Acura RSX, ILX and Honda EP3 Forum banner
1 - 20 of 777 Posts

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #1 · (Edited)
UPDATES:

New offline version of RazDyno 1.0 -- includes the ability to tune number crunching parameters.
New offline windows program written by theta called ThetaDyno that uses same principals.

Original text of first post in thread is below...

Okay... I've had somewhat of a itch in my brain in the problem solving area.

Premis:
Tuning for POWER with kpro should be 'doable' by examining acceleration at fixed increments across the rpm band given static test conditions (or fairly static). When tuning for power, you are NOT really interested in an actual HP/TQ number as much as finding the combination which produces the highest one. In order to understand the value of what I am suggesting, you should already be familliar with the general process of tuning a 4 stroke internal combustion engine and more specifically, the general process that hondata suggests using when tuning k series motors with kpro. If you are not familiar with kpro tuning... it is best to become familiar with it before reading further.

Background:
Basically, I have seen a few people talk about offering a 'street tuning' service where they use acceleration calculated from your datalogs to determine optimum cam angles. This is a somewhat novel idea, perhaps because it was using the same principal I was using with my in car dynorad system to see if tuning for power was possible without shelling out large quanities of money for dyno time to get everything perfect. A thread with some of my results from installing and using the dynorad product are here. I just want to make it 100% clear that I do not support the idea of sending your datalogs somewhere and blindly running a tune created by another person who hasn't actually been working with your car.

My Goal / Proposal:
I am a software engineer. I know a bit about physics from college. This means that I can bang my head on the wall and write an very well optimized routine to explain exactly why it hurt.

My goal is to potentially develop a tool for people who are already tuning a kseries motor with kpro that would remove the dependency on a chassis dyno such as a dynojet. I don't intend to develop something that will allow joe bob noob to go out and tune with kpro -- I simply want to remove the dependency on spinning large drums to gather data. Also, real world loads where the air moving around your car is different than a dyno should provide greater real world results compared to that which is available by simply blowing a fan at your car while gunning it in 3rd or 4th on rollers.

The tool I am suggesting would be a simple tool that would analyse exported datalogs that were done under fairly static test conditions (time of day, temp, same stretch of road / highway going the same direction, same wind conditions, same gear, same weight, tire pressure... etc). The tool would go through the datalog to find the first place the map sensor readings went wide open (say 80 kpa or above) and the tps readings indicated the pedal was to the metal (say 95% or more), and use the data about how fast rpm was rising to calculate 'acceleration units' at a fixed set of rpm intervals.

Using this data, graphing etc, should provide one with the data required to tune for power in the multiple times tuning for power is desired when tuning with kpro. The most useful of which is probably identifying cam angles by comparing fixed cam results for the high and low cam, and tuning ignition for a power plateau.

Physics Explanation / Justification
Based on the analysis of the acceleration based on rpm change over time, the tool would then spit out 'acceleration units' at a given interval.. say 100 rpm steps that can easily be loaded into say an excel spreadsheet and graphed. Comparing these 'acceleration units' are in essence identifying where more power is generated. The reason for this is simple: Acceleration (the rate of change of velocity with respect to time) is the variable which changes FORCE. The force I am speaking about is TORQUE. Torque in combination with RPM generates horse power. Given that MASS is static (static enough, burning fuel mass going away reduces mass... but you get the picture), force can be calculated based on acceleration alone. This means that when you reduce it all down, what you are really wanting to know is what generates the most acceleration at each rpm increment.

Don't believe me? Please review the following:

What is Force?
What is Torque?
What is Work?
What is Power?

Also, measuring acceleration is how a dynojet works -- they simply have a more static environment. The fundemental difference between doing a dynojet run is a dynojet measures the acceleration of BIG drums your wheels turn while we will measure the acceleration of the car. Given that the acceleration of your car has MANY more variables than just spinning drums, you cannot simply calculate HP from acceleration -- although it is in there at some level. The dynorad system I purchased does a pretty good job at giving you hp numbers that match up with a dynojet once calibrated, but you need to provide weather station info, test weight, drag factors, and rotating mass factors.

How a chassis dyno (dynojet for example) works

Now given that we really don't care WHAT the hp/tq number, just to know that we have generated the highest possible, we really only need to compare acceleration. Going any further will only provide you with a number to tell your friend when he asks how much power you are pushing. All you need is acceleration to determine if condition A generates more power than condition B.

Given that acceleration is harder at 60 (requires more power) than at 5 mph, one can say that acceleration is not a valid test since if you are going faster, more power is required to generate the same acceleration. This is not a problem either... the reason being is the manual transmission that makes my 03 type-s go 65mph right at 6k in 3rd gear. The rate of change from 6500 to say 6600 rpm always happens at the same speed... so if the change to 6600 rpm happens faster than previous, more power is generated!

Conclusion:
I am probably going to write such a utility for my own personal useage in analysing my datalogs just to see if it's a viable idea. I would like feedback to know if anyone else would be interested in this utility. I plan to make it basically operate in a few modes... something like comparing run files for tuning ignition and comparing a set of run files for finding cam angles and vtec change point.

