Clay, as to "why 20 ohms out", there is a long-time-nominal 50 ohm PPS
convention that calls for 5V pulses to be delivered into a 50 ohm load.

If the driving voltage was 5V and source resistance was 50 ohms, then you'd
never get more than 2.5V into the load.

Different references across the net show 5V chips driving parallel 5x100
ohms, or 5x47 ohms. These are all attempts at getting more than 2.5V and
closer to 5V into the load. If you run the driver chip at 6V (which is
actually permissible for some HC-derived logic chips) then you can get even
closer to 5V into a 50 ohm load.

Most devices that consume this 5V-into-50-ohm signal are actually fine with
less than 2.5V but that's very vague.

There are real coaxial cable driver chips as well as arrangements of
bipolar transistors that do a very good driving long 50 ohm coax cables to
5V. These were very popular in the 1960's and 1970's as coax drivers for
mainframe/minicomputer computer peripherals, and are still available today
but not nearly so popular.

A much more popular cable driver standard from the past 20 years would be
LVPECL or LVDS but those are nominally differential and everyone seems to
insist on coax even today.

Tim N3QE

On Sat, Jun 17, 2017 at 4:15 AM, Clay Autery cautery@montac.com wrote:

HI

If the “output stages” are all going to switch at the same time, you want to hit their inputs
with a fast edge. As a practical point in a stand alone buffer, the invert and then invert
process keeps you out of a classic “oops” mistake.

Not everybody does :)

It all depends on what you are trying to accomplish. If the spec from the OEM you are
building it for wants a 5V level into 50 ohms, you use a bit smaller resistor … You need
to check the spec from your customer.

Gates are not spec’d for output impedance. The output stages are MOSFET’s and they
go into current limit during switching. In the full on or full off state, they are close to a
short circuit. A good guess is that they are in the 20 to 50 ohm range. For more detail you
would have to grab a network analyzer and pick a frequency range.

The AC04 gates are not as high output drive as some of the others. There are gates
designed as high(er) current buffers that likely will do better. Usually they will supply
1.5X or 2X the current of a normal gate.

Yes, but power is the primary issue. Will you be running into a short? (again, back
to the spec requirement). Do you need to run into a short and supply 5V (or 4V or >2.5V)
into a 50 ohm load? Current (and thus power) can get pretty big pretty quick.

========

The much more basic set of questions revolve around what I’m calling “the spec” above.

If you need to drive 5V CMOS logic on the other end, that gives you one requirement. If
you need to drive 5V TTL on the other end, the requirement is much different. The same is
true if you have 3.3V logic of different families on the other end. There is no “one size fits all”
here. 5V logic is not a favorite anymore. The easy answer on the Rx end is to run TTL level
5V stuff if you must have 5 volts. Then it can be driven with a 3.3V source. The only “works
with everything always” solution is to drive 0 to 5V into a 50 ohm load from a 50 ohm source.
Indeed, even that does not work with everything. The 10V p-p signal into an open circuit will
nuke a 5V gate.

Do you need termination on both ends of the cable? We have gone around on this a lot of
times. Terminating one end or the other is generally adequate. Terminating both ends does work
better. Terminating neither end is never recommended, but it can work out. If you need terminations,
do they have to be DC terminations? …hmmm….

What range of “PPS” signals will the buffer be handling? Is it’s entire life going to be spent running
a 10 us wide pulse? Might that pulse get inverted somehow? Will you ever be running a 50/50 duty
cycle signal (or 50/50 pulse per every other second …). All of this drives the numbers on the power
in your driver.

Will you have multiple PPS outputs and are they all in sync? A great big high power driver is really
neat. It also is a great way to generate a massive spike on ground and supply. Isolating multiple
drivers to reduce cross talk is a bit exciting when the signal goes down to DC and up to GHz. LVDS
gets fast signals moved without a lot of power or a lot of swing.

What will you be driving with the PPS signal? The answer of “I don’t know” is not at all uncommon.
Stop and think about what actually uses a PPS around your lab. Counters can be adjusted to any
level and any termination. PPS inputs on digital gizmos are most likely TTL level. What else will you
be feeding?

How far / where will your PPS signals be traveling? Driving a 5 or 10 foot cable is fine. Sending a signal to
the next building takes a very different approach. Output isolation is generally ignored in a lab setup.
It may be a really big deal if high voltage or high power gear is involved.

No, it’s not simple. No there is no single right way to do it.

Bob

Why 100R?  If using a 74xxx logic part as the driver is has a maximum
drive current.  100R limits current in case of a short to ground or 5V
supply  to only 50mA  It is about the lowest value resister i'd want to use

If you need 50 ohm output them go with an small RF amplifier or a
transistors driver.  These can drive a 50R load without burning up.

On Sat, Jun 17, 2017 at 1:15 AM, Clay Autery cautery@montac.com wrote:

--

Chris Albertson
Redondo Beach, California