[time-nuts] Correcting jitter on the 1 PPS signalfrom a GPS receiver.

Mon Sep 15 14:14:13 EDT 2014

Hi Dave,

what you are describing is reflected wave switching, which works perfectly  
in applications such as the PCI bus. The PCI bus uses it because it  lends 
itself to extremely low power consumption.

Your scenario does not work with end-termination either if you have  
multiple taps, because the taps would have to a) have very  high impedance so as 
not to disturb the signal traveling down the pipe, and  b) not have any cable 
length associated with them so as to prevent the edge  going down into the 
stub and reflecting back from it, and the stub causing a 25  Ohms impedance 
drop (due to two 50 Ohms transmission lines in parallel).

With the rise-times of sub 1ns (the CSAC GPSDO has a rise and fall time of  
<500ps typically) any cable length becomes an issue, including the couple 
of  inches of wiring inside the counter and through the T-splitter etc. 
Another  issue is that the timing of the instruments would be all off because of 
the  differing propagation delays through the additional cable.

So in the end-terminated world having a stub would only work if the stub  
does not have any cable length associated with it and very high impedance. 
How  many folks distribute their 1PPS signal through a single coax cable to 
multiple  instruments? This should be done through a 1PPS distribution amp.

I agree though that in the analog video world you don't want reflections  
due to even small impedance mismatches because they cause ghosting. In the  
digital world where all we care about is that one single rising edge per  
second ghosting is not an issue.

My original point was that the video world got it right: use a 75 Ohms  
output impedance for a 75 Ohms coax. The 1PPS world did not get it right by  
driving a 50 Ohms coax with a 10 Ohms output impedance.

Bye,
Said

In a message dated 9/15/2014 10:43:24 Pacific Daylight Time,  
dave.martindale at gmail.com writes:

I will agree that the end termination is optional if you are  delivering a 
pulse signal to just one input, which is at the far end of the  coax.  

However, I think there's still a problem with series-only termination  when 
the pulse signal is daisy-chained through multiple inputs.  When you  apply 
5 volts through a 50 ohm terminator to a 50 ohm cable, the instantaneous  
voltage on the coax is only 2.5 V.  A pulse of amplitude 2.5 V travels  down 
the cable, and reflects from the open far end.  The reflection  travels back 
along the cable to the source, raising the voltage from 2.5 to 5  V as it 
passes.

A device input located at the far end of the cable sees a single edge of  5 
V amplitude, so it's happy.  But anything located somewhere along the  
cable run sees two edges: one from 0 to 2.5 V, then a constant 2.5 V for a  
period equal to twice the delay of the remaining cable, then another edge from  
2.5 to 5 V.  Depending on the input threshold, this in-between device  might 
trigger reliably on the first edge, the second edge, or not reliably on  
either.

Having proper far-end termination is critical for analog video, where  
daisy-chaining is common, and a reflection that's even 1% of the amplitude of  
the original signal is likely to be visible as a ghost image.  With pulse  
signals, maybe it makes more sense to use one cable per device input, input  
plus lots of distribution amplifiers and splitters.

- Dave

On Mon, Sep 15, 2014 at 1:13 PM, S. Jackson via  time-nuts 
<_time-nuts at febo.com_ (mailto:time-nuts at febo.com) > wrote:

Hi  Dave,

yes there is a reason.

The "standard" 1PPS signal  termination (Thunderbolt etc) used to be 5 Ohms
or less series  termination into a 50 Ohms coax (yikes), then end-terminate
to  get  rid of all the undesired reflections.

Your example below is properly  terminating a 75 Ohms coax with a 75 Ohms
series termination. The  end-termination then becomes optional and affects 
the
signal level  at the sink. So if a higher signal level is desired, simply
leave   off the 75 Ohms end termination.

But in the case of the Thunderbolt  they don't use a 50 Ohms output
impedance, they use something around 5  Ohms. That is the problem here: the 
total
impedance mismatch from the  very low source impedance into the 50 Ohms  
coax.

The reason  they do that is so that they can generate a "proper" signal
level that is  approaching 5V across the 50 Ohms end termination so that the
signal  remains CMOS compatible. Otherwise if they properly terminated  the
driver  with 50 Ohms they would have a voltage divider and would  only 
generate
2.5V  at the sink.

bye,
Said

In a  message dated 9/15/2014 06:04:34 Pacific Daylight Time,

_dave.martindale at gmail.com_ (mailto:dave.martindale at gmail.com)   writes:

Is there  any reason (other than cost) not to both  series-terminate the
source and  parallel-terminate the  sink?

When I was dealing with analog video, the  standard  distribution method was
:

1. Buffer amplifier with high  input  impedance, very low output
impedance, and a gain of 2 (so 1 V  P-P input  becomes 2 V P-P out)

2. A series 75 ohm resistor from  the amp output to  each individual video
output.  This formed a  2:1 voltage divider with  the 75 ohm coax to give
1 V P-P on the  cable.  It also isolates the  loads from each other.

3. A  single video signal could be looped through  multiple high  impedance
loads.

4. 75 ohm parallel termination at  the far  end of the signal path
(usually on the last device).

This   way, every device along the way saw an undistorted copy of  the
signal.  The buffer amplifier sees a simple resistive  load.  And  any
reflections are absorbed at both ends of the  cable.

-  Dave

On 15/09/2014 02:04, Fuqua, Bill L  wrote:
> A lot of devices  have a low output impedance so that  the signal can be
split using a TEE  adapter with little loss or  need for a distribution
amplifier.
>  However, the cables must  be impedance matched at far end, scope input,
to  prevent  reflections which are the source of the ringing.
> You can match   the impedance at the source and you will get a reflection
which will then  be  absorbed by the source resistance. One way to do this
> is to  get a  small 15 turn pot about 100 Ohms put it, in series with  the
input source and  adjust it until the ringing is gone or you can  put it at
the far end
>  ,input of the scope, to ground and do  the same.  But the best solution
is  to get a good feed thru 50  Ohm terminator and put it on the input of 
the
scope.
>   Bill
>
>

_______________________________________________
time-nuts   mailing list -- _time-nuts at febo.com_ 
(mailto:time-nuts at febo.com) 
To unsubscribe,  go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and  follow the  instructions  there.
_______________________________________________
time-nuts  mailing list -- _time-nuts at febo.com_ (mailto:time-nuts at febo.com) 
To unsubscribe,  go to 
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and  follow the instructions  there.