Page 219 - Embedded Microprocessor Systems Real World Design

real-time capacity of the 8031. The tables were calculated by software running on
a PC, and the resulting tables were incorporated into the 8031 code.
The spacing errors occurred on every machine produced; it was not limited only
to certain machines. There were several possibilities for why spacing errors still
occurred after the original spacing/throughput problem was solved:
The original fix had a bug (software problem).
The 8031 was making an incorrect prediction about the size of the post-
acceleration spacing, and therefore not correcting some documents (software
problem).
The 8031 was making an incorrect calculation about the amount of correction
needed to fix the spacing (software problem).
The sensor hardware delivered an incorrect signal to the 8031, resulting in an
incorrect gap measurement (electrical hardware problem).
The motor or motor driver was not responding correctly to the rampdown/
rampup commands (electrical or mechanical hardware problem).
The first acceleration wheel was not always accelerating the documents correctly
(mechanical hardware problem).
The problem only occurred on documents that were physically damaged in some
way and could not be controlled/sensed properly (media problem; possible spec-
ification problem if these were not taken into account in the design).
The first items checked were the media and the acceleration wheels. We found
that most of the documents with incorrect spacing were not damaged so it was not
a media problem. The acceleration wheel was more difficult to measure, but it was
based on an earlier design and appeared to be working correctly. Had the problem
not turned out to be elsewhere, we would have returned to this and examined it
in more detail.
We measured a number of rampup and rampdown cycles and determined that
the feed motor was working correctly, at least as far as the shaft encoder was con-
cerned. Spacing measurements were made on a number of documents and it was
determined that the variation in the output was minimal and insufficient to cause
the problem.
With all the possible errors other than code eliminated, a means was needed to
determine what was happening. Unfortunately, errors were not detected (at the
second sensor) until 150ms after the spacing corrections had been made at the
feeder. We did not have an emulator with a trace buffer that deep. This type of
issue is worse with modern processors, in fact, because newer, faster processors will
execute more instructions in a given time period, requiring an even deeper trace
buffer for problems of this type.
Considerable work was done to verify that the original software fix was correct,
verifymg the thresholds that determined the feedback characteristics. Eventually,
as described in Chapter 6, action codes were added to the firmware to indicate mea-

200 Embedded Microprocasor Systems