Is the loss of precision that important?

You may be able to get a little more precision by first dividing by 64, then
multiplying by the 0-45 advance figure and then dividing by 16.

Any chance of putting on a web site when it's finished?

Regards

Mike Rigby-Jones

> -----Original Message-----
> From: Mark Crankshaw [SMTP:mic1000@HERMES.CAM.AC.UK]
> Sent: Thursday, May 20, 1999 1:01 PM
> To:   PICLIST@MITVMA.MIT.EDU
> Subject:      Slightly [OT] Timing calculations
>
> As part of an engine ignition project I need to calculate an advance value
> using a lookup and a measured RPM. Seems simple enough but I've come
> across a small problem.
>
> The time for 180 degrees of crank rotation is measured giving a 16 bit
> number which is used in a lookup to give an 8 bit advance number
> corresponding to a range of 0-45 degrees of crank rotation. This number
> then must be converted into time, based on the RPM to feed into a delay.
>
> The numbers look friendly - dividing the 180 degree time by 4 gives the
> time for 45 degrees, dividing that by 256 then gives the time for an
> advance number of 1 so then just multiply by the advance number. Since the
> division is by 1024 overall, I intended to just shift the bits right,
> saving a slow division routine. The problem with the 'divide by 1024 and
> multiply by advance value' process is that a lot of the precision is lost
> if I divide by 1024 first. Multiplying by the advance number first would
> require 24 bit maths routines.
>
> Anybody got any cunning ideas? The calculation must obviously be quite
> quick
>
> Thanks,
>
> Mark.