GBT MEMO ---- : One Second Pulse Distribution.

by R. Norrod & F. Ghigo

Addendum: December 1, 2004.

Delay measurements for new pulsar backends

(The diagram below has been updated with these new measurements) 
The delay to GASP is 139nS; to CGSR2 128nS.  The same procedure was used as
described in the web page, except the Racal/Dana counter in the Timing Rack
was used instead of the Fluke 6681 Timer/Counter.  That should be fine, 
but it has 1ns resolution.  I measured to the end of the NRAO 1pps cables; 
there are of course delays from these points down into the bowels 
of the backends.

Summary for pulsar backends:
  Site 1Hz to:      delay
  -------------   ----------
   BCPM            191.2 ns
   GASP            139   ns
   CGSR2           128   ns
   Spigot          107.4 ns
   SpectralProc     96.3 ns

February 8, 2001

Introduction

For observations that require precise timing, a "one pulse per second" (1pps) signal is generated and is distributed to several locations in the GBT control and equipment rooms, and to the GBT receiver room. This memo reports measurements of the time delays of the 1pps signal to the various devices.

SiteTime

As shown in the diagram, the master clock generates a 1pps signal, locked to a reference signal from the maser frequency standard. This 1pps, referred to as "SiteTime", is always kept within a few microseconds of UTC by comparison with time signals from a GPS receiver.

Logging delays

The SiteTime 1pps is compared with the GPS signal and with other 1pps signals by selecting pairs of signals with the switch and measuring the relative delay with the Racal/Dana counter, which reads to 1 ns resolution. The switch and counter are computer controlled, and automatically record the time delays between SiteTime, GPS, and the other signals. These delays are recorded about once every two minutes and stored in log files.
These log files in various formats may be found in the timing archives.

On 1/31/2001, delays of the 1PPS signal at various points in the Equipment Room were measured. A Fluke 6681 Timer/Counter was used to measure the delays. Two RG223 cables about 25 feet long, and matched to 0.3 nS were used to connect to the test points. The Fluke counter A input was connected to the cable normally connected to the A input of the Racal/Dana counter in the 1PPS Measurement system located in the Timing Center rack. The 1PPS Mux was locked to the X0 input, the Site Time 1PPS from the Trak clock. The Fluke counter B input was then connected to various points in the 1pps distribution system in the Equipment Room:

  1. To the cable normally connected to the Racal/Dana B input (mux locked to X0). A to B measured 2.8 nS. With oscilloscope, both pulses had rise times of about 20 nS. (The trigger level of both inputs to the Fluke counter was set to 1.0 Volts, 50 Ohm input impedance for all these measurements.)
  2. To the input of the 1PPS Buffer in LAN Rack RNG3-8, by connecting to a BNC Tee at the chassis input. A to B measured 55.3 nS.
  3. To the "L" output of the 1PPS Buffer in RNG3-8. A to B measured 78.5 nS. (Checked five or six of the unused outputs and all were within 1 nS of this value.)
  4. To the cable normally connected to the 1PPS input of the DCR, WC541. A to B measured 94.2 nS.
  5. To the cable normally connected to the 1PPS input of the Spectral Processor, WC542. A to B measured 96.3 nS.
  6. To the cable normally connected to the 1PPS input of the GBT Spectrometer, WC543. A to B measured 107.4 nS.
  7. To the cable normally connected to the 1PPS input of the LO Reference Receiver located in the Analog Filter Rack, WC54. A to B measured 129.2 nS.

graph