Seems like you're saying that you went with what I listed as option (2). =
=20
Which I have started doing as well.  I'm also doing an interpolated looku=
p=20
like you.  Read 16 samples of 10 bits, continuously summing them (so I ha=
ve=20
16 bits now), then lookup the table (256 values) for the upper 8 and the=20
upper 8 + 1.  Then interpolated proportionally based on the lower 8.  I'v=
e=20
found that only 3 or 4 bits on the lower is enough, but overkill is fine.

I'm using a 16F872 though, and have room for 512 lookup values, so consid=
ering=20
that.

Cheers,
-Neil.



On Thursday 31 July 2003 01:56, Bob Axtell scribbled:
> I know what you are getting at. I had a similar situation.
>
> In my case, I was extracting 24bit strain-guage values and calculating =
the
> engineering
> value (weight in kg). The math was bewildering and took too much time. =
The
> strain-guage and the PIC / A/D involved formed a UNIT as far as the HOS=
T
> was concerned.
> The problem was that I had errors from so many different sources that t=
he
> math was Min/Max (the possible solutions, rather than an ACTUAL solutio=
n).
> Here's what solved everything:
>
> A test fixture was fabricated that subjected our SENSOR/PIC unit under
> known pressures at fixed points. For example, a value at 00.1Kg was app=
lied
> and the PIC measured results captured (from the PIC) and stored on a lo=
cal
> HOST computer. For each value from 00.1kg to 99.9kg the values were
> extracted automatically (using a hydraulic press to ensure constant for=
ce).
> Then the PIC was loaded with a 1000-place PGM MEM table, using almost 1=
K of
> program memory space. The 24b value was shaved so that the LS 10 bits w=
ere
> ignored except that when the table was built, were round up or down to =
the
> next useable bit. Loading this table became a permanent part of that UN=
IT,
> and it took a few minutes to upload and FLASH in the values. But it onl=
ty
> had to be done ONCE.
>
> When a value was received, the lookup routine rummaged through the list
> until it discovered what UPPER and LOWER values the value received. The=
re
> were two versions of this software, i.e. a 4-digit version (99.99kg) an=
d a
> heavier 6-digit version for larger strain-guages (9999.99kg). To obtain=
 the
> last digit value between the upper and lower points, we took the LS 10b=
its
> and cranked on them to interpolate the last digit. In that way, only a
> 1000-place table was needed so it easily fit into a 2K PIC16F876.
>
> Sounds complicated, but is the easiest way home because it compensates =
for
> EVERYTHING in the UNIT: PIC, resistors, strain-guage, 24b A/D converter=
,
> local 5V supply, and bad teeth.  It worked VERY well.
>
> --Bob

--
http://www.piclist.com hint: PICList Posts must start with ONE topic:
[PIC]:,[SX]:,[AVR]: ->uP ONLY! [EE]:,[OT]: ->Other [BUY]:,[AD]: ->Ads