Rubidium Disciplined 26 MHz Oscillator
When experimenting with GSM protocols and Radio frequencies, it's useful to have your own, accurate and stable 26 MHz oscillator.
Why 26 MHz? because it's exactly 96 times the bitrate of GSM. Why stable and accurate? Because the GSM specs demand 50 ppb tolerance (that's 1.3 Hz @ 26 MHz) for a BTS.
There are GPS disciplined oscillators available and you can even use a mobile phone (has 26 MHz oscillator inside and tunes itself from receiving GSM signals around it) to get an accurate 26 MHz signal. But they all require reception of public radio signals in your lab, which might not always work. Especially if you perform your GSM tests in a shielded environment.
Rubidium oscillators, especially the 10 MHz FE-5680a, are quite affordable from online auction houses. They mostly come from China, where they were just disassembled from an old GSM BTS.
See in the following chart the working principle of the Rubidium disciplined 26 MHz osillator.
The 10 MHz from the Rubidium standard get divided down to 500 kHz. The 13 MHz from the Voltage Controlled and Temperature Controlled Crystal Oscillator (VCTCXO) also get divided down to 500 kHz. A phase comparator (simple XOR gate) compares the frequencies. The output of that gets low pass filtered and the residual dc voltage, aided by a biasing voltage of a 10 turn potentiometer, feeds the voltage control input of the VCTCXO. That's a Phase Locked Loop (PLL).
I chose the 13 MHz VCTXO because it was much cheaper than a comparable 26 MHz VCTCXO. So I had to multiply that 13 MHz by 2 with a ICS501 clock multiplier.
I produced this 26 MHz oscillator out of off-the-shelf (literally what was in my drawers) parts. So yes, you could simplify things or make things even more precise. For me it works well and from what I can measure the result produces exactly 26 MHz square (kind of) wave with high frequency accuracy.
I built the circuit in the dead-bug style on a piece of double sided FR4 material. The whole lot was built into a tin case for good shielding. And the whole assembly went into the case where my Rubidium standard was already in.
Works like a treat.
About the schematics in detail: I didn't have programmable divider (like 74HCT4059) lying around so I constructed the 26-times divider out of a binary counter (74HC393) and some AND gates (74LS08) which feed the Reset Input of the binary counter. The 20-times counter was built from a 74HC390 and the phase comparator is a 74HC86. The VCTCXO I used was an IDQ LFTVXO009902 (about EUR 10,-) which is still easy to solder with its 5mmx7mm outline. Based on this schematic you can of course build oscillators with any frequency as long as your dividers are quick enough (the versatile 74HCT4059 gets up to 40 MHz).

















