A 10-MHz frequency standard based on the Trueposition GPSDO

Photo of finished project

The finished GPSDO frequency standard

Project interior Project interior

Project interior showing the Trueposition GPSDO board (bottom), Extron distribution amp (back panel), Raspberry Pi Zero (front panel), and power supplies (side panels).

A useful tool for the lab is a frequency standard of known accuracy and stability. 10 MHz has become the "standard standard" frequency for this purpose, and many pieces of test equipment are designed to accept a 10-MHz reference signal through a back-panel connection; common examples are function generators, RF signal generators, and frequency counters. Besides imparting a known accuracy to measurements with these instruments, a frequency standard is useful when fine-tuning oscillators of lower stability in other equipment. For many engineers, simply possessing and comparing highly accurate time and/or frequency standards becomes a hobby and an end unto itself; these people can often be found hanging out on the time-nuts mailing list.

Thanks to modern technology, what used to require exotic atomic clocks can now be achieved more easily and cheaply with a GPS disciplined oscillator, or GPSDO. The atomic clocks are still there, only now they are in orbit in the form of GPS satellites. When a GPS receiver calculates its position, it is using the measured time differences between satellite signals to jointly solve for 3-D position and time. (From the perspective of navigation, the time comes for free, and from the perspective of precision timing, the positional coordinates come for free.) The time solution has excellent accuracy when averaged over long periods (hours to days), because it ultimately derives from the atomic clocks on board the satellites. Over shorter periods (seconds to minutes), the GPS-derived time is not quite as good as a good quartz oscillator, owing to random delays on the radio signals passing through the ionosphere, and other uncertainties. In a GPSDO, the short-term stability of a good OCXO (a quartz crystal oscillator in a temperature- stabilized chamber--an "oven") is married with the long-term stability of GPS to achieve the best of both worlds. GPSDOs are specialized gear but thanks to the mobile comms industry, surplus GPSDOs are easy to find on sites like Ebay; cell sites are a major consumer of precision timing, and a GPSDO is often the way they get it.

Trueposition GPSDO

The Trueposition board is a surplus cell-site GPSDO which was widely available on Ebay between 2017 and 2019 for about $40. There is a long thread in the EEVblog forums discussing it, and a slide deck showing some of the connector pinouts. Communication with the board is by a serial UART, and everything known about it comes from reverse engineering the commands, so there are still mysteries, but basic operation is well documented. All that's needed to make the board work is a GPS antenna, power supply, and microcontroller or host system to send it serial commands. (Note that the PDF slide deck suggests a 15V supply, but it should be 12V max.)

I decided to integrate this GPSDO in an enclosure with power supplies and a Raspberry Pi system monitor. This is fine for a single 10 MHz output, but what if we need multiple outputs to drive different pieces of gear?


Extron distribution amp

Someone on the EEVblog thread linked above suggested a solution to the distribution problem: Now that video production has completely switched from analog to digital, there are plenty of surplus "video distribution amplifiers" to be found on (where else?) Ebay, with enough bandwidth for the 10-MHz signal. Extron was a popular brand of distribution amp. The Extron enclosures are mostly empty except for the PCB along the back panel where the connectors are, which makes them well suited for cramming in some extra electronics like a GPSDO board. The Extron amplifiers are popular enough for this purpose that someone even wrote a guide with some circuit modifications to improve performance. I went ahead and performed the recommended mods on my Extron PCB.

Getting the GPS antenna signal into the Extron box requires adding an SMA connector someplace, and on my Extron 300MX every available bit of space on the back panel was already taken by connectors and switches, with the main Extron PCB right behind it. (The taller Extron models have a separate PCB strip along the bottom which is easily removed, providing a space to pass signals in and out.) I eventually gave up and resorted to desolding one of the Extron's existing BNC connectors, drilling a 3/4" hole in the PCB behind it, and mounting the SMA connector in its place, with the hole patched by a piece of brass stock screwed into the back panel.

System monitor

A Raspberry Pi Zero is mounted behind the front panel and connected to a parallel 800x480 parallel RGB LCD (from buydisplay.com) via Adafruit's DPI Kippah adapter board. The Pi is connected to the GPSDO serial port via a 3.3V serial-USB dongle, and it automatically boots into Mark Sims' excellent Lady Heather software under Raspbian Linux. This provides a nice display of the current time, GPS signal strength, temperature, OCXO control voltage, and other stats. The LCD bezel was custom designed and 3-D printed on a friend's printer.

Putting it all together

Power supplies were the most challenging aspect of this project. The Extron box had a cheap switching supply mounted to the roof of the case, which I discarded and replaced with two Meanwell switchers (5 and 15 volts); these also supply the GPSDO board and the Pi Zero. In a preventive bid to reduce switching noise, I added linear post-regulators on the power rails going to the Extron amp and GPSDO. The sheer number of power rails means that there are more wires zigzagging around the enclosure, and less extra space, than I had anticipated. In addition to the power wiring, other cables connect the GPSDO serial port to the Raspberry Pi and distribute the 1 PPS (pulse per second) output of the GPSDO to a front-panel LED and the "sync" channel of the Extron distribution amp; thus the 1 PPS signal is available on the bottom row of back-panel outputs should it be needed. The Trueposition GPSDO board is mounted to the bottom of the enclosure, with just enough room to stand up the two Meanwell supplies on either side.

Thermal performance was a concern, both for stability of the ovened oscillator (OCXO) and for yjr rated temperature of the components. Fortunately, in ambient indoor conditions the internal temperature levels out around 50-55 degrees C, well within specs. In operation, the OCXO control voltage seems quite stable, more so than it was in open air. Actually verifying the stability at various time scales, or measuring the output phase noise, would require another frequency standard of equal or (preferably) higher performance, so for now I am content to assume this one is good enough!