[Cameron Meredith] starts the Hackaday.io page for one of his projects by quoting a Hackaday write-up: “A timepiece is rather a rite of passage in the world of hardware hacking“. We stand by that assertion, but we’d say most of the clocks we feature aren’t as capable as his project. He’s made a real-time-clock module controlled by a rubidium frequency standard, and since it also includes a GPS clock he can track local time dilation effects by comparing the two.
Surplus rubidium standards are readily available, but each description of one seems to feature a lot of old-fashioned hardware hacking simply to get it working. This one is no exception, an unusual connector had to be replaced and an extra power supply module attached. Once those modifications had been made and a suitable heatsink had been attached, he was able to bring the rubidium standard, an RTC module, and GPS module together with an ATMega32U4 miniature Arduino-compatible board and an LCD display. The firmware is functional, but he admits it is not finished.
All the project’s files can be found on the Hackaday.io page linked above. Future plans include also monitoring the NIST WWVB radio time signal from Fort Collins, Colorado, for an extra time dilation comparison.
We’ve featured innumerable clocks over the years here at Hackaday, but among them have been a few based upon atomic standards. More than one has been used as a lab reference standard, but most similar to this build is [Max Carters] experiments to check the accuracy of an atomic standard, also using the WWVB transmissions.
For many years I worked in the electric utility industry and I’ve built several GPS synchronized clocks for work and home use. My most recent clock, that isn’t finished yet, has been comparing the GPS time to the power line frequency. It seems like the Eastern US power grid can run perhaps 6 seconds fast or slow some days. Several of my clocks have power line frequency inputs and with a PC can log the frequency to 3 decimal places every second. It’s possible to see fluctuations during the day and major fluctuations if nuclear reactors scram half-way across the country. I recorded the frequency changes during the earthquake that damaged the Washington Monument a few years ago. At least one nuclear reactor scrammed during that event. These are things you can do when you have a highly accurate clock.
So what you’re trying to say is that we shouldn’t use AC power as a reliable 60hz time reference? /s
You mention earthquakes – living in California this is always a concern. I am curious if your fluctuations were due to the reactor scramming, or if fluctuations can be used as a pre-warning? Did you have a chance to compare the time with the USGS time of the event and see how it lined up?
I never thought about using power as a potential early warning system, but the fluctuations might travel faster than earthquakes.
Display display
Looks like a standard 2mm header for power, no big deal.
I sell on Tindie a GPS discipline board for the FE-5680A rubidium oscillator that’s widely available on eBay.
The 5680 has excellent long term stability (like GPS), but it’s short term stability (tau < 2-3 sec) isn't as good as a decent OCXO. If you're making a clock, that's not a big deal but if you're feeding a UHF or SHF frequency multiplier, you'll need something with better phase noise than the 5680A offers.
The best application of the 5680A is as an offline standard for use when GPS reception isn't available. If you calibrate a 5680A, you can count on the rubidium physics package keeping the calibration locked in much better than an OCCO would be able to do without discipline.
I’m wondering if adding a linear power supply (Instead of the regulating psu) will improve the stability. It all depends on the switching frequency of the regulators. A good power filter can also do the job.
Testing this would require 4 of the same circuit 2 of each and preferably of the same batch… A bit to much for just playing around. As you can get the time with NTP or GPS pretty cheaply.
The regulating PSU actually powers a set of linear regulators within the frequency standard. Hopefully together with the filtering caps not much of the switching noise makes it through.
The 5v rail is switching only however, but it has some nice filtering caps. I may consider adding a follow-on linear regulator just to be sure…
Another option is the Microsemi Quantum™ SA.45s CSAC that costs under USD $1000 and is a chip scale atomic clock. That is if your project’s aim is a stable clock, rather than also having the requirement to incorporate surplus rubidium gear that you already have. Once you do have a stable clock on your LAN you can distribute it with PTP, https://en.wikipedia.org/wiki/Precision_Time_Protocol
The CSAC is very impressive but this FE-5650A has better long term stability: <2 x 10^-10/year
compared <1×10^–8/year typical(*) for the to the CSAC module.
Also this disclaimer(*) in the CSAC datasheet does not inspire confidence:
"However, continuous operation of CSAC over extended period of time
may yield unpredictable aging performance, resulting in failure to meet the aging specs and may not be suitable for certain applications."
(http://www.microsemi.com/document-portal/doc_download/133305-quantum-sa-45s-csac)
The way they acheive short term accuracy in the cesium beam atomic clock (5060 series HP / Agilent / Keysight ) is by EFC correction of an ovenized crystal oscillator (10811).
However comparing to WWVB isn’t really practical as it would take almost a week to truly certify a Rubidium standard. It takes near a full day just to certify an ovenized crystal. Rubidium is at least a full magnitude or two beyond a good ovenized rock. Problem is WWVB is only 60KHz. Takes time to compare a 5 or 10 MHz standard to that.
Thanks for the tip! OK, I just ordered a muRata OC2545-DT 10.00 MHz OCXO Crystal Oscillator so that I can implement this and improve the short term accuracy.
Nice little osc…. If it lives up to its spec of 0.3 ppm per year. (3 Hz @ 10 MHz). The 10811A is spec’d at 5 x -10 per day, or 1.82 Hz per year. So doing REAL GOOD to come in under 2x that spec in a micro can OCXO. Only problem I see from spec sheet is the +/- 4 ppm EFC voltage range of 0 – 6 V. On a 10 MHz rock that’s a range of 80 Hz, or 13.333… Hz/V. You’ll need a very stable correction circuit. (Think mylar or glass caps)
Add an ESP8266 and you’ll have a wireless Stratum1 NTP server!!
http://www.absolutelyautomation.com/articles/2016/09/15/esp8266-based-wireless-ntp-server-no-internet-connection-needed
I have a similar project here: https://hackaday.io/project/5012-gpsatomic-clock-build
I decided to go with several 8×8 LED arrays so that I could incorporate menus that allow for selecting time zone, 12/24 hour display, brightness, etc. I can also “turn off” the GPS so I can measure the drift between the GPS and the Rubidium source. Went with a generic 2U case and cut a hole for the display.
I still haven’t “finished” the project, as I’d like to switch to the 16KHz source from the DS3231 as the internal timekeeper, which is significantly better than the Arduino crystal.