News:

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

Main Menu

Scatter Plots for Beginners

Started by whittlebeast, August 05, 2015, 07:45:07 PM

Previous topic - Next topic

0 Members and 3 Guests are viewing this topic.

whittlebeast

Quote from: wolf_59 on August 05, 2015, 05:33:38 PM
Andy, it would really help us if you would start a thread "scatter plots for beginners" or something of the sort, I try to read all your post and the scatter plots are fine but I'm not seeing what your talking about, all I see is "OH Pretty Picture" but what does it mean? and what should it look like? How do we get it there?
Interested in what your saying just help us understand what your saying

I have had lots of questions regarding Scatter Plots and and how do they work.  Let me start with the most basic use of a scatter plot.  Let's call it "Input vs Response".  I like to put the thing the rider has a direct input on the X or horizontal axis. 

In my first example, let's put Throttle (TPS in my case) on the horizontal axis.  The response to this TPS input is the Position of the throttle blade (in the throttle body) or the resulting Response from the input.

http://www.nbs-stl.com/motec/Motec%20Throttle%20Mapping.png

In this example

0% TPS at the hand feed on the X (or horizontal) axis gives 0 Throttle Blade Pct on the Y or vertical axis.

100% TPS at the hand feed on the X (or horizontal) axis gives 100 Throttle Blade Pct on the Y or vertical axis.

Now look where the two white lines cross in the field.

27.77% TPS at the hand feed on the X (or horizontal) axis gives 15.60 Throttle Blade Pct on the Y or vertical axis.

The color or Z axis in this example, is the also Throttle Blade Pct just to keep things simple.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

98fxstc

#1
So , the color has no significance at all in this example but is there by default?
( z axis always utilizes the graded colors ? )

And this is TBW not a cable throttle
because a properly adjusted cable throttle would plot as a parabola and not a scatter graph ?

glens


whittlebeast

Yes this is a DBW.  There is a curve in the SW that I can set like 15 points in the finger position and what I want the desired resulting position of the throttle blades to be.

Actually, a properly adjusted cable throttle would most likely be a slightly curved line.

In this example, the color is not real helpful.  But keep in mind, the color can be the average of any other field, like Speed or average RPM.  Whatever you may want to look at.  Whenever you look at a scatter plot, you need to think thru what is the horizontal axis (the Input), what is the vertical axis (the Response) and what is the color representing.

Each dot represents one of thousands of samples.

This will get lots more interesting fast if you hang in there.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

#4
Here is one that is a little more interesting.

http://nbs-stl.com/motec/Motec%20Throttle%20Mapping%20to%20HP.png

The X (or the Input) is TPS or where my finger is in the throttle.

The Y (the Response) is the estimated HP based on fuel flow that I call HPEstMotec

The Z shown in color is the average RPM that the motor was at when the X and Y were at this point.

In this example,  50% throttle on the X results in about 140 HP and based on the color and looking at the right for the scale, about 6300 RPM.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

#5
Here is the same sort of plot but I am comparing a supercharged motor on the left and a turbocharged motor on the right.

http://www.nbs-stl.com/motec/Blower%20vs%20Turbo%20Power%20Delivery.png

The driver input on the bottom is throttle position.  The Response on the vertical axis is Inlet Mass Flow (essentially HP) and in color is the Vehicle Speed in MPH.

Look how on the graph on the left, Power is very much a near 1 to 1 response thing.  If your finger is here, the power is that.  The turbo motor on the right is very different.  45% throttle can get you almost anything for power.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

#6
The second type scatter plot that is very common in the EFI Tuning world has the bottom axis RPM and the vertical axis MAP or Manifold Absolute Pressure.  Lets consider both of these axises to be basic inputs,  The third axis out in the center is the Response.  In this case, these are the same axises that are used for a "lambda" calibration for the Volumetric Efficiency table.  In the Harley world, it is called the air flow table front and rear as I recall.

