Building a Stratum-1 GPS Based NTP Server with a Soekris net4501

Overcoming Obstacles & Finishing The Build:

As I said before, my net4501 board is an OEM model and did not include the internal header for the second serial port. I had contemplated running a cable from inside the case out to the back of the board to connect to the serial port, or even soldering some wires to the underside of the PCB where the serial port is. Either way that would have meant the loss of using the only serial port for direct communication with the net4501, not something I really wanted to do. My other option was to just hook up the PPS signal and let the system get its coarse time from another source other than the GPS (like another NTP server).

Neither of the above choices really seemed that appealing, and fortunately I came across a third choice after booting up the net4501 with a quick build of NanoBSD. The system detected two serial ports in the startup logs, so that made me examine the physical board a little deeper. The serial point on the rear of the board has a TTL-to-RS232 chip near by, so I made the assumption that the location for the second serial port header used the same chip. I looked up the specs on the Maxium website, and using my trusty oscilloscope I verified that TTL signals were arriving on the pads where the converter chip goes. Eureka!

Since the Oncore receiver module also sends/receives data at TTL levels I could just wire it to the board directly! No need for the typical TTL-to-RS232 conversion. Below you can see where I wired through the DCD, TX, and RX lines @ U17 where the converter chip would go. I also added the row of pins @ JP11 which connects to the wiring I just did.

After working out the serial port issue it was pretty much easy going to finish up the wiring. I used a little 5V linear regulator to power the GPS module, the power is connected under the board to the main DC input. I could have used a spare 5V output from the main board, but I didn't want to chance drawing too much current and damaging anything, better safe than sorry.

In the left picture all the blue wires are the communication between the GPS and net4501. There's the TX & RX lines which allows the board to communicate with the GPS (for initialization and configuration) and also the GPS will output time & position data. There's also a wire from the GPS that sends the PPS (Pulse Per Second) to a GPIO pin (it's the wire with a resistor on it in the upper left corner). From that same pin a wire runs down to the front of the board where a pair of resistors are located (R61/R62) and connected on the front (common line between the two). This connection is the TMR1IN aka the high resolution timer that was mentioned on the second page.

To better explain how the time stamping works, here is part of the comment in the elan-mmcr.c source code:

The Elan has three general purpose counters, and when two of these are used just right they can hardware timestamp external events with approx 125 nsec resolution and +/- 125 nsec precision.

The rising edge of the signal will start timer 1 counting up from zero, and when the time counter polls for PPS, both counter 1 & 2 is read, as well as the GPIO bit. If a rising edge has happened, the contents of timer 1 which is how long time ago the edge happened, is subtracted from timer 2 to give us a "true time stamp".