Let me know what you think. I am already aware of the issue of test conditions being similar leading to inaccurate results -- I am hoping that given a small enough test window and the same stretch of road, comparison is possible.
 

·
Registered
Joined
·
7,639 Posts
Those are very good assumptions; however, the precision of the data that the kpro records is not great enough to determine acceleration numbers.

A long time ago I tried using acceleration along with some assumptions to create a 'street dyno' graph from the k-pro's datalogs. Here is a screen shot of the data:



The column labeled 'a' is the acceleration. Notice how there are too many voids in the data because the precision of the velocity data is not high enough, creating many points labeled as '0' for 'dv' (change in velocity).
 

Attachments

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #3 ·
park83rsx said:
Those are very good assumptions; however, the precision of the data that the kpro records is not great enough to determine acceleration numbers.

A long time ago I tried using acceleration along with some assumptions to create a 'street dyno' graph from the k-pro's datalogs. Here is a screen shot of the data:
This problem has also been adressed in a 'smoothing' and averaging of the results across a larger window to generate more error free results. This is what my dynorad does. I would be interested in knowing more about your attempts so I can see if I have any new ideas or can spot anything you may have missed.

I see that you did this in 3rd gear from the ratio -- that should be plenty to get granular enough data -- were you using the speed reading only or did you try RPM. The reason I say this is because RPM is much more granular than the speed sensor. If you note, it's always an integer. There may be more luck in using RPM since in a manual, rpm and speed are fixed and speed seems to have precision cut off drastically where as rpm would seem to have almost 2 decimal points worth of data compared to speed.
 

·
Registered
Joined
·
7,639 Posts
Razathorn said:
This problem has also been adressed in a 'smoothing' and averaging of the results across a larger window to generate more error free results. This is what my dynorad does. I would be interested in knowing more about your attempts so I can see if I have any new ideas or can spot anything you may have missed.

I see that you did this in 3rd gear from the ratio -- that should be plenty to get granular enough data -- were you using the speed reading only or did you try RPM. The reason I say this is because RPM is much more granular than the speed sensor. If you note, it's always an integer. There may be more luck in using RPM since in a manual, rpm and speed are fixed and speed seems to have precision cut off drastically where as rpm would seem to have almost 2 decimal points worth of data compared to speed.
You may want to try one with RPM. I was using speed because I wanted to get hp and torque numbers. This was my final graph with smoothing. Lots of smoothing.

 

Attachments

·
Registered
Joined
·
13,797 Posts
I'd definately be interested in this as I do not have the time nor money (it's mainly money) to sit on the dyno for hours for a full tune. Measuring acceleration has been thrown around a few times and if you can peg it perfectly..even nearly perfect, a useful utility it will be to many.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #9 ·
park83rsx said:
You got my mind working now (weird too, i was just at the bar...lol) so I did this. This is change in RPM with time, graphed over the RPM range:

Looks a lot like our torque curve.....mostly coincidence though.
Why wouldn't it look like it. If you really are capturing a 'unit of acceleration' at a point in time, it should mirror the torque curve in every aspect except apmlitude. Think about it, acceleration is one of the parts of force .. how you get to torque. The reason it won't look like HP is because HP has rpm thrown in there.. which, for lack of a better way to put this, makes hp crawl up the graph ;).

I think that given the proper test data range (length of say, 3rd gear), and the granularity of the rpm and distance between sample length, a pretty accurate graph of acceleration could be produced. My dynorad does a pretty darn good job in 2nd gear using the SPEED sensor (which it measures pulses directly from, of which there are 4k + per mile). I think if we crunch the numbers a bit, we can nail it down.

I think i'm going to go grab some datalogs tomorrow of a few 3rd gear pulls and see what I come up with.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #10 ·
Another thing that you weren't doing that I am planing to do is to interpolate a bit so that my figures land on even 100 rpm increments. That's a requirement any way you slice it since to have a fixed sample length to have vertical comparison points. Also, it seems that you may have been tracking things without relevance to difference per time t. When calculating a delta value were you simply taking the delta between the two samples or additionally comparing the time between samples in making an acceleration over time comparison?
 

·
Registered
Joined
·
7,639 Posts
Your right, it should mirror the torque curve pretty closely, I dont know what I was thinkin. :drink:

Seeing the data in larger increments would also smooth things out.



Also, I have a write-up I have done on another method and hope to hit up the dyno jsut to see how close it is to that of a map which gave me good results on the dyno.
 

·
EFI Tuning Professional
Joined
·
8,139 Posts
How do you guys create hp/tq graphs using your datalogs?

Thought I might chime in here since I like this method of tuning

I have tested the acceleration method on my car, 2 other cars and palmerblocks car. Also, i came to understand that tuning someone elses car while not being out there will not work due to the fact that there are many variables that you have to watch (example is swerving through traffic to avoic accidents while trying to log your data, IAT temps, tire pressure, basically everything a dyno-jet figures, as Razathorn clearly stated). Hence, it would have been wrong to charge due to the fact that data will be incorrectly incorporated into this method.

