In statistics this is commonly referred to as sampling error. A sampling error is a deviation of the sampled value from the real value. The method of reducing error is usually to capture more samples, given that uniformly distributed random errors tend to cancel out when averaged. By making the bins narrower, it spreads out the data among more bins, capturing less samples per bin, yet each sample still retains 100% of the sampling error in each sample. I think the solution here is to capture more samples, and focus on taking those samples at steady-state moments where there is less systematic error.
Thank you for the correct term for this. While I won't argue about your statement in general about statistics: "By making the bins narrower...each sample still retains 100% of the sampling error..." I will say that in terms of MAF tuning, the the amount of error is highly reduced when combining (1) bin size reduction with the (2) Dynamic Air formulas I posted in other vides and on the HPT forum. It works like magic.
I don't worry so much about cell hits for MAF (for VVE it is much more important IMHO). For MAF I like to hold each cell and I watch the value change. At first it changes a lot and changes rapidly. After a second or two you will see it start to settle on a final number. I also prefer to drive for at least 30 minutes, but 45 minutes even better, with a slow and smooth throttle pedal input. After 45 minutes you should have enough MAF cell counts no matter what. AND none of this counts until the engine...I should say OIL...temp is at full temp. A simple drive around the block on a cold motor isn't going to work if you want the best data which leads to the best results.
@@stevenwilliams263 when you first start on your VVE quest more cell hits is important....perhaps 50-100. Once you start to tame the shape after a few revisions, then you can reduce that down to say 10-20 hits. When you look at the RPM and MAP breakpoints there is a HUGE amount of averaging going on in the scanner for each cell, so I look at the VVE scanner data as clues to the shape it wants. I am just looking for a trend that I can interpret the final shape.
It still looks bumpy, I tuned my 3800 v6 MAF in two rounds by logging LTFT with hz vs MAP then used a high order polynomial fit in excel to smooth. Went from +\- 10% to 1%. Simple procedure, complicated spreadsheet. Adding MAP to the MAF log was very interesting… For a given hz/flow you will in some spots find quite a difference in LTFT vs MAP. I figure this is because of double counting, and should get worse with higher overlap….. I ended up filtering by 10 hits per cell minimum, linearize with a power curve fit and then smooth with a 5th order quadratic. Next up is an L83, I’ll be using your tools for that one.
I think you're referring to "double counting" air that reverses past the hot wire a second time as a result of resonance. You don't have any control over that in MAF tuning. With MAF tuning you only control the relationship between Hz and mass flow. What you find in logs is essentially a normal distribution, so averaging is the best approach. To address the limitations of simple MAF control (which actually does work pretty damn good in the real world!) the concept of VVE was created to combine the accuracy of directly measuring mass flow with the beneficial 3D aspects of MAP tuning. @@snivesz32
Using the maf analyzer, so far so good. but if i try to highlight the Hz Range +/- and i try to type 25. It buffers for a minute and just goes up or down 1. Trying to go from 50 to 25 would take me like 30 minutes. I just downloaded this today 8/12/2024 and I downloaded the most recent one (jan 10) off the google doc. May be something to look into!
Hmm I am not sure...everything work ok on my end. There is usually some delay when changing the value as the code has a lot of data manipulation, but nothing more than a second or so (at least on my laptop). Are you typing in the number or using the up/down arrows? I would definitely type in a number (never use the arrows). What happens if you put in 25 Hz before you paste in your first MAF table?
Are you saying to allow the user to input the MAF table from the Editor (the grams/sec) and then apply the fuel trims and visually graph it all out showing before and after?
@@SilverSurfer77 Do it like the VVE tool, where you can put the grams/sec table back into the tool after making corrections to see the expected trims afterwords - this way if you smooth the MAF table any you can check the work :)
I tune with a wideband...I'm guessing nothing changes with the cell expansion and data extrapolation if I'm gathering fuel trims by percentage? This would totally explain why I'm getting such weird data fluctuation day to day when tuning. Thanks for the information.
@@TheSaxon25 narrowbands are essentially an EPA controlled device. They only do one thing, but they do it exceptionally well...which is to report back on stoich. Since they are so accurate and you should exploit their accuracy at every chance you get. Commercially available widebands on the other hand are not super accurate and you should always treat their results with a grain of salt. If you tune in OL with a wide band and then enable CL narrowvbands...now you are switching the input to the computer and the computer does not care about what the wideband says. Basically I feel like you should tune the ECM with the conditions and equipment that it is going to operate in.
I just did 2 separate MAF logs. I have 1 in Closed loop and 1 in open loop. My closed loop (STFT only) is vastly different (leaner) than my open loop log. Any ideas on why this is?
Cringer, you are really stretching it with this maf software. Tuning maf is pretty straight forward. 3rd party software just makes it complicated. Your maf/ve/cl/ol/math/filters will make beautiful logs but will not eliminate false knock. You need to eliminate fuel trims from your math (fuel trims are already factored into the DA calculation) and take a closer look at your airflow mode boundary filters. Do this and you won't need to overly smooth your spark table to get rid of knock. Tuning maf and ve at the same time has been around for years, but very few have the math correct. As tuner with a computer science background, I think your videos are great...Right or wrong, makes me think about better ways to tune.
