At 06:35 PM 4/17/2003 -0400, Bob Ammerman wrote:
>I would think a comparison between two readers using the same target
>barcodes would be meaningful, no matter what shape the barcodes were in.

IF you can assure the scan happens over the same points in the target.
Otherwise the input to the algorithm can be significantly different.

If you're testing decode algorithms, you can code up a micro to send
perfect, and variously damaged waveforms, so that it's always repeatable,
but it's a giant PITA.
Good for regression testing though.

What I've done in the past, is to output the scan widths in hex form, prior
to the decode, then add the decoded data appended as a comment.
Capture the whole thing in hyperterminal, then weed out the good scans.
Take the bad scans, include them in the next code assemble, then hand-fly
the decoder through it's logic to see if the scan should have read ok, or
where the logic failure is.  It's somewhat less of a PITA that way. Of
course you need a micro with a uart, and/or enough spare code space to do
that.

>In fact, if I were tasked with this project I would expend significant
>effort to get a good sampling of different typical barcodes to use with the
>two readers.

I have a little book of nasty barcodes from real products.  Given the HUGE
penalties that large retailers put on bad barcodes, it's amazing how many
are out there in the real world.

>With a large enough sample size you could certainly determine which reader
>would be better in the real word.

As long as you're careful about what sorts of errors you allow.
Some are out of spec, and could bias the results unfairly.

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