After i tuned a/f and compared the acceleration times per 500 rpm, it just so happened that it came to the cam angles that were dyno tuned. Palmerblock can testify to this.

However, the 2 other cars that i street tuned using acceleration comparing showed no gains over provided maps due to very restrictive exhaust systems. Also, when we dynoed, the ecu retarded 4 degrees timing, so we really couldnt tell if it gained any. I didn't touch ignition on these 2 cars, just a/f and cam angle.

The interesting thing is that comparing acceleration times for every cam angle will net you a very similar/almost exact map as provided by k-pro.

For example, if you have a custom header that doesn't have a calibration on k-pro, it is essential that you make some sort of basemap and then head to the dyno to compare the cam angles of another header to the cam angles you made for your custom header.

I am currently on brake from testing due to very hot temps(ecu retards 4 degrees of ignition to prevent knock) here in Ca along with my car needing a custom fuel setup (which i have to install in this hot weather)

However, when temps go down to the 80's, I will take one of those 2 cars i tuned and do some more tests with ignition timing ( i do most tests at night where there are less cars and temps are reasonable) and compare acceleration times to find out if the extra 2 or 4 degrees netted a faster acceleration

What would be nice is a excel spread sheet formula that takes the data from the datalog and incorporates it into excel, to decrease street tuning time (the first time i did the a/f and cam angle maping took 5 hours, 2 hours just to right everything down and then go drive to find the optimum vtec point. Also, this spreadsheet should take consideration of IAT temps (for example, if u dataloged 10 degrees at 76 degrees IAT and 30 degrees at 78, the formula can calculate a new acceleration time)

I am not good with excel, so if we all work together, I think we can all make this work. Of course, head to the dyno numbers to compare maps and hp differences between different cam angle setups and ignition timing

Mikey
 

·
Speedy
Joined
·
2,648 Posts
there are canned software packages (minitab, matlab, addins for excel) for you guys are going (numerical methods). just import your data and go..

i've had the classes, but so long ago all i remember is that it can be done. anyone a little fresher?
 

·
Registered
Joined
·
24,665 Posts
while this is all nice on data analysis....but the big flaw in this is aquisition of data. on a dyno, the load stays the same. on the road, you're dealing with wind, bumps in the road, non perfect flatness, etc etc. in order to try to replicate the same data, you would have to drive down the same exact section of road from a start marker to end marker and make sure that day it's 0mph wind and your IAT did not heat up from idling around while you were making changes.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #16 ·
HardcoreRsxS said:
How do you guys create hp/tq graphs using your datalogs?
You don't. You get acceleration differences. Acceleration combined with a bunch of other unknowns (and furthermore, not constant across the rpm range) gets you hp and torque. You don't want torque and hp as much as you want 'something' to show what produces more power. That something is acceleration.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #17 ·
krystar said:
while this is all nice on data analysis....but the big flaw in this is aquisition of data. on a dyno, the load stays the same. on the road, you're dealing with wind, bumps in the road, non perfect flatness, etc etc. in order to try to replicate the same data, you would have to drive down the same exact section of road from a start marker to end marker and make sure that day it's 0mph wind and your IAT did not heat up from idling around while you were making changes.
Yes, this is the one major crux of the whole thing. But on a dyno, the same thing is present as far as the air temp thing. My 3rd run on a dyno is ALWAYS the best for some reason.

The solution for this is to do multiple pulls under 'nearly' static conditions and average them out. A trend should appear if there is one.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #18 ·
TheSilverBullet said:
there are canned software packages (minitab, matlab, addins for excel) for you guys are going (numerical methods). just import your data and go..

i've had the classes, but so long ago all i remember is that it can be done. anyone a little fresher?
Hopefully something I would make would simply be a give away that was a simple 'run this' with your datalogs and it would produce answers given correctly taken datalogs. A bit simpler and less expensive. Plus, it sounds like it's pretty heavy stuff considering there were classes for it ;).

I would love to know more about this software / plugins you speak of in case there are things I can learn from them.
 

·
glory to the hypnotoad
Joined
·
6,295 Posts
Discussion Starter · #19 · (Edited)
Btw guys -- if you haven't checked out my dynorad thread, check it out. The link is in the first post of this thread. Basically the dynorad has produced results that when averaged are VERY useful in determining things. You should take a look at my comparison of before and after ESMM install to see exactly what small differences the dynorad was able to produce. I *could* tune entirely from my dynorad -- BUT, entering in power data BY HAND for 100 rpm increments is just too time consuming.. I wanted to DATALOG, view acceleration, compare with others, all in a very short time frame so I could actually do some tuning instead of typing.

That is the major push for me to implement something like this -- basing things from datalogs that come right off the computer.

BTW: page 2 on the dynorad thread has the best graphs.
 

·
x type-s club
Joined
·
5,387 Posts
Interesting stuff, I have worked with Matlab/Excel quite a bit, PM me if you need a hand in some programming, I think a simple excel sheet with some macros should be able to do the job here, to evaluate your process anyways...
 
1 - 20 of 777 Posts
Top