I think the answer to your comment deservers a full video response, but I am not going to do that at this time. You are correct. Basic MAF tuning is simple. But ask yourself this question, "What is the difference between a guy who has been tuning for a month vs the professional who has been tuning for years? How do they tune the MAF differently?" The professional will wait for the car to get in closed loop, and wait for the coolant and oil to get to full operating temperature, and drive with a slow and steady throttle input...and he will also use filters in the scanner to filter out junk data. Other than that their methods are very similar. What does the average tuner (or person in general) know or care about statistics and data sampling? Not much if I had to guess. However, the tuner in the above scenario did use the Scanner filter formula to filter out junk data! So part of him understands that statistics is important. The MAF Assistant takes this concept to the next level. The goal of this tool, and all the tools I write is to get you, not to just get to a "done tuning state", but rather a "perfect" tune state quicker and easier. Now you referred to me as my forum name, Cringer. So surely you have read my signature line, which states: A standard approach will give you standard results. I am not looking for standard, basic, quick, or everyone does it this way types of approaches. I am always on a quest to learn and do more, so if you have new ideas to contribute and share, I am all ears!
Tuning a 1.4T E78 Cruze and this tool works like a charm! Onto VVE Assistant next 👍 thanks
In statistics this is commonly referred to as sampling error. A sampling error is a deviation of the sampled value from the real value. The method of reducing error is usually to capture more samples, given that uniformly distributed random errors tend to cancel out when averaged. By making the bins narrower, it spreads out the data among more bins, capturing less samples per bin, yet each sample still retains 100% of the sampling error in each sample. I think the solution here is to capture more samples, and focus on taking those samples at steady-state moments where there is less systematic error.
Thank you for the correct term for this. While I won't argue about your statement in general about statistics:
"By making the bins narrower...each sample still retains 100% of the sampling error..."
I will say that in terms of MAF tuning, the the amount of error is highly reduced when combining (1) bin size reduction with the (2) Dynamic Air formulas I posted in other vides and on the HPT forum. It works like magic.
All of your videos are golden man.
This guy is on a roll !!
Awesome mate! I can't wait to put my engine back together to try.
Your a legend mate love all of your videos great work appreciate it.....PS loved the intro Bumpy Maf
Thank you for the kind words! Glad I can help the community out!
Thank you so much for your hard work 💪🏻 you rock
You are welcome!
Another excellent video, keep them coming please.
Awesome and informative! Quick question, how may cell hits do you suggest for our MAF logging tables?
I don't worry so much about cell hits for MAF (for VVE it is much more important IMHO). For MAF I like to hold each cell and I watch the value change. At first it changes a lot and changes rapidly. After a second or two you will see it start to settle on a final number. I also prefer to drive for at least 30 minutes, but 45 minutes even better, with a slow and smooth throttle pedal input. After 45 minutes you should have enough MAF cell counts no matter what. AND none of this counts until the engine...I should say OIL...temp is at full temp. A simple drive around the block on a cold motor isn't going to work if you want the best data which leads to the best results.
10/4. So MAF will stay at default. What would you say is a good number for VVE? I appreciate you!
@@stevenwilliams263 when you first start on your VVE quest more cell hits is important....perhaps 50-100. Once you start to tame the shape after a few revisions, then you can reduce that down to say 10-20 hits. When you look at the RPM and MAP breakpoints there is a HUGE amount of averaging going on in the scanner for each cell, so I look at the VVE scanner data as clues to the shape it wants. I am just looking for a trend that I can interpret the final shape.
X2 I'm going to try this tool to chekc if I can get rif of a gaggy maf or to avoid blending much for just to get a nice curve maf.
It still looks bumpy, I tuned my 3800 v6 MAF in two rounds by logging LTFT with hz vs MAP then used a high order polynomial fit in excel to smooth. Went from +\- 10% to 1%. Simple procedure, complicated spreadsheet. Adding MAP to the MAF log was very interesting… For a given hz/flow you will in some spots find quite a difference in LTFT vs MAP. I figure this is because of double counting, and should get worse with higher overlap….. I ended up filtering by 10 hits per cell minimum, linearize with a power curve fit and then smooth with a 5th order quadratic. Next up is an L83, I’ll be using your tools for that one.
I am not going to lie, that sounds intense! Can you make a video and show me how this is implemented and executed???
Sure, I’ll clean up the excel sheet and send it to you. You should be able to find a library that accomplishes the same thing.
@@shane-222 thank you!
@@SilverSurfer77 I shot you a PM on HPtuners.
