Odd VE table behavior

[Hib Halverson]

I wanted to copy the entire VE table from one E38 cal to another without making any changes.

I did that by selecting all the cells in the first table, copying them to the clipboard, then selecting all the cells in the second table then pasting all the values from the first into the second.

My belief is there would be no need to generate coefficients or VVE but after pasting the new values on the second table, I see the eclamation point telling me I have to generate new coefficients and generate VVE and, after that when I compare the first table to the second, there are differences.

Also, this problem is not consistent. I can try the same procedure multiple times and different cells end up being changed.

I'm stumped.

Anyone have any idea why that's happening?

[joecar]

Hi Hib,

If you changed the VVE table then you would want to generate the coefficients.

Or, copy the VE coefficients table.

Is this on the 2014 E92...? I'll try it.

Joe

[Hib Halverson]

Hi Hib,

If you changed the VVE table then you would want to generate the coefficients.I didn't change anything in any of the cells, I simply copied all the data.
Or, copy the VE coefficients table.Ok.

Is this on the 2014 E92...? I'll try it.

JoeNo. It's a 2012 E38.

[joecar]

If you changed the VVE table then you would want to generate the coefficients.


I didn't change anything in any of the cells, I simply copied all the data.Ok.No. It's a 2012 E38.

Ah, what I meant to say: if you changed the VVE table (even by pasting into it), then you would need to generate coefficients.

i.e. the ECM uses the coefficients... you manually generate (by clicking the button) the coefficients from the Virtual VE table.

[joecar]

Ah, ok, 2012 E38.

[joecar]

I...

Also, this problem is not consistent. I can try the same procedure multiple times and different cells end up being changed.

...This is odd... I'll let the software guy know.

[Blacky]

I wanted to copy the entire VE table from one E38 cal to another without making any changes.

I did that by selecting all the cells in the first table, copying them to the clipboard, then selecting all the cells in the second table then pasting all the values from the first into the second.

My belief is there would be no need to generate coefficients or VVE but after pasting the new values on the second table, I see the eclamation point telling me I have to generate new coefficients and generate VVE and, after that when I compare the first table to the second, there are differences.

Also, this problem is not consistent. I can try the same procedure multiple times and different cells end up being changed.

I'm stumped.

Anyone have any idea why that's happening?

The "Virtual VE" table data is computed from the "VE Coefficients" table using complex, non-linear, polynomial equations. Those equations will always produce exactly the same VVE table data from the same coefficients. However, going the other way - i.e. generating the coefficients table data from the VVE table data may not (and usually will not) produce exactly the same coefficients that were used to generate the VVE data in the first place. Rounding errors, curve fitting techniques, and even the number and distribution of sample points all play a part in "reverse engineering" the coefficients from the VVE data.

You can see the same result if you edit just a single cell, then edit it so that it has its original value (but not using undo), so that the software "thinks" the VVE data has been modified (i.e. at least one cell has blue corners). Then click [Generate Coefficients]. That will update the coefficients table with new values estimating the different curves of the VVE table zones. Then click [Generate VVE] which will use those modified coefficients to plot a slightly different version of the VVE data. Each time you perform that cycle the VVE data is further estimated/modified/updated etc.

Updating the coefficients is where the estimations are done and small rounding errors may appear. So if/when you want to copy a VVE table without change from one calibration to another, copy the coefficients table then generate the VVE from those coefficients.

Regards
Paul

[Hib Halverson]

Hi Paul-

The "Virtual VE" table data is computed from the "VE Coefficients" table using complex, non-linear, polynomial equations. Those equations will always produce exactly the same VVE table data from the same coefficients.That part, I know.


However, going the other way - i.e. generating the coefficients table data from the VVE table data may not (and usually will not) produce exactly the same coefficients that were used to generate the VVE data in the first place. Rounding errors, curve fitting techniques, and even the number and distribution of sample points all play a part in "reverse engineering" the coefficients from the VVE data.

Ok, so....pasting data into one VVE table from another, functionally, is trying to "reverse engineer" the cooefficients in the receiving file from the VVE cell data one just pasted?

[joecar]

Hi Paul-
...

Ok, so....pasting data into one VVE table from another, functionally, is trying to "reverse engineer" the cooefficients in the receiving file from the VVE cell data one just pasted?Yes (after you click Generate Coefficients).

[Hib Halverson]

Yes (after you click Generate Coefficients).

Got it.

Thanks Paul and Joe.
You guys are your usual responsive selves.

I appreciate that.

[Naf]

You know i have seen changes in my VVE as well.

I will make my changes,
click on Generate Coe...
it will all hold steady, then when i click on the save/save as the table changes in the slightest way...

[joecar]

You know i have seen changes in my VVE as well.

I will make my changes,
click on Generate Coe...
it will all hold steady, then when i click on the save/save as the table changes in the slightest way...Can you post some screenshots before/after saving.

[Taz]

Think of it as a "rounding" adjustment (as it is not really an error). With the VVE tables, you are using a 4 digit whole number to reverse generate numerous coefficients.

You will see a similar effect with T43 (6L80) shift times, if you use more than 3 decimal places - there will be a very slight change when reopening the tune.

[joecar]

That is called bit-representation-error...

each bit position in a binary fixed point point number represents a fractional power of two, i.e. (1/2)^n where n is bit position.


For example, how would 0.6 be represented in simple fixed point binary...
0.500 = 0.100
0.250 = 0.010
0.125 = 0.001
you combine those to get other values:
0.750 = 0.110
0.875 = 0.111
but you can't represent 0.6 exactly, the closest to 3 or 4 bits is 0.101 = 0.625 or 0.1001 = 0.5625
if you go to more bits (binary decimal places), you get closer, but never exact; and the word has some fixed number of bits (say 8 or 16 or 32 or 64).


why-cant-decimal-numbers-be-represented-exactly-in-binary (http://stackoverflow.com/questions/1089018/why-cant-decimal-numbers-be-represented-exactly-in-binary)

[Naf]

Here are my screen shots, i did it twice to show there are also changes to my VVE after i save...