Len:

Right now I am running a 3.8Ghz processor clock with a 466Mhz memory clock
on a bus capable of 266Mhz.

The fastest keyer would leave the
hardware/software killing time to await his next di or dah...

The purpose of the "error file" is to catch new abreviations so they can
be added to the table, let them come up with as many variations as they
possibly can, in the end the reader is only made stronger...

John

On Thu, 04 Aug 2005 12:01:04 -0700, LenAnderson wrote:

From: John Smith on Aug 3, 10:51 pm

Len:

Yep, that is one way alright, and produces good results, there are others,
some better.

Adaptive learning by the program is the key, and the program must learn
what the senders' length of a di to a dah is, and the breath of the width
he is spanning of each the di and the dah.

The amateur abbreviations are in a table, and the dictionary from a spell
checker can be borrowed to check decoded morse words against which are not
abbreviations.

There's no need to use a table of abbreviation...those can vary
from time to time and operator to operator. The MAJOR problem
is determining the "dit rate"...once that is done, the "dah" can
be separated, also the inter-character spacing.

You are right, a high speed machine affords you time to do
abundant error checking--and here is where you gain close to 100% accuracy
from, final fall back is the ear and the mind, to correct any mistakes the
program cannot, yet, handle...

Those who've never gotten far INTO computer programming will NOT
fully understand how blazing fast a 2 GHz clock PC really is!
My platform isn't top-of-the-line at 2 GHz clock and 100 MHz
RAM access...but it can almost blow the mind on how fast it can
handle anything in the Win32 family...I am getting slowly into
that through PowerBasic Compiler, calling the canned Win32
routines directly.

Back in '92 I shifted over to a moderate PC with a 20 MHz clock
and 1 MHz RAM access rate...and was checking various forms of
complex number calculation combinations to handle LARGE two-
dimensional arrays. Had those runing at tens of thousands of
random-quantity repetitions in order to get the fastest. That was
for a ported-over circuit analysis program from RCA in the 70s
(which I had helped improve - along with others in Central
Engineering - then). In the 1970s, any mainframe with a 10 MHz
clock and 1 MHz RAM access was considered "top of the line." :-)
I dug out the same PC time test routines and ran them on THIS
platform and the runs were just an eyeblink long. :-) In order
to actually time them, I had to increase the number of iterations
a hundred times (from about 10K) in order to see some semblance
of actually working hard!

All words which do not match the table of abbreviations or the dictionary
have a copy of that word thrown into an error file, along with di's
represented by periods and dah's represented by underscores or hyphens, of
the word thought to be an error. This error file can be studied later and
the program "tweaked" to handle such errors in the future.

I disagree. The first task is to ADAPT to the going rate. That
requires only a temporary memory (but a large one at that since
one dimension MUST be time) to set the approximate received rate.
When rate is approximated, there can be a built-in weighting on
time duration to determine dit from dah. Simple conditional yes-
no on duration but the trick seems to be arriving at a good
decision point in time.

"Abbreviations" tables aren't needed. In fact, from seeing such
a program working (and being able to look at the flow diagrams
of the routines...source was in C++ but flow diagrams were in
standard box-diamond form...the common abbreviations and Q codes
could be left as they were in ASCII on the screen. Since morse
operators tended to have a great variation in inter-character and
inter-word spacing, those spaces were just left in the screen
display and the human reader could do the final "adaptation" on
what the spaces meant (or didn't mean).

However, what interests me most is your knowledge on the subject, you most
certainly have a good grasp of the logic necessary to begin to put one
together.

In 1970 (at an RCA division in Van Nuys, CA) I got a chance to
do programmed calculations on an HP 9100 desk calculator...did
some statistics runs on aircraft collision avoidance estimates,
part of a long-range R&D project at RCA devised by the late
Jack Breckman (a genius type who could extemporaneously speak
in ordered paragraphs). Found that programming and I got along
very well and a "romance" of sorts happened, went full-flower
with successful negotiations with bean counters to get corporate
computer time (then horribly expensive). Got Dan McCracken's
softcover on FORTRAN IV Programming to explain the program
ordering (damn good book, Dan became President of the ACM for
a time later), won a steak dinner bet with another on being able
to make a running program and off it went to bigger, better
things. The epiphany happened due to a supplier delay on some
small inductors for a vital hardware delay line...could I use
a "close, but not quite right" inductor which was plentiful?
Did a simulation run for pulse shape on the corporate computer
using the possible replacement, decided it was okay. When the
replacement parts arrived, I quickly tack-soldered the delay
line together, viewed the pulse shape though it and found the
computer simulation waveform matched the real waveform EXACTLY!
I was sold and a definite believer in accurate simulation ever
since.

Perhaps you have programmed and played with such yourself? Perhaps you
have a relative or friend in the field?

I never bothered with a "morse code reader" program. Wayyyyy
TOO MANY OTHER kinds of calculations that would be of immense
value. When my group at the RCA division was disbanded in '75,
I had six programs in the Central Engineering software library
and have had four other programs as Shareware back in times
before the Internet went public. Those are all Freeware now,
not that it matters much with Windows and other GUI-ey graphic
screens being "what all want." :-)

As a former voting member of the ACM, courtesy of cross-membership
privilege of the IEEE, I've worked with/known a bunch of computer
programmers. The kind that can DO THE WORK and DEMONSTRATE it
without pointing to a bunch of framed/plaqued certificates on the
wall. One of those was the guy I described...one who had the
HOBBY of programming as well as doing it every day for a living.
We were friends enough for him to let me look at all pages of his
project notebook...and myself letting him use my Icom receiver as
a morse signal source. He thought it was a fun program to do,
while I thought morsemanship wasn't worth bothering about...but,
it was a fascinating challenge to mechanize and to make work.
His development platform was rather faster than my 20 MHz clock
thing and - as it was written then without final optimization -
wouldn't work with high rates of the speed-freak hams (I knew of
two, one in Frisco, the other near San Diego, both retired and
busy beeping each other most every day then). Right now I can
order a Microchip PIC from DigiKey running at a 50 MHz clock,
get a couple large EPROMs to hold the source code (if ported)
and it would work fine, I'm sure. PIC's RISC instruction set is
NOT compatible with 80x86 instruction set and takes a lot of
translation. Thank you, I'll take canned PIC programs, use
those and be done with it.

I do NOT have ANY sort of decoder for morse, TTY, commercial
SSB TTY tones, any of the TORs in-house. Not my cuppa either.
The 'TOR peripheral boxes are cheap enough that I could buy
one and use it if the interest struck. No problem. Someone
else did the design, debug, engineering and I respect that;
no sense in re-inventing wheels unless it's to make them more
rounded, smoother, etc. :-)

Note: Watch for Jimmie Noserve, the Nun of the Above, to pick
up on that last sentence and use it later in chiding postings
agin' me...it might be weeks before he do dat, but he gots a
memory like an effluent and will issue it later. Predictable.

NSA BSA