I think you're referring to "double counting" air that reverses past the hot wire a second time as a result of resonance. You don't have any control over that in MAF tuning. With MAF tuning you only control the relationship between Hz and mass flow. What you find in logs is essentially a normal distribution, so averaging is the best approach. To address the limitations of simple MAF control (which actually does work pretty damn good in the real world!) the concept of VVE was created to combine the accuracy of directly measuring mass flow with the beneficial 3D aspects of MAP tuning. @@snivesz32
Using the maf analyzer, so far so good. but if i try to highlight the Hz Range +/- and i try to type 25. It buffers for a minute and just goes up or down 1.
Trying to go from 50 to 25 would take me like 30 minutes. I just downloaded this today 8/12/2024 and I downloaded the most recent one (jan 10) off the google doc. May be something to look into!
Hmm I am not sure...everything work ok on my end. There is usually some delay when changing the value as the code has a lot of data manipulation, but nothing more than a second or so (at least on my laptop). Are you typing in the number or using the up/down arrows? I would definitely type in a number (never use the arrows). What happens if you put in 25 Hz before you paste in your first MAF table?
You need to put the expected trim return function in this one too as a separate page or something like that just to keep that part separate.
Are you saying to allow the user to input the MAF table from the Editor (the grams/sec) and then apply the fuel trims and visually graph it all out showing before and after?
@@SilverSurfer77 Do it like the VVE tool, where you can put the grams/sec table back into the tool after making corrections to see the expected trims afterwords - this way if you smooth the MAF table any you can check the work :)
@@GHuggins this is a good idea and a reasonable request. It shall be done!
Do you need to re-scale the MAF columns in Editor to include/match the new column values on the MAF tool before pasting in the values?
Nevermind... I just saw the note in the tool that it doesn't copy the dummy cells. I suck at reading. LOL
I tune with a wideband...I'm guessing nothing changes with the cell expansion and data extrapolation if I'm gathering fuel trims by percentage? This would totally explain why I'm getting such weird data fluctuation day to day when tuning. Thanks for the information.
Yes same process for wideband. But I don't like to use wideband tuning for stoich operations.
@@SilverSurfer77 is it because the resolution for stoich is more precise on narrow bands than wide bands?
@@TheSaxon25 narrowbands are essentially an EPA controlled device. They only do one thing, but they do it exceptionally well...which is to report back on stoich. Since they are so accurate and you should exploit their accuracy at every chance you get. Commercially available widebands on the other hand are not super accurate and you should always treat their results with a grain of salt. If you tune in OL with a wide band and then enable CL narrowvbands...now you are switching the input to the computer and the computer does not care about what the wideband says. Basically I feel like you should tune the ECM with the conditions and equipment that it is going to operate in.
@@SilverSurfer77 thanks for the explanation. Makes perfect sense
Do I need to go into speed density mode to log for maf with your layout and channels
hi , the utility is not in the drive file only .pptx and .txt
Sorry! I updated it the other day and I guess I forgot to upload the new version. Should be all now.
@@SilverSurfer77 thank you, very useful tools
I just did 2 separate MAF logs. I have 1 in Closed loop and 1 in open loop. My closed loop (STFT only) is vastly different (leaner) than my open loop log. Any ideas on why this is?
Too many variables to say for sure. Trust your CL log. Narrow bands O2s are more accurate and that's what your ECM uses to control fueling.
Thanks
I would like to try your tool too. If you could throw that link too please.
I have a high and low which would be the numbers of 12161 and 12162
You will need to do this twice, once for the Low and High values.
Cringer, you are really stretching it with this maf software. Tuning maf is pretty straight forward. 3rd party software just makes it complicated. Your maf/ve/cl/ol/math/filters will make beautiful logs but will not eliminate false knock. You need to eliminate fuel trims from your math (fuel trims are already factored into the DA calculation) and take a closer look at your airflow mode boundary filters. Do this and you won't need to overly smooth your spark table to get rid of knock. Tuning maf and ve at the same time has been around for years, but very few have the math correct. As tuner with a computer science background, I think your videos are great...Right or wrong, makes me think about better ways to tune.
I think the answer to your comment deservers a full video response, but I am not going to do that at this time.
You are correct. Basic MAF tuning is simple.
But ask yourself this question, "What is the difference between a guy who has been tuning for a month vs the professional who has been tuning for years? How do they tune the MAF differently?" The professional will wait for the car to get in closed loop, and wait for the coolant and oil to get to full operating temperature, and drive with a slow and steady throttle input...and he will also use filters in the scanner to filter out junk data. Other than that their methods are very similar.
What does the average tuner (or person in general) know or care about statistics and data sampling? Not much if I had to guess. However, the tuner in the above scenario did use the Scanner filter formula to filter out junk data! So part of him understands that statistics is important. The MAF Assistant takes this concept to the next level.
The goal of this tool, and all the tools I write is to get you, not to just get to a "done tuning state", but rather a "perfect" tune state quicker and easier.
Now you referred to me as my forum name, Cringer. So surely you have read my signature line, which states: A standard approach will give you standard results. I am not looking for standard, basic, quick, or everyone does it this way types of approaches.
I am always on a quest to learn and do more, so if you have new ideas to contribute and share, I am all ears!