Refael Whyte
Date: 3 July 2008
The PPS signal has a propagation delay of 22.5ns through the system. A prototype successfully propagated a signal from a GPS to several measurement devices, and GPS software running on a workstation was able to interact with the GPS through the device.
DAG cards are used to capture network traffic. At The University of Waikato they are used to capture all inbound and outbound traffic. They are designed by Endace[2]. When capturing network traffic the DAG card can time-stamp the arriving packets. This aids in analysis at a later stage. When comparing time-stamps of packets captured with different DAG cards, the same time reference is needed.
GPS satellites have accurate atomic clocks on board. GPS receivers use the transmitted time for positioning, thus can tell the time down to nano-second accuracy. This can be used to synchronize the time-stamping of network packet capture all over the world.
The Trimble Acutime2000 and Gold are GPS's designed for accurate time keeping [5]. To provide accurate timing the GPS uses two signals, a serial packet and a PPS signal. The PPS is a signal that transmits a narrow pulse every second to indicate when the time contained in the serial packet is valid. This system is used because of the unknown latency of servicing interrupts on a CPU. In most cases the time is on the rising edge of the PPS.
At present the GPS receiver can only provide time for one DAG card. There are 8 DAG cards in the WAND Group server room that need to be synchronized. A PCB is needed to split the PPS to 8 DAG cards, and provide serial interfacing to the GPS.
This paper goes through the design, testing and results of making this PCB. In this process two version were created.
RS422 is a differential pair standard. The differential signals can have a faster bit rate, greater noise immunity and a longer cable run. The longer cable run is needed for the GPS on the roof.
The latest version of the gEDA tools can be found at http://www.geda.seul.org, along with adequate documentation. The gEDA suit is not monolithic, but a bundle of different applications in the UNIX style. The main applications are:
Pcb the program is a reasonably advanced PCB layout package. It has a number of advanced features like auto-route and auto-place. Pcb has recently been greatly improved after being accepted into the Google summer of code in 2007. However it is still missing some key features, and could do with a number of bug fixes.
For this board the assembly was to be done by hand. This limited the pitch of the leads on the IC. A general rule of thumb is any pitch less than 0.5mm is too small to solder by hand. It also limited the use of devices requiring reflow soldering e.g. BGA's.
An RJ45 female plug was used to output the PPS because it is used on the DAG cards for the PPS input. In an RJ45 plug there are 8 different wires. The DAG card input is only on pins 3 and 6. Pin three is positive input and pin 6 the negative input[8].
One of the design parameters is that the board could support both 8 RJ45 outputs and 4 RJ45 outputs, to reduce the price of the board. The initial idea was to use two 2x2 RJ45 plugs, however these were costly. Another way is to have one 2x4 RJ45 socket, and one 1x4 RJ45 socket. Then integrate both footprints so the sockets can be interchanged on the PCB. The 1x4 RJ45 plug cost $17.50 from Farnell Electronics, while the 2x4 RJ45 plug cost $55.82 from Farnell, in July 2008.
On the second revision 2.54mm headers were used for the serial I/O because it takes up less room on the board. The needed DB sockets were mounted in the box. Version 1 the PCB size was limited by the number of connects needed around the outside, resulting in a large amount of wasted space. The second version was logic bound, PCB sized limited by the number of components.
A switch mode power supply was chosen for the power input, because they are more efficient than a linear voltage regulator. For a linear regulator to step down a 24
to 5
while supplying a current of 1A it would result in a power loss of 19W. A switch mode power supply is normally greater than 70% efficient.
National semi-conductor produce a range of switch mode power supplies and provide application notes. From summing the maximum current draw of each IC, the total possible current draw is 1300mA. Using safety a factor of 2, a maximum current of 2A was used when designing the switch mode power supply components and copper tracks. The power supply was designed easily with the National Semi-conductors data-sheet, which suggests components.
In the PCB design the top layer was chosen for the ground plane, and the bottom plane for the power plane. 0.1
F decoupling capacitors were placed next to all the ICs, to filter ripple and increase response time. Ferrite beads were used between I/O to stop high frequency ripple on the ground planes.
