[GreenKeys] Why this stuff matters

[Jeffrey Angus]

An interesting point was brought up.
Yes, the inductance of the selector magnet changes as the
movable plate touches the coil poles.
Could somebody send me a model 15 selector magnet so
I can measure the before and after inductance? And how much
current it takes to close the armature.
This can be modeled in LTSpice by using a normally closed
switch that will open and add some series inductance at a given
current to model the closing of the armature.

As the selector armature makes contact with the pole piece(s)
the inductance will go up. This will result in a longer time
constant.

However, this is AFTER the armature has contacted the pole
piece. The critical timing is the initial rise time of the current
BEFORE to armature makes contact. It has to do that BEFORE
the TTY machine says, "This is when I set the mechanical
levers to either Mark or Space, then wait until the next sample
time."

This where the range finder comes into play. At Center:
The initial start pulse is supposed to be 22 mS long.
The selector mechanism waits about 11 mS to determine if
the is the signal is Space. Then decides every 22 mS to decode
the subsequent Data bits at 33 mS, 55 mS, 77 mS, 99 mS and
then finally at 121 mS.

As long as the time from contact closure to armature contact
with the pole piece is 10 mS or less, the selector magnet is
going to properly set each sequential bit prior to the end of
the data stream.

If the rise time is greater that 11 mS, the machine will print
garbage, (unless you adjust the range finder to lag.) Meaning
it test for "valid" later during the sequence. For example, 16 mS,
38 mS 60 mS etc.

THIS is why the time constant of t = L/R is crucial.

Trivia: Quality of signal.
There are three levels of quality (typically called distortion.)
Copying from a source such as ITTY.
Distortion should be next to zero.
Long lines, (from previous examples) 10-27 miles.
Radio TTY, all sorts of fun things, such as selective fading.
(Missing Mark or Space signals) and delays caused by
propagation.

 ??Trivia #2: Receiving 66 WPM signals on a 60 WPM machine.
Set the range control to lead the signal.

The middle of the 5th data bit is at 115 mS instead of 121 mS.
The selector mechanism waits about 7 mS to determine if
the is the signal is Space. Then decides every 22 mS to decode
the subsequent Data bits, 29 mS, 51 mS, 73 mS, 95 mS and
then finally at 117 mS. This spreads the validation from 3 mS
early to 2 mS late. If the received signal is free of distortion it
will probably decode properly.

No, the range finder does NOT change the speed of the machine.
It just skews the validation timing to hopefully hit all 5 data bits
when they are valid. It most certainly have nothing to do with the
transmit timing. To receive 60 WPM on a 66 WPM machine, that
range finder on the 66 WPM has to be set to an appropriate
amount of lag. 2-3 mS.

Trivia #3. I have way too much free time.
Having a logic simulator program, I designed a discrete version
of a UART. The trick was adding a range finder function to it.
Starting with a 1 mS time base (1 KHz)
A divide by N counter for the first validation point. 11 mS -/-
5 mS (i.e. 6 mS to 16 mS in 1 mS steps.)
Then N = 22 for each subsequent validation time.
I was able to copy a 66 WPM data stream with a 60 WPM UART
by decreasing the initial time from 11 mS to 7 mS.
For the range finder I used a ten position thumb wheel switch.
I suppose I could have use two switches, a 10 KHz time base
and an N = 220 for a little finer tuning, but this was just an
exercise in "proof of concept." It was NOT a trivial task.

-- 
"I am a river to my people."
Jeff-1.0
WA6FWI
www.foxsmercantile.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Then a miracle occurs.gif
Type: image/gif
Size: 12871 bytes
Desc: not available
URL: <http://mailman.qth.net/pipermail/greenkeys/attachments/20190911/f041a806/attachment.gif>