Smoothing or Interpolating
Smoothing or Interpolating Hey all,
What do you guys do when you when tuning, smooth or interpolate, and how often.
Do you select the entire VE table and smooth?
Do you select 3-4 cells horizontally or vertically and smooth?
Do you select 3-4 cells horizontally or vertically and interpolate?
So what I am asking is how do you get your VE table smooth after making changes to only certain cells?
Thanks
What do you guys do when you when tuning, smooth or interpolate, and how often.
Do you select the entire VE table and smooth?
Do you select 3-4 cells horizontally or vertically and smooth?
Do you select 3-4 cells horizontally or vertically and interpolate?
So what I am asking is how do you get your VE table smooth after making changes to only certain cells?
Thanks
When I was doing VE tuning, I would only "interpolate" the cells I couldn't hit. Ideally, my best results for a given cell in the VE table came from good data filtering. What I mean by that is, the data points in the cell represent a value for a given MAP & RPM. The PCM will interpolate the data in between the cells for you. Smooth might look nice...but it doesn't make it accurate.
For example, say you're adjusting the 50kPa/3,200RPM cell. The data that is most accurate for that cell is the data that is focused around that cell's "center" (50/3,200). Getting creative with filters and only looking at say +/-1kPa and +/-50RPM from that "center" is a better way of tuning then to say, "I'll rely on my software's smoothing feature to manipulate the figure I'm going to use." So in this case, your observable range might be a set of rules that say, exclude data where:
RPM is <3150
RPM is >3250
MAP is <49
MAP is >51
This will give you better data that doesn't involve as much interference from the inaccuracy of neighboring cells. Without using said filters, your log for this cell will include data points where the MAP may have been 52.4 or where RPMs were 3395. Why is this a bad thing? Again, if the 55/3,200, 50/3,600, and 55/3,600 cells are wrong, they will affect the changes you're making for the 50/3,200 cell because the PCM was blending those values in during your drive as you drifted from the 50/3,200 cell's center.
Make sense?
For example, say you're adjusting the 50kPa/3,200RPM cell. The data that is most accurate for that cell is the data that is focused around that cell's "center" (50/3,200). Getting creative with filters and only looking at say +/-1kPa and +/-50RPM from that "center" is a better way of tuning then to say, "I'll rely on my software's smoothing feature to manipulate the figure I'm going to use." So in this case, your observable range might be a set of rules that say, exclude data where:
RPM is <3150
RPM is >3250
MAP is <49
MAP is >51
This will give you better data that doesn't involve as much interference from the inaccuracy of neighboring cells. Without using said filters, your log for this cell will include data points where the MAP may have been 52.4 or where RPMs were 3395. Why is this a bad thing? Again, if the 55/3,200, 50/3,600, and 55/3,600 cells are wrong, they will affect the changes you're making for the 50/3,200 cell because the PCM was blending those values in during your drive as you drifted from the 50/3,200 cell's center.
Make sense?
I only smooth cells I can't hit. As mentioned, if the data says the cell should be that value, then don't change it, unless it's obviously incorrect. But filtering the data should prevent most incorrect values. The PCM itself interpolates between cells when running, so it will run a lot smoother than it looks.
When I was doing VE tuning, I would only "interpolate" the cells I couldn't hit. Ideally, my best results for a given cell in the VE table came from good data filtering. What I mean by that is, the data points in the cell represent a value for a given MAP & RPM. The PCM will interpolate the data in between the cells for you. Smooth might look nice...but it doesn't make it accurate.
For example, say you're adjusting the 50kPa/3,200RPM cell. The data that is most accurate for that cell is the data that is focused around that cell's "center" (50/3,200). Getting creative with filters and only looking at say +/-1kPa and +/-50RPM from that "center" is a better way of tuning then to say, "I'll rely on my software's smoothing feature to manipulate the figure I'm going to use." So in this case, your observable range might be a set of rules that say, exclude data where:
RPM is <3150
RPM is >3250
MAP is <49
MAP is >51
This will give you better data that doesn't involve as much interference from the inaccuracy of neighboring cells. Without using said filters, your log for this cell will include data points where the MAP may have been 52.4 or where RPMs were 3395. Why is this a bad thing? Again, if the 55/3,200, 50/3,600, and 55/3,600 cells are wrong, they will affect the changes you're making for the 50/3,200 cell because the PCM was blending those values in during your drive as you drifted from the 50/3,200 cell's center.
Make sense?
For example, say you're adjusting the 50kPa/3,200RPM cell. The data that is most accurate for that cell is the data that is focused around that cell's "center" (50/3,200). Getting creative with filters and only looking at say +/-1kPa and +/-50RPM from that "center" is a better way of tuning then to say, "I'll rely on my software's smoothing feature to manipulate the figure I'm going to use." So in this case, your observable range might be a set of rules that say, exclude data where:
RPM is <3150
RPM is >3250
MAP is <49
MAP is >51
This will give you better data that doesn't involve as much interference from the inaccuracy of neighboring cells. Without using said filters, your log for this cell will include data points where the MAP may have been 52.4 or where RPMs were 3395. Why is this a bad thing? Again, if the 55/3,200, 50/3,600, and 55/3,600 cells are wrong, they will affect the changes you're making for the 50/3,200 cell because the PCM was blending those values in during your drive as you drifted from the 50/3,200 cell's center.
Make sense?
Very well said. I agree 100%
What you do depends on what you're looking to fix.
I avoid the "easy way" table modification schemes
(no "paste histo" for me). I work with data I believe
in, and reject data I don't, by eye and hand. When
you update specific cells they change color, so you
know what's been touched. Then I look at the response
surface and where the old data looks out-of-place, I
interpolate between known-good points or, if there's
nothing on the other side, fiddle by hand so the shape
just "looks right".
Only at the very end, will I "smooth" things. Smoothing
touches every data point, and some of them ought to
be left alone - you don't get to say, which.
I avoid the "easy way" table modification schemes
(no "paste histo" for me). I work with data I believe
in, and reject data I don't, by eye and hand. When
you update specific cells they change color, so you
know what's been touched. Then I look at the response
surface and where the old data looks out-of-place, I
interpolate between known-good points or, if there's
nothing on the other side, fiddle by hand so the shape
just "looks right".
Only at the very end, will I "smooth" things. Smoothing
touches every data point, and some of them ought to
be left alone - you don't get to say, which.
