* I232 Communication Components for RS232 *
* Electrical Isolation, Fast Data Transfer, and Long Cables. *
* Photographs of CAT-5E Cable Waveforms at Destination for *
* 150 and 1000 feet With Different Drivers. *
* As applied to RS232 applications. *
For any practical directly connected RS232 system the combination of
the RS232 line driver and the communication line can be approximated
as a lumped resistance-capacitance low pass filter. Any transmission line
reflected energy effect is negligible. This low pass filter has some
bandwith that determines the maximum baud rate. There is nothing that
would cause a resonance effect at 9600 baud.
Because the line driver is of a fixed source impedance the bandwidth
will be inversely proportional to cable length. Total cable capactance
is capactance/foot times cable length. Thus, maximum baud rate is
inversely proportional to cable length.
If you stay adequately within the bandwidth limits of the communication
link (driver, cable, and receiver), then the only major problems you
have are related to noise. This noise can be common mode ground path
currents, capacitively and magnetically coupled signals, and radiated energy.
The most likely major source of noise is in the common ground path.
The following photographs show the waveforms at the destination (the receiving end)
of a CAT-5E cable.
The tests were run with two different cable lengths, 150 feet of
an import of unknown manufacturer, and 1000 feet of Belden 1583A. These appear to
have slightly different propogation velocities, an observation, but this is of no importance.
Three different drivers were used. These all had a source voltage of a rectangular pulse of
approximately +/-12 v open circuit.
They differed only in source resistance.
The first source was a Tektronix CFG250 Function
Generator with only its internal source resistance of about 50 ohms.
The second source
was the same except that a 250 ohm resistor was inserted in series with the output.
third source was the same generator driving the input of an MC1488 (circa 1978).
The 1488 is a standard RS232 driver
with a 300 ohm output resistor preceeded
by a +/-10 ma current limiter supplied with +/-12 vdc. Thus, the output is an approximately constant current source
until the 300 ohm resistor becomes dominant. This causes the leading edge to be more linear than
for a straight exponential RC curve. The initial current thru a 300 ohm resistor
charging a capacitor from a 24 v source would be 80 ma. This is 8 times the current
limit and thus the reason for the slow rate of rise when the 1488 is the source. The initial slope of the
curve is 1/8 of what it would be if the current limit was not present. Compare photo P7 with P9.
Any of these combinations will probably work at 9600 baud. This does not mean that they should be
used. There are no resonance effects except at the leading edge of the pulse. For the 150
foot cable this dies out within 1.6 microseconds. For the 1000 foot cable it dies out in
8 microseconds. After this transient period at the leading edge the signal is steady state.
A typical RS232 receiver has an input impedance of 3000 to 7000 ohms. This would have little
effect in comparison with our 10 megohm scope input. Our goal here was to create a worst
case condition which is represented by an open circuit transmission line.
The tests were run with unshielded CAT-5E cable because this is low capacitance cable with
about a 100 ohm characteristic impedance. The capacitance is about 13.6 pfd/foot. Belden
8723 which contains two twisted pairs each individually shielded is about 66 pfd/foot. Roughly
speaking an 8723 cable will fail to work at about 20% of the length of CAT-5E cable with other parameters
Within useable cable lengths there is nothing in the characteristics of normal RS232 drivers,
receivers, and cable to cause any resonance at 9600 baud.
Most CNC RS232 communication will be asynchronous. This means that the UART (Universal
Asynchronous Receiver Transmitter) has an internal precision clock programmed for the baud
rate of the data being transmitted and received. Today virtually all communication within a
given path is at
the same rate for transmission and reception, and the same number of data and stop bits,
and parity in each direction. In the past send and received might have been at
at different baud rates. Parity must always match in one direction. Transmit stop bits can
any number greater than or equal to the setting for the receiver. There is always exactly
one start bit, and at least one stop bit.
Bit, byte, and word need to be defined. A bit is a single binary digit that can have one of
two states, 0 or 1. A binary word is a group of binary digits. Some examples are 4 bit, 8 bit,
16 bit, 32 bit, etc. A one bit word can represent 2 items, a 4 bit 16 items, 8 bit 256 items, etc.
Generally a 4 bit word is called a nibble, an 8 bit word a byte, and after this usually
16 bit word, 32 word, etc. However, some of these may be referred to as double word. But that
useage becomes specific to a particular computer or language. So double word might mean 32 or
64 bits, etc. Sometimes byte is associated with character, and sometimes characters are
represented by less than 8 bits. In fact CNC machines are usually communicated to via RS232
with 7 bits per character. 7 bits allows 128 entities.
The UART looks for the leading edge of a start bit. This is
identified by a change from binary 1 (the rest state) to binary 0. An internal counter is
started and counts to the midpoint of the start bit and checks that it is still binary 0.
If so, then the remainder of the incoming data word is processed with testing of each bit
position in the middle of its time period. At the end of the word there is a stop bit which is
binary 1 and therefore equal to the rest state. Now the UART waits for the next start bit.
Thus, the UART resynchronizes on each received word. If you have 7 data, 1 parity, 1 start,
and 1 stop bits, then the UART word is 10 bits long. What distortion exists on the leading edge of a bit
position does not matter so long as it does not continue to the center of the bit position, or
cause sampling of bit centers to be incorrectly located.
Some peculiar distortion on the leading edge of the start bit might be of concern, but the
following waveforms do not illustriate such a problem.
P1 --- Photo 1608A. Shown are superimposed plots with scaling of
200 nanoseconds/cm, and 5 volts/cm.
This is a test using a source pulse
generator with an output impedance of about 50 ohms. The open circuit source
voltage is +/-12 volts. The cable is 150 feet of CAT-5E from an import manufacturer.
The destination (receiving end) is open circuit with only a scope probe load
of 10 megohms. Open circuit is a worst case condition for producing
The first plot is the input to the CAT-5E and shows distortion from the
reflected energy from the open circuit destination. The second plot shows
the destination voltage. The transit time to the destination is about 200 nanoseconds.
The total time of waveform distortion from the reflected energy is about
1.6 microseconds. Beyond this the signal is in a steady-state condition for
any length pulse.
Thus, a pulse width of 3.2 microseconds
would be approximately the shortest one would want to use in communication.
That is about 300,000 baud. But you should not work that close to your
The RS232 maximum threshold points are +/- 3 volts. Thus, there is a good
margin between the signal and threshold points.
P2 --- Photo 1607A. This is the same as P1 except that only the
destination plot is shown. The +/-3 volt threshold points are marked on
this photo only. You can judge their position on the other plots.
P3 --- Photo 1610A. This is the same as P2 except that here the
cable is 1000 feet of Belden 1583A.
Now the waveform distortion lasts about 8 microseconds. Using the same
criteria as in P1, then maximum baud rate would be 62,000 baud. Note very
little timing error at +3 volts.
P4 --- Photo 1612A. This is the same as P2 except that the
sweep time has been changed from 200 nanosec/cm to 20 microsec/cm, and
markers have been inserted to indicated the sampling points at
9600 baud. Note that the sampling times are in a perfectly steady-state
portion of the waveform. The pulse width has been set to equal one bit period
at 9600 baud.
P5 --- Photo 1611A. This is the same setup as P4 except cable length is
1000 feet instead of 150 feet.
P6 --- Photo 1613A. This is the same as P4 except the source impedance
has been increased from 50 to 250 ohms. Note all overshoot is gone and
there is very little effect from the reflected energy.
P7 --- Photo 1614A. This is the same as P6 except that here the
cable is 1000 feet of Belden 1583A.
P8 --- Photo 1617A. This is the same as P4 except that a standard
RS232 line driver, Motorola MC1488, is used to supply the input pulse.
This chip has an internal output resistance of 300 ohms preceeded by
a +/-10 ma current limiter.
This would probably work up to about 50,000 baud rate. There is
no overshoot and little effect from reflected energy. A comparison of
P6 and P8 shows a slower rise time for P8 resulting from the +/- 10 ma
P9 --- Photo 1619A. This is the same as P8 except for the cable
which is 1000 feet of Belden 1583A.
Maximum baud rate here is probably about 9600 or less for direct RS-232 because of the
slow rise time.
PO Box 1288, Ann Arbor, MIchigan 48106-1288
This web site is http://www.beta-a2.com
Our e-mail address is firstname.lastname@example.org
Copyright© 2005-2014 Gordon A. Roberts
All rights reserved. 141201-2127