OS/2 Professional/OS2PRO194.ISO/os2/wps/utils/sample/sample.txt

Text File | 1993-09-17 | 8KB | 182 lines

Documentation for sample.exe, a program
written by Helge Hafting.

There are 2 executable files:
sample_n.exe - with norwegian texts.
sample_e.exe - with english texts.
Install the one you want to use and rename it to sample.exe

sample.exe is a program for high-speed 8-bit sampling under
os/2.  The program turn the pc into a simple & cheap logic
analyzer, maybe useable for hobby electronics and such.
Sampling rates of approximately 700kB/s has been achieved
on a 25MHz 386sx, with ISA bus.

Sampled data may be viewed as 8 bits that take the values
0 and 1, or as a continous graph.  The latter is useful if the
data is sampled from a DA converter.

The program can be set up for sampling from any I/O address,
so it will work with any i/o card that provides input at an
i/o address.  Even a standard parallel port (printer port) may
be used.


USING THE PROGRAM

The program has a PM interface, and should therefore be
almost self-explanatory.  Experiment a bit, and you'll learn it.
Numbers (such as the sample buffer size) are converted by the
c function strtoul().  They can be decimal without a
leading 0: 1024, hex with the 0x prefix: 0x800 or
octal with a leading 0: 02000.

The data is displayed in two windows.  One is a digital
representation showing each bit.  The other is an analog
representation, assuming that you are using a DA converter.
(Be sure that you connect the LSB of the DA to the LSB of the
digital port, or the data will look very strange.)
Both windows are scrolled with a scroll bar below them.
There is also a slider between the windows.  It is used to drag
a 1-sample wide line across the data windows, making it easier
to see exactly where data changes.  It is also possible to zoom
in on data, using the zoom menu.

The program has three ways of searching through sampled data:
"x=a"  This is an exact match.  Type in the byte value you
want to find, and start the search.
"a<=x<=b" Type in upper and lower limits, and start the search.
This is useful for analog signals sampled with a DA converter.
"01*" This is useful if you are looking for situations where
some bits are 0, some are 1, and some are don't care conditions.
Use the radio button array and specify 1, 0 or * (don't care)
for each bit.  Pressing the search button again will bring
you to the next sample that matches the search criteria.

Interactive help and file operations are not implemented yet.

For a quick test, use configure to set the sample buffer to
100kB and the i/o port to 64.  Then press ok in the configure
dialog, and press the sample button.


TECHNICAL INFO:

Sampling rate:
The program use the rep insb instruction for sampling.  A 386
uses 6 clocks per sample, a 486 uses 8 clocks per sample.
A 386 may therefore sample faster than a 486.  The 486 is
optimized for other things than I/O.  You will usually get
sampling rates lower than cpuclock/6, as the cpu isn't the only
limit.  An ISA bus run at only 8MHz, and has several waitstates
per 8-bit i/o address access.  I haven't tested EISA,MCA or
local bus sampling.  RAM speed and possibly waitstates inserted
by the i/o card will also limit sampling speed.
If you want to find out what sampling rate you get with
your particular machine, try sampling a signal (preferably a
square wave) with a known frequency.  100kHz is a good test
frequency.
You may loose a few samples now and then because of ram refresh.
You will also loose 1-2 samples per 4k block, when the MMU
has to read a new page table entry.


Os/2 considerations:
The program achieves its high sampling rate by using all
the cpu time when sampling.  You will see that the mouse stop
responding a few seconds if you sample several megabytes.
The program reserves the cpu by turning off interrupts during
sampling, this is a somewhat nasty thing to do.  There will be
no multitasking or running of background programs when sampling
is going on.  Even device drivers will stop.
Processor exceptions and non-maskable interrupts can still
occur.  This is a problem, because os/2 will re-enable
interrupts and resume multitasking if it gets an NMI or
exception.  This will give strange sampled data, as the machine
will sample at full speed for 1 timeslice (usually 30ms), then
do something else for a while, then sample at full speed another
time slice, and so on.
NMI's don't usually happen, unless your machine is equipped
with a watchdog timer.  The only processor exception you may get
is page faults.  The sample buffer must be allocated in real memory, virtual memory on disk won't work.  This is because:
1. Os/2 re-enables interrupts after a page fault.
2. The time needed to get a page from disk is long, tens of
   thousands of samples will be lost.
You will know that a page fault happened if you get disk
accesses during sampling.  If that happens, try again with a
smaller buffer.  Allocating 3MB on an 8MB machine seems to work
fine.

The scroll bar will probably not work properly with buffers
larger than 14MB, due to PM limitations.


PLANNED FUTURE ENHANCEMENTS:
(not in order of priority.)

1. Enabling all the menu choices etc. that currently don't
   do anything, such as help and a possibility to save
   and load data.

2. Optional slower sampling.  Many uses don't require 700kB/s.
   If you want to take a look at the serial port (in order to
   find out why a modem connection don't work), then a lower
   rate could be useful.

3. Changing colors and resizing the window.

4. Support for setting up i/o cards:  Some i/o cards may require
   that certain values are written to certain i/o ports or
   memory addresses before sampling can be done.  (In order to
   set up a port as an input, for example.)

5. 16-bit and 32-bit sampling.  32-bit won't be very useful
   on an ISA bus.  16-bit sampling may be faster than 8-bit
   on an ISA bus, as fewer waitstates are specified for 16-bit
   i/o...

6. Automatic trigging.  For now, you have to use the sample
   button, the menu or ALT S in order to start sampling.

7. Fix known bugs, such as:
   a: the window don't look good in 1024x768 resolution
   b: The slider may get wrong size when zooming, if you are
      using better than 640x480 resolution.
   c: Moving the slider around sometimes leaves a black line
      that shouldn't be there.
   d: english texts are sometimes a little too long.

None of these enhancements are difficult to implement.  But my
time is limited, so I will only implement what people want.
Feel free to mail me and suggest improvements.
---------------

This is a shareware program.  You may freely use and copy it
as much as you want.  Please include this text with any copies.

Commercial users will have to pay me $10 in order to use the program commercially in any way.  Use the snail mail
address when sending money.

The program will be updated according to what people want. So
mail me if you have suggestions, questions, or bug reports,
or find the program useful.

My email addresses are:
[e-mail]
[e-mail]
I won't be available on the internet forever.  If the above
addresses don't work, try my snail mail address:

---------------------
|   Helge Hafting   |
|   Poverudvn. 22   |
|   N-3440 R¢yken   |
|   NORWAY          |
---------------------

(The "¢" in "R¢yken" is supposed to be an "o" with
 a slash through it.  If this is a problem, use an
 ordinary "o" instead.)