Lessons learned with an external clock / USRP1 interface
In order to use even a high-priced SDR for GSM experimentation, an accurate clock source is indispensable. According to the GSM standard, a clock with “absolute accuracy better than 0.05 ppm for both RF frequency generation and clocking the timebase” is required. This means that with a standard USRP1 with frequency accuracy of 2.5 ppm, an external clock is definitely required. (An external clock port is provided on the front panel.) Using an accurate external clock source is essential when using OpenBTS.
The crew at hack4fun wanted better USRP clock accuracy for OpenBTS, so they decided to interface a FA-SY 1 external clock from Funkamateur.de with their USRP1. The results are chronicled in their project log. They eventually discovered that *coincidentally*, their RFX900 daughterboard on the USRP was fried, and they also experienced issues with the external clock’s calibration.
The experience is leading the hack4fun crowd in the direction of developing a new, easy to use external clock for USRP1 OpenBTS enthusiasts.
(For those interested, a succinct explanation of the proper way to install an external clock to the USRP1 can be found on page 13 of Nicolas Krassas’ GSM Fun article. It is also important to note that the FA-SY 1 is a CMOS output, and requires modification to decrease its output voltage to prevent USRP input burn out.)
This entry was posted in project logs, RF, SDR and tagged FA-SY 1 clock, GSM, OpenBTS, USRP.
Comments
Hi,
There is an opensource hardware alternative to FA-SY1, the clock-tamer.
The price tag is much higher, but it’s much more flexible and accurate (0.28 ppm or 50 ppb with GPS option) .
There is an excellent installation guide which is also valid for other USRP reclocking options.
Wil
The FA-SY1 uses a Silicon Labs Si570 programmable oscillator, which has crystal-like stability but cannot be easily disciplined. There is an Si571 that can be disciplined, but only in a loop. An Si571 disciplined SDR L.O. would be difficult to tune over large frequency ranges. Another option is to use the FA-SY1, choose your Si570 part carefully for highest stability (use part number builder on the Silicon Labs site) then temperature control the Si570. That would do the trick for experimental purposes, but I would not go to production with such a scheme.
The Clock Tamer uses some National (now TI) clock synthesizer parts. The phase noise of these parts is significantly worse than the Si570, but still pretty good. The Clock-Tamer can be GPS disciplined, this is where it really matters in this application.
National/TI have a relatively new clock generator chip that may be worth looking at for this application, the LMK03806.
Also, Conner-Winfield/Navcom make a GPS module with an on-board NCO that goes to around 30MHz (I can’t remember the model number, but the previous version that went to 10MHz was the CW25-TIM). The CW GPS modules go for less than $100 bucks each the last time I looked. I’m not sure how good (or bad) the phase noise performance of the Connor-Winfield GPS module NCO output is, but they claim it is used for base station reference clocks.
Maybe just for experimentation with the likes of Open-BTS via an SDR, phase noise may not matter as much as stability.
Regards, David
“They eventually discovered that *coincidentally*,
their RFX900 daughterboard on the USRP was fried,”
Sounds like the good old 5 V on a 3,3V citcuit issue :)