A 
termination resistor is needed between the incoming positive and negative receivers[6]. This is sufficient to maintain signal integrity and reduce ringing on the transmission line. According to[3] termination between differential pairs should be carried out differently. However due to the slow rise time a resistor between lines is adequate.
The PCB's were testing first by doing a ``smoke test'' on the power supply to detect any possible shorts that could be fatal to components. Then the serial connections were tested by plugging into the GPS and monitoring the outputs on the scope. Then communicating with the computer was undertaken. Finally after the GPS gained satellites the PPS was generated and monitored on the scope. The final test was plugging into a DAG card in a box for doing research for WAND. The DAG card automatically detected the PPS and worked!
Unfortunately some components were chosen without checking there commercial availability. This caused some problems.
Figure 1 shows the pulse per second arriving at the PCB with no termination resistor. Notice how the reflection generated by the falling edge is half the height of the original signal. Also notice the noise caused by reflections on the rising edge of the PPS. Any noise and reflections on the PPS is undesired. Over and under shoot can cause damage to the device, and possibly cause false triggers, which for a timing application destroys any reliability.
When this data set was captured the PPS signal width was set to 10
s. However the PPS signal can be set anywhere from 10
S to 500mS[5].
Figure 2 shows an incoming RS422+ packet from the GPS, with the outgoing RS232 signal at the same time. The incoming RS422+ signal is shown in red, and the outgoing RS232 signal in grey. The graph shows there is only a small latency between incoming and outgoing serial communications. There is little distortion or loss of signal integrity through the board. The graph shows how the PCB translates effectively between the two different communication standards of RS422 and RS232, with little time delay.
The Trimble Accutime software for monitoring the GPS worked without trouble using COM1 serial port on a Windows XP machine. The software was downloaded from http://www.trimble.com and is less than a megabyte. The GPS did take about 1/2 hour to get satellite lock the first time it was turning on.
Figure 3 shows the incoming RS422- PPS in green against the outgoing RS422- PPS in red. From the graph the incoming signal has a lower starting voltage. This is properly due to the attenuation in the cable.
The incoming PPS signal has reflections on the falling edge with affects the outgoing signal. This noise is only introduced in the lower part of the signal. This shouldn't affect the integrity and affect of the output PPS.
Figure 4 shows the rise time and propagation delay of the PPS signal. The green is the incoming signal and the red the outgoing. The rise time measured in this capture was 21ns. Slightly larger than given 20ns [5]. This is alright as a slower rise time is better in this situation.
The propagation delay measured between the PPS arriving at the DB-25 plug and exiting at the RJ45 ranged from 22.5ns to 23ns. This would be easily compensated for by changing the GPS settings.
Figure 5 shows the PPS signal input with termination resistor of
on the receiver. The reflection only peaks at -0.6V. Compared to PPS input without termination had a peak reflection of -3V, which is 5 times larger. With termination the reflections have damping out after 
without termination the reflections take about the same time.
This is the first commercially fabricated PCB that I have designed, I've learnt a lot from the experience. It was enjoyable creating the board from the start of research to the end of powering up the board and finally seeing it work.
The schematic files are provided in a .tar.gz file, because the are in four .sch files, along with all the symbols. One for the power supply power2.sch, one of the LED controler leds2.sch, one for the pulse per second disturbution pps.sch, and the final one for the serial communication serial_io.sch. files/schematics.tar.gz
The pcb layout, with the needed symbols has been file is provided below. files/gspv2_design.tar.gz
Any question or comments feel free to contact me at:
rzw2 at students dot waikato dot ac dot nz
This document was generated using the LaTeX2HTML translator Version 2002-2-1 (1.71)
Copyright © 1993, 1994, 1995, 1996,
Nikos Drakos,
Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999,
Ross Moore,
Mathematics Department, Macquarie University, Sydney.
The command line arguments were:
latex2html -split 0 website_report.tex
The translation was initiated by Refael Whyte on 2008-08-26
