The calculation in TM1 always takes the rounded 4 decimal places, regardless of their format.
The ROUNDP function can be used only for a specified cell.
How could we set the default on the calculation to incorporate higher # of decimal places ?
If you are trying to say that calculations only use 4dp, then you are incorrect. Calculations in TM1 will always use all available decimal places (which, as you have noted, you can control with functions like RoundP).
However, the Trace Calculation window will only display 4dp. As far as I know there isn't a way to change this.
You can prove this to yourself (taking a guess at your calculation) by multiplying 5374368948.2000 by 0.0385 and you will see that you do not get your calculated value of 206945450.7194.
Now multiply it by 0.038506 instead, and you will get that value - to 4dp.
Just a small point of note - the floating point issue can arise if you're dealing with some very large numbers.
eg, we generate our costing / pricing rates against a VERY large number and proportionally spread over a large number of areas.
Because the spread %ages are calc'd, they tend to give 0.1x10E-25 type numbers. Multiply one by the other and TM1 gives up eventually (working on the basis that there's no point in giving any further detail as >15dp* seems a reasonable level of granularity to use in the multiplication). Multiply out a huge number as we do for our total costbase and you find 20k drops through the net due to what is effectively system rounding.
We get round it by multiplying out the % by 1000000 and then dividing the answer back down again after the multiplication has been done to effecitvely move the decimal place further to the right and avoid the floating point issue. System knows what the number is, it in essence just rounds it internally before the calc.
If anyone's got a better way of changing the FP (maybe some arcane, undocumented config setting?) then I'd love to hear it, would get rid of some unrequired calcs in the system.
* Point of note - no idea if it is 15dp that it stops working, that's just a high lvl observation but think it's somewhere around that level of granularity.
However, there might be a workaround (depending on what you want to achieve). You can create additional rule-calculated trace measures that would be computed by multiplying your original measure by 1000, 1000000 etc. Then you can run your trace on those new measures - the precision will stay at 4 decimal places, but those places have moved to the right, so you've increased the effective display precision.
