Dyno-Less tuning?

[Walter Petermann]

Thor Johnson wrote:
> 
.......
> I have some nice long data logs that look like (like hondo?):
>   CamDT, Throt, PSI, Tp1, Tp2, Tp3, Tp4, Ox1, Ox2, Ox3, Ox4
> 
> I've loaded these into excel to assist me... does anyone have a
> "tuning your ecm with excel" faq?
> I have lots of points, but I have no idea whats going on....
> they make pretty lines...
> 
> Currently, I'm leaning the whole map out until the engine hardly
> runs, finding that MAP value, twiddling with it (slowly) until I
> get Ego's in the range I want (which changes the RPM!), doing the
> same to the new RPM... I can't quite get it to converge.  To get
> the upper PSI points, I have just *completely made up numbers*
> that attempt to elimanate idle->WOT stumble.
> It doesn't help 1/2throt->WOT very much.
> 
.....
> 
> As you can see, I'm confused when we try to get these things to
> work in reality instead of theory...
> 
> TIA,
> Thor Johnson

Thor, you might find some use for this identification
technique (learned it in a controls course but never made
use of it):

let x=map output, y=RPM , z=throttle position and T=injector
pulse length.

then:  ax + by + cz = T

take several measurements,

  ax1 + by1 + cz1 = T1
  ax2 + by2 + cz2 = T2
  ax3 + by3 + cz3 = T3

set up a matrix equation,

  | x1  y1  z1 |   |a|     |T1|
  | x2  y2  z2 | * |b|  =  |T2|
  | x3  y3  z3 |   |c|     |T3|

now all you need to do is take the inverse of the data
matrix and
multiply both sides,

                       -1
 |a|     | x1  y1  z1 |    |T1|
 |b|  =  | x2  y2  z2 |  * |T2|
 |c|     | x3  y3  z3 |    |T3| 

and you have the solution for the coefficients. In theory
the minimum
set of equations must equal the number of variables/sensors.
In practice you should have at least 3 times that many.
However, your data matrix will no longer be square and the
inverse will be more complicated to calculate (pseudo
inverse).
If you are interested I can dig it up...

 regards,
   Walter

PS I have no idea how this would be done on a spreadsheet.