The more logical things to have in the field could be AFR, O2 Voltage, Knock Retard, VE, Spark Advance, even a special calculated field you created like Fuel Injector Duty Cycle... Anything you may want to look at.  Remember that each dot in the field represents the average of every time the motor got to that exact combination of the first two axises.

Here is an example of

X=RPM  (Engine Speed is what this SW calls it)
Y=MAP (Inlet Manifold Pressure is what this SW calls it)
z- AFR  (Exhaust Lambda External is what this SW calls it but I configured the SW to display data in AFR, that's just me...)

http://www.nbs-stl.com/motec/Motec%20Typical%20Ride%20100hz.png

Note that if you ran the crud out of a NA Harley, the data would look very similar but the top half of the data of this screen shot would not be there. Any data in this screen shot above 100 kpa is boost.  NA motors get little to no boost.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

Here is a Harley that was run fairly hard.  The RPM and MAP axises are flipped compared to the other plots but that is no big deal.  Max MAP was 104 as reported by the PV.

http://www.nbs-stl.com/PVTune/HitsTypicalRide.png

On the left is Speed Density (MAP vs RPM) and on the right is Alpha-N (TPS vs RPM)

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

Here is a little peak at where you can roll a scatter plot into a histogram.  The break points on the histogram are totally adjustable (count and value) and you can have any combination of X,Y,Z as required to create any table you want.  Showing you guys this is the fundamental foundation for creating your own autotune.

This feature also gets real handy for dealing with a motor that knocks.

http://www.nbs-stl.com/motec/Scatter%20Plot%20to%20Histogram%20View.png

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

98fxstc

Interesting stuff Andy
Am I jumping the gun in asking about filters and how they work ?
Looking at your scatter plots I'm wondering if it may be more usefull to look at data gathered by separating into say
Accel , Cruise , Decell ?
Thanks

whittlebeast

Here is the instructions for the math section of and list of math functions built into the software.

http://www.efianalytics.com/MegaLogViewerHD/formulas.html

You have the ability to do nested if statements to build any field or filter you want.  I can give some examples of my filters if it would help you get a handle on the possibilities.  Keep in mind that each filter can be applied separately.  Filters like TPS moving too quickly, Engine temp out of your acceptable range.... are all easy to do.

Hope this helps.

Andy

Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

Is this stuff starting to make sense?  You guys got quiet.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

98fxstc

Been having a bit more of a play with Megalogviewer
Can see how the filters work
I'd like to see some of those that you have found most usefull
I was looking at combining the info in DM3 files from a same day run
Can the files be combined in Megalogviewer for a single plot ?

Just Nick

I have been real busy but I am following along its all pretty simple so far
I'm never wrong , once I thought I was wrong , but I was wrong

wolf_59

Still here reading and looking, still haven't got the jest of the scatter plot and what I should be looking for.
Since this is a Harley site could you post up some more Harley data showing "Bust in the tune"and what we need to do to correct it.
I use TTS will I be able to use this to help finalize my tune?
May have to try the Megalogviewer and have a look see at some of the Vtune and Spark data files I have.

whittlebeast

Quote from: 98fxstc on August 07, 2015, 05:38:58 PM
Been having a bit more of a play with Megalogviewer
Can see how the filters work
I'd like to see some of those that you have found most usefull
I was looking at combining the info in DM3 files from a same day run
Can the files be combined in Megalogviewer for a single plot ?

Sure, simply open MLVHD, then <File> <Open> hold the control button and pick all the logs you want to open.  Hit <Enter> and they all load up.
Dynos are great for getting the motor close enough to get on the data loggers.

whittlebeast

#16
Quote from: wolf_59 on August 07, 2015, 05:59:55 PM
Still here reading and looking, still haven't got the jest of the scatter plot and what I should be looking for.
Since this is a Harley site could you post up some more Harley data showing "Bust in the tune"and what we need to do to correct it.
I use TTS will I be able to use this to help finalize my tune?
May have to try the Megalogviewer and have a look see at some of the Vtune and Spark data files I have.

What you are looking for is data that has a clear patterns but with some conditions that do not fit the pattern.

I realize but this is the sort of thing that is the holy grail that you find looking at lots of data.

http://www.nbs-stl.com/motec/Motec%20Prop%20Test.png

I realize that this is not Harley related but it is really cool when you can find clear patterns in a 1/2 million records.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

FLTRI

Ok Andy,
We're all like little kids listening to you go on about how cool these scatter graphs are and how it's so easy to make them out of just about anything you can imagine.
NOW! PLEASE! Take ONE friggin scatter graph of your choise and circle or point to where the bust you identify is in the graph and tell us what the bust is and how you see that. SIMPLE!
THEN! PLEASE! Show us the scatter graph with the fix to the bust and circle or point to where the picture changed after the bust was fixed. SIMPLE!
So please Andy just show us the value now that we know the utility.
Bells and whistles are great but we need to know why we need scatter graphs BEYOND THE PRETTY PICTURE.
THANK YOU,
Bob
The best we've experienced is the best we know
Always keep eyes and mind open

Sunny Jim


whittlebeast

Bob, I simply need a big data log off a really well tuned Harley being run hard on the back roads so we can all figure out what the signature patterns for a Harley should be.  From there we can look at data of one misbehaving.   

All the big data logs I have are off other systems.  There was a thread that Jason posted data that did do some of what you are asking.  I need a log.

Andy
Dynos are great for getting the motor close enough to get on the data loggers.

Mirrmu

Must be a tuner who can tune it properly, take for ride and post up?

whittlebeast

Or even a stock bike on a trip thru the mountains. .....
Dynos are great for getting the motor close enough to get on the data loggers.

FLTRI

Nope! Don't need Harley data to do the following:
"NOW! PLEASE! Take ONE friggin scatter graph of your choice and circle or point to where the bust you identify is in the graph and tell us what the bust is and how you see that. SIMPLE!
THEN! PLEASE! Show us the scatter graph with the fix to the bust and circle or point to where the picture changed after the bust was fixed. SIMPLE!"

Let's stay focused on the above request....

Bob
The best we've experienced is the best we know
Always keep eyes and mind open

Tsani

#23
Quote from: whittlebeast on August 08, 2015, 05:07:52 AM
Bob, I simply need a big data log off a really well tuned Harley being run hard on the back roads so we can all figure out what the signature patterns for a Harley should be.  From there we can look at data of one misbehaving.   

All the big data logs I have are off other systems.  There was a thread that Jason posted data that did do some of what you are asking.  I need a log.

Andy

Here's the problem:
1)Your asking for a log from a well tuned Harley run hard so that you can establish a pattern. Stock bike or modded? Because in a Harley, they all run different. That is part of the problem with tuning them.

2)Define big log file because I don't think your going to get what you want there either. At least not from anything short of a racer, the MOCO, or a Tuning shop/Tuning aid maker and I don't think they are going to give one up.

I just got my computer straightened out. Ready to post something up given what you asked for in another thread. But now that request has been modified. I cannot give you what you just asked for.  I will say I can pick out patterns based on changes I have made in a tune and that was just using the plots provided in the graphing function of my tuning software, but it was a evolving process. You want someone to post baseline info. You should have asked for this a long time ago, right from the start. It took all this time to get to this point.  I'm done. Observation time.
ᏣᎳᎩ ᎤᏕᏅ ᎠᏴ ᎠᎩᎸᏗ ᏔᎷᎩᏍᎩ ᎠᏂᏐᏈᎵ
ᎠᏎᏊᎢ Leonard Peltier

turboprop

Andy, has the MegaSquirt hardware been packaged yet into a bolt-on plug and play system for harley yet?
'We' like this' - Said by the one man operation.