AVR-Chat

At 04:19 AM 7/16/04 +0000, you wrote:
>Excel shows 65,536 lines and reports that 32,000 points is the limit
>for a single column for a chart.
>
>It takes about 3 seconds to get a graph of the 32,000 points.
>
>With a small amount of mathematical reduction, that offers a
>highlight  of the points of interest.
>
>A single point reading once a minute is 10,080 readings per week.
>That means that a reading once each 10 seconds would be 60,480 points
>and would fit in one spreadsheet column.
>
>I did a little more on Excel.  12 columns of 24,000 readings per
>column and it takes almost a minute to render the graph.
<snip>

>This means the brute force method of 'just log everything all the
>time' could be done, but with something on the order of 130,000
>readings per week including a time stamp. All the points can be 3
>digit integers including the time.

OK, that sounds quite doable.  That was my big worry.


>I'm thinking that means the brute force reading will eclipse the
>limit of a ram or eeprom.

I doubt space will be a problem, with even sampling (the just log 
everything approach) you can get rid of the time column and you are just 
under 128K which would fit into 4 of Ramtron's biggest FRAMs.  Not a lot of 
room for extra but some and 4 so-8's won't take much room.  That might be a 
little pricey but I haven't checked that.

Alternatively there are flash packages of serial memory that are more than 
big enough The downside is they have the same complexity issues of 
EE.  Both Atmel and ST have such devices.  For example ST's M25P series a 4 
Mbit (512K) so-8 is about 3.25 at Digikey in singles.  They go up to 32Mbit 
(4M) in a single package. The downside is that write have to occur in pages 
so you would have to buffer the writes. ST also has the M45PE series which 
has fewer write restrictions (only one family member so far 8Mbits (1M)), 
but I didn't see the price on that.

Since you can handle the data I'd say go for it, sample everything at the 
same rate using a low pass filter on the light (just so they don't get 
aliased with any transient shadows)  and let the user do any further 
decimation and filtering with the acquired data.  They'll learn more about 
data handling that way than they would repeating the measurements because 
they set the sampling too slow.

Uniform sampling and a slightly more complex storage mechanism is probably 
easier than multiple sampling rates and a simple storage technique.  It 
should be easy to reduce the data on a PC.

That just leaves, finding time to learn how to do it :)

Good luck, this sounds like an interesting project.  If you want any 
feedback or suggestions while you are doing it let me know.

Robert

" 'Freedom' has no meaning of itself.  There are always restrictions,
be they legal, genetic, or physical.  If you don't believe me, try to
chew a radio signal. "

                         Kelvin Throop, III