[time-nuts] Why are 1PPS signals so skinny?

[SAIDJACK at aol.com]

Yes, but the point is to not use end-termination for all the reasons  
mentioned by others in this thread, such as massive spike in power consumption  
once per second, over-voltage spikes if the termination is faulty or  
missing, higher ADEV due to power supply modulation, etc etc..
 
Your test clearly shows the ringing when the transmission line is left  
open, and it shows the massive current (5V into 50 Ohms) when end-terminating  
the cable. It also shows that the cable is ringing up to 7V or more! Which 
could  actually kill your driver circuit by overvoltage if you forget to 
enable the 50  Ohms termination on your counter or scope for example. The 
voltage could  theoretically spike all the way up to 10V as long as the pulse  is 
traveling back on the coax. All more reasons why this is an  undesirable 
mode of operation.
 
To make this work without the unnecessary power consumption simply remove  
the end-termination resistor, and use it as the series termination resistor 
(R1  in your schematic)! Done.
 
Attached are two plots of a series terminated (~55 Ohms) high-speed  1PPS 
transmission from our CSAC GPSDO board zoomed-in  and zoomed-out to show the 
actual rise-time, and a longer time frame  view.
 
The 1PPS pulse was run through about 30 feet of LMR-195 cable, directly  
connected to the CSAC GPSDO 1PPS CMOS 5V output. There is no massive voltage  
over-shoot, the output is short-circuit protected, and no matching resistor 
is  required, just a 50 Ohms coax cable. Use an additional 25 Ohm series 
resistance  for 75 Ohms cables.
 
The output rise time is 1.25ns at the end of the 30 foot cable, and the  
signal fully settles within 800ns, and never goes below 4V after the initial  
1.25ns rise.
 
The current spike on the power supply is only there during the time that  
the cable is being charged up, which is about 30 feet * 2 * 1.5ns/foot = 
~100ns.  That is short enough for the power supply caps to filter the current  
spike.
 
In short, one could easily modify the Thunderbolt 1PPS output circuit which 
 is probably a bunch of parallel AC240 gates with some low value series  
resistors, and modify these resistors to have the equivalent  of 50 Ohms 
impedance. That would alleviate the need for end-termination on  the coax, and 
provide very clean rise time, fall time, and no ringing.
 
BTW: one advantage of this in the lab is that you can connect multiple  
instruments to one 1PPS output. The signal will take slightly longer to settle  
as it has to traverse and charge more cable hubs, but in the end there will 
be  5V on the cable with no DC current flowing, and there won't be any 
positive  ringing above 5V.
 
You cannot drive more than one input if you are using 50 Ohm  
end-termination without possibly over-loading the driver, and causing  massive impedance 
mismatch, and getting the associated cable ringing etc.
 
bye,
Said
 
 
In a message dated 5/15/2012 11:49:14 Pacific Daylight Time,  
shalimr9 at gmail.com writes:

The  Thunderbolt's output impedance is much less than 10 ohms. However, it 
is only  necessary to filter the end of the line for a clean pulse.

See  http://www.ko4bb.com/Test_Equipment/CoaxCableMatching.php

I used the  Thunderbolt's PPS output as a source in those  measurements.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: CSAC_risetime.GIF
Type: image/gif
Size: 9261 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20120515/a58e3dbe/attachment.gif>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CSAC_risetime_zoom.GIF
Type: image/gif
Size: 10836 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20120515/a58e3dbe/attachment-0001.gif>