Scatter Plots for Beginners

[FLTRI]

The proof is in the silky smooth power delivery and instant throttle response from any load.  That is all difficult to document but you know when you get it right.  I like to help people learn how to get it right.

Always remember, nobody wants your bike to run great more than you.  Nobody else cares to spend the effort.

Andy

Which is why we need to see what you see. Please show us. Just a before and after tune scatter plot will suffice to show us what we need. No more hype the feature, show us how to use it to become better tuners.
Or are you actually saying there is really no before and after picture to look at because it doesn't help tuning but makes brilliant plots?
Bob

[whittlebeast]

HEHE.  Do you want to watch screen shots as a tune dials in?  I am doing a race motor right now.  I normally am teaching people to tune. 

[FLTRI]

HEHE.  Do you want to watch screen shots as a tune dials in?  I am doing a race motor right now.  I normally am teaching people to tune.

Nope, not necessary...just the first screen shot (busted tune) then the last screen shot (after the tune is dialed in)
Is way it'll be much easier to show us the contrast. Are you actually listening to these requests?
If so you are evading the requests...hmmmm, why would you do that? Secret information only you are able to understand? HEHE
Bob

[98fxstc]

The only thing that concerns me is on the MAP X RPM vs Duty Cycle1 (really front DC) gets messy up and to the right of the circle.

Andy

[attach=0]

When you say gets messy
you mean a wider band (variation in values) rather than closer values along the line ?

And likely cause ?
Probably not the injector , but O2 sensor readings wandering a bit and ECM calling for different pulse widths ?

[FLTRI]

Are those two plots from a well tuned engine?
Or are they from an engine that needs tuning?
Bob

[98fxstc]

Bob   See post #37
I have done a lot of  DIY tuning for Start , Idle and Cruise and EGR
Think its "well tuned"
Want to get it onto a dyno to finish the job with > 80kpa

[98fxstc]

Here is the cal and data files for anyone who wants look at them in TTS
https://app.box.com/s/7ej3j7o7cttv88wzw612t4hcuxlhqs7o
https://app.box.com/s/boo0jpzmfstq6370nh3egbi56q4p8d9v
https://app.box.com/s/t35f38qtxo0vdkprf8na0qpn07tve60d
https://app.box.com/s/zwaq3ay5hbp6863w2fd1zx4qtqhvhn0f
https://app.box.com/s/bhvoj9lw2dv7thok0rwnl199vhsws0q3
https://app.box.com/s/9u05tkvo7d682rx4kqyghziysjmq97ul
https://app.box.com/s/1azkzzz22fywgqep9dpb9xhmut1urvoe
https://app.box.com/s/fu8yqfvpzz6khq8m52cetnh16dncmuaw
https://app.box.com/s/r2x1tzna56kyowc5gqb3yaqcfrv5jlht
https://app.box.com/s/s53p6316erg4mkpl2g1qsn9lpjrtsw5c

[whittlebeast]

Are those two plots from a well tuned engine?
Or are they from an engine that needs tuning?
Bob

Bob, I stated that I have concerns in the areas up and to the right of the circles.  Below and to the left it is very consistent.  It tuns up that the majority of that upper right area is under spirited acceleration.  It appears the autotunes were having issues as the throttle is it around 15%.  I personally would consider hand tuning above that point.

http://www.nbs-stl.com/HarleyTuning/Harley_MAPxRPM%20vs%20DC%20and%20TPS.png

Andy

[98fxstc]

Thanks Andy
Hadnt picked up on the 'spirited acceleration' 

[FLTRI]

Are those two plots from a well tuned engine?
Or are they from an engine that needs tuning?
Bob

Bob, I stated that I have concerns in the areas up and to the right of the circles.  Below and to the left it is very consistent.  It tuns up that the majority of that upper right area is under spirited acceleration.  It appears the autotunes were having issues as the throttle is it around 15%.  I personally would consider hand tuning above that point.

http://www.nbs-stl.com/HarleyTuning/Harley_MAPxRPM%20vs%20DC%20and%20TPS.png

Andy

Ok, if I understand you, the tune is busted above 15% and should be manually tuned above that!
So the tune should only be left in closed loop below 15% throttle due to unreliable Autotune data? How do you determine that from these plots?
Bob

[whittlebeast]

The tune is suspect above that point as there is no clear pattern in the plot.

If you are running on a dyno, at some point in the pull the motor stumbles, you get a puff of black smoke and the o2s indicate lean.  At that same point in the VE map, you have a huge VE number trying to cure the "Lean" spot.  What would you do?  This is simply one more tool to find this sort of thing.

I simply find/locate this sort of thing with math.

[FLTRI]

The tune is suspect above that point as there is no clear pattern in the plot.

If you are running on a dyno, at some point in the pull the motor stumbles, you get a puff of black smoke and the o2s indicate lean.  At that same point in the VE map, you have a huge VE number trying to cure the "Lean" spot.  What would you do?  This is simply one more tool to find this sort of thing.

I simply find/locate this sort of thing with math.

Ok, again the question, how do you see the bust is from 15% throttle up? Please point to where this is on the plot.
Also, what condition does it take for O2 sensors show lean as black smoke comes from the tailpipe? Unless, of course you are looking at no load/decel.
Maybe you can show the data to support that statement...right after showing the before and after plots that shows bust in a tune and the fixed calibration plot so we can see what you are referring to when you make these matter-of-fact statements.
Wonder if anyone else is noticing you have once again evaded numerous requests for before and after plots.
Bob

[Onthefence]

How do we take the tune bust at 15% throttle and up and apply to the map to unbust it?

[FLTRI]

How do we take the tune bust at 15% throttle and up and apply to the map to unbust it?

Excellent question.
Bob

[Tsani]

So I take it according to your idea, this would be an example of a "bust"? This a grocery run done this morning and its the first time I have seen it do this.

[whittlebeast]

It sure is suspect.  Plot Z=VE Front  to see if something pops out

try Z = o2 volts Front

Try Z = throttle position

Tag the lower dots and see on the trace what you were doing to get it to happen.

[FLTRI]

So I take it according to your idea, this would be an example of a "bust"? This a grocery run done this morning and its the first time I have seen it do this.

How about sending that log to Andy for his massage of the signals and see what he creates to analyze and plot?
Then, after you follow his instructions for what to do to correct it, you can post up the log to show the new, corrected tune side by side as I requested from him numerous times?
Thanks,
Bob

[Onthefence]

I still don't see a clear path to getting VE's corrected using this scatter plot.  It's pretty simple to hoover over an errant Dot and find out what MAP and RPM or  TP and RPM the Dot is riding in.  Now that I know that, what do I do mathematically to come up with a New VE setting in that exact spot with an expected outcome that I will herd that DOT toward the line.

Secondly, I not putting together how we know by the curve of the line that we are hitting targets to begin with.  If the whole tune is junk at 80 and up.  Won't the plot show a consistent line of junk?

[whittlebeast]

Onthefence

I generally do not tune based on what I see in the MAPxRPM vs DutyCycle.  I use that to only locate where I have concerns.  Often I find bad VEs that never got hit on previous rides.  What I noticed a few years ago is the closed the tune got, the better the MAPxRPM vs DutyCycle. got.  Now I always look there first when I get a log from anyone.  If that looks good, they very seldom have a tuning issue on the fuel side of things.  The most common problem I see is people run the closed loop way too much.  I fatten the motor up about .75 AFR and they are amazed how much better the motor feels.

When looking for knock issues I tend to use the RPM vs MAP and Knock Retard in the Z (Color)  I normally create a filter that filters out 0 Knock.  That way I can toggle on and off the normal running and find the problem cells quickly.  I like to use the Control P and Control S keys to see where I was right before I got in the knock situation.  Things like the knock only happens only after a long idle may show up.

Andy

[Tsani]

I loaded the native DM3 files up, but as far as I can tell, MLVHD will not load them unless converted to CVS files first. I have tried.
That graph is a combined trace of both files together. Don't know if that makes a difference or not. It has been my understanding that you can use multiple data files made with the same cal. But I don't think that holds true. Neither of the two files displayed separately show a clear picture of the split. But done together, they do.

http://harleytechtalk.com/htt/index.php?action=media;sa=album;in=221

[98fxstc]

I'm having trouble loading DM3 files as well 

[Tsani]

Files made with Datamaster, not Datamaster2, will open. So basically, the claim about opening DM3 files is not really correct. They haven't kept up with changes. Shame.

[whittlebeast]

If enough people ask the guys at MLV, they may elect to do it.  It is just typing as the programmers say.  All sorts of interesting things may show up in the logs.

[Tsani]

A DM3 file is a DM3 file. It says it can. They advertise it can. It can't. Shouldn't have to ask pretty please. They are making money with it, they should stay on top of it. It's the little things that tend to piss people off.

But you can export it to CVS format. Just don't forget to change the time column header. A PITA, but doable.

[98fxstc]

Have sent them an email