Having recently acquired a 'spare' Icom R-75 receiver it seemed like a good time to finally try locking the receiver to a GPS disciplined 10 MHz standard. The military receivers in use here, when referenced to the 10 Mhz GPS standard, have mHz (millihertz) accuracy and superb frequency stability. Over time, these receiver characteristics become addictive. Accuracy and stability of this level are seldom required on HF but are handy on LF and MF where long duration weak signal modes with very narrow bandwidths are common.
The Icom R-75 looked like a perfect candidate for 'GPS lock' as all internal frequencies are derived from a single 30 MHz oscillator. Stock R-75s are supplied with a 30 MHz crystal and an optional TCXO (Icom CR-282) is available. My other R-75 has the TCXO option and the receiver is remarkably accurate and stable. It was hoped that 'GPS lock' would take the receiver to another level - frequency accuracy and stability similar to expensive military receivers.
The concept appeared to be a relatively simple matter of multiplying the 10 MHz GPS reference signal X3 to arrive at 30 MHz to replace the crystal or TCXO in the R-75. One important concern in this process is to not increase the phase noise characteristics of the 'clean' 10 MHz oscillator beyond the unavoidable X3 multiplication. A search turned up an interesting circuit* that appeared to exactly fill the requirement. It was developed by Charles Wenzel, of Wenzel Associates, well known for their low phase noise commercial and military products. A schematic of the circuit, along with the addition of an MMIC amplifier stage is shown below.
R-75 GPS 10-30 MHz tripler.pdf
The circuit is fairly straightforward and simple to build. Fixed value capacitors are shown for the 120 pF and 10 pF capacitors and tuning is accomplished by spreading or compