Turn off PE, AE/DE and collect some data.  This will take care of tables we have control over and get the interpolations of cells out of the equation.
Semper Fi

hrdtail78

Quote from: Jamie Long on January 06, 2022, 04:18:22 PMClosed loop is active during PE and even under DE, we filter decel out using min MAP threshold so it doesn't try to follow & learn what's happening during transients.


How do the tables of Closed loop AE DE freeze Level and Closed Loop AE DE Freeze Time come into play with this?  The software has a recommendation of setting to use with TT.  It recommends that if transient conditions last for 2 milliseconds to freeze closed loop operation for 15 engine cycles.  But I don't see a distinction between AE or DE.

Power enrichment mode activation is dynamic.  Which also makes the target AFR dynamic.  Since the most basic calibrations are based on TPS and time.  The clock starts when tps hits the parameter.  Meaning that the clock can start at various rpm's.  If main target AFR table is set to .880.  If we go 95% tps at 2000.  Time base target AFR from PE mode table has a new target of .870.  As rpm goes up.  A richer mixture is called out.  My understanding from what your have explained.  Corrections will be made in the trim tables.  But now we hit the tps parameter at 4500 rpm's and the clock starts and the target is .870.  But wouldn't the trim tables be set for when we activated PE mode at 2000 and at 4500 PE mode was targeting .840 at 4500 rpm's?  This is what I mean by a dynamic moving target, and can skew fueling trims and the predictions that the ECM does.  Sure the TT will straighten it out next time it is in the area but first it needs to read and make those calcs to trim it back in.  Instead of just using a mapped VE table for the air flow and shooting for a richer target over time after PE mode is activated.

Trim tables are on the air calculations of the ECM. Fuel Mass= Air Mass  x (STFT x LTFT)  Air Mass is represented by VE.  The above describes the very basic PE modes.  Later TBW cals can have PE mode activated with TPS and KPA.  Which could cover even more of the VE tables. Seems to me for what it is worth.  Close loop in the area where PE mode is activated at the same time can cause the trims in that area to really chase it's tail, and I can't see with my limited understanding of how TT works.  How it can be a better strategy than just proper mapped VE tables.  Or turning PE mode off w/ TT and just asking for a richer target in the main AFR target table as RPM increases.

I turn PE mode off when mapping VE on a NA engine.  On a dyno WOT pulls don't take as long and PE mode is rarely activated for enough time to go to much richer mixtures. 5-9 seconds depending on power output of the engine.  Compared to a 2500 to 6200 6th gear pull on the street.

Semper Fi

hrdtail78

Quote from: hdmanillac on January 07, 2022, 09:44:38 AMhrdtail78, you have put the finger on ! I sent you an email.

 :up:

After going through the logs you have sent me.  It is targeting .981 in that area.  A couple of things with that.  There are multiple hits in that area but no two hits back to back.  I would suspect something to do with how Vision needs the base cal to build on, but I am not 100% on that.  I haven't seen this.  It looks like the changes made in the cal didn't take affect when loaded, or how the data is being read in the log.

If the vision uses the base cal and then has an adder table to change the Target AFR table.  The one hit might just be the base AFR table and might take more data hit in a row to see the adder table on top.  Not sure how it works but I think DJ would be interested in seeing this.
Semper Fi

hdmanillac

Quote from: hrdtail78 on January 10, 2022, 09:49:50 AM...I think DJ would be interested in seeing this.

Sure, but I don't know how to get them to deal with this problem...

I tried but the guy that answered me did not undertood the problem.

:scratch:
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Well, at first I wasn't following. :embarrassed: Reason for my answer about other tables.  Then you posted the histogram pic and I figured out what you were saying.  I did reach out to Dan H at DJ.  Maybe he will pop up or give me a call back. 
Semper Fi

hdmanillac

Well, it's very nice of you and if it leads to something concrete I would be very satisfied.

I thank you very much for your help.

 :wink:
2017 FLHR + 2019 FXLR + 2007 XL1200R

Jamie Long

Some quick clarification on how the PV sends the tune. It doesn't use the OEM calibration, rather it uses data within the calibration that's currently in the ECM versus sending all of the redundant files. When you initiate a flash it checks the existing file vs the file you are sending, any edits to the calibration that are different get sent, the rest of the files stay in place as these are values that are already there. This is why some tunes flash in as little as 15 seconds, however if you change tune strategies (i.e Level 723 to 921) it can take about a minute or so as its sending the full calibration in that case.   

hrdtail78

I changed the scaling up a bit to more pin point the area of concern.  Not many hits in this area at all and never two hits in a row. 
Semper Fi

hdmanillac

I've got other logs with more hits in this area of concern.

But when you log on the road, you can't always stay in this critical zone for a long time.
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Does anyone else see what hdmanillac and I are seeing?  I will send to anyone that wants the calibration and log file to open it up with whatever you are comfortable with.  Have a couple of cals and log files.
Semper Fi

Jamie Long

Quote from: hrdtail78 on January 13, 2022, 10:50:41 AMDoes anyone else see what hdmanillac and I are seeing?  I will send to anyone that wants the calibration and log file to open it up with whatever you are comfortable with.  Have a couple of cals and log files.

send them to jamie@fuelmotousa.com and I will take a look

hrdtail78

You cannot see attachments on this board.You cannot see attachments on this board.
Semper Fi

hdmanillac

2017 FLHR + 2019 FXLR + 2007 XL1200R

Jamie Long

I took a look at your log and I can see what you are referring to. I'm not quite able to make to a complete conclusion based on one log however I have a couple ideas what may be going on. The first thing I would have you try is to set the entire Lambda table as well as the PE table to .980 and take 10 minute log, then set both tables to .920 and take another 10 minute log, and post the logs so we can have a look. Being that you are overseas I'm assuming you have an HDI model and i'm also curious how your tune was derived; essentially what you used for the base map, the more info you can provide the better.

hdmanillac

Ok I'll do that asap. The temp ouside is very cold at the moment and I hope I don't turn into ice.

As I said before:
- whether I use my original 824 cal or this 823 cal from the previous model, the problem remains the same.
- I have also observed this phenomenon on all the M8 motorcycles that I have tuned.
2017 FLHR + 2019 FXLR + 2007 XL1200R

hdmanillac

From the DJ Technical Support:

"With the latest information received from you, I got our engineers involved again.  As it turns out, they were able to identify the specific issue and already have the solution.  This should be included in the next round of updates, which will be coming soon.  Thank you for pointing this out and helping us to more clearly understand the issue.

I suggest that you either connect that PV to your computer once a week to check for updates, or keep an eye on the following downloads page for updates.  Once updated, give it another try.

If you have any further questions or concerns, please feel free to contact us again."

This is very good news and I look forward to the next update of my PV.

 :smilep:
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Quote from: hrdtail78 on January 14, 2022, 11:57:39 AMYou cannot see attachments on this board.You cannot see attachments on this board.

Looking through the log.  At 243200 (2643) (First number is the time stamp.  Other is row in excel) You can see that the engine is 2100 rpm's, 64 kpa and set lambda is .981.  Throttle is opening slowly.  Rpm and kpa are climbing. At 245789 (2671) Engine is 2490 rpm's, 84 kpa and set lambda is still .981.

The calibration is set for .800 from 70 kpa to 100kpa for all rpm's.
Set lambda was .981 for 2.5 seconds.

You cannot see attachments on this board.

You cannot see attachments on this board.

When we look at other times the engine is in this area. 418519 (4512) to 419139 (4519)  Engine is 3011 rpm's at 71 kpa.  Set lambda .981. 

272567 (2958) to 279808 (3035) Engine is 1999 rpm's and 75 kpa.  Set lambda .981.  Both of these last two seem to shoot for .801 once hitting in the 80kpa's.

Some other interesting data as you are looking.  This bike is running on narrow band sensors.  Set lambda never did go to .800.  Maybe .801 is the limit.  Bad news for corn guys.  WBO2 front and rear are reporting and VE new is changing.  1223448 (13152) and 1223809 (13156) are just two example to view in excel.  This can easily be seen with MLV.

You cannot see attachments on this board.

I hope they got it fixed and am glad they recognized that there is a problem.  I went from "that is a lugging area" to "I had no idea this bike was not set up for target tune."

You cannot see attachments on this board.


Semper Fi

hdmanillac

January 18, 2022, 11:02:52 PM #41 Last Edit: January 18, 2022, 11:19:01 PM by hdmanillac
In conclusion, do you think that the problem comes from the fact that the ecm is partially configured for a TT cal while it is a cal for narrowband probes?

This reminds me of a cal reflash problem.

I noticed that WBO2 F and Front O2V are correlated in the log.
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

I really haven't came to a conclusion.  This might not be a calibration problem at all.  Might be how the PV requests and gets data from the ECM.  From what I have seen and discussed with other tuners.  This has seem to be a problem in the past.  Case in point might be AE and DE traces that PV is unaware of the m8 putting out.  It would be interesting to collect data with DTT ad look at it's log.  That is how I collect data when doing a PV tune, and then I use my tune.  This has always worked out for me better than using what is supplied in the PV software.

Usually when I am collecting narrow band and there is no wide band data or a wide band controller goes off of line.  The trace reads the leanest the controller puts out.  The leanest the frt is reporting is .815 and rear is .808, but the rich's they report is .685.  But as was explained earlier by Jamie.  WB's have a range and this is reporting out of that range.

Quote from: Jamie Long on January 06, 2022, 04:18:22 PMTT uses closed loop over the entire AF/Lambda table regardless of Lambda value; closed loop will follow any commanded AF value in the calibration as long as the command is within the range the sensor can measure (.70-1.15 Lambda/ 10.5-17.0 AFR).

Seems if this bike was running as rich as indicated by the log.  Narrow band sensors should be showing the same extremely rich condition, and the trim tables should be showing it in the closed loop area.  So, I really don't know and hope they figure something out. 
Semper Fi

hdmanillac

To be honest, I don't use autotune either because it produces VE tables that are too rich in the open loop area. I use another personal technique based on narrow band probes. And this technique uses, among other things, the "Set Lambda" signal. If the signal is false, the VEs will be false.

Dynojet said they found the problem and they have the solution. So we cross our fingers and wait for the next firmware of the PV...
2017 FLHR + 2019 FXLR + 2007 XL1200R

hrdtail78

Quote from: hdmanillac on January 19, 2022, 11:31:15 AMTo be honest, I don't use autotune either because it produces VE tables that are too rich in the open loop area. I use another personal technique based on narrow band probes. And this technique uses, among other things, the "Set Lambda" signal. If the signal is false, the VEs will be false.

Dynojet said they found the problem and they have the solution. So we cross our fingers and wait for the next firmware of the PV...

That is why it would be interesting to collect the data with another device that will show set lambda.  This could possibly tell us if it is just a data collection problem or a calibration problem.  If another device collected it and set lambda was displayed as we think it should, based on AFR lambda table.  Ding ding.  Yes, set lambda is very important to a speed density EFI system.  The whole system is based on having this correct.  VE's are for nothing without it.
Semper Fi

hdmanillac

Personally I don't have any other device to make recordings on this motorcycle.
But I will ask friends if they can do it.
2017 FLHR + 2019 FXLR + 2007 XL1200R