Fanuc Karel REAL variable rounds up. How to avoid?

Hi

We are doing accurate measurements with robot and probe. We are passing robot positions through position register to Karel program. Problems appear with big values and they cause inaccuracy to the end results. In Karel programs values are passed as REAL values.

For example:

We want to calculate next expression from position values:

val = - (x_val ^2 + y_val^2 + z_val^2)

When measured values are

x_val: 1554,669 mm

y_val: -468,354 mm

z_val: 9,543 mm

We get from Karel program result: -2 636 442,000

From excel and wanted value is: -2 636 442,238

Is there any way to avoid the roundups? We know that the Karel can handle limited size values, but still we those digits shouldn’t be too much?

(How we see it, our math should be correct in the program )

Thanks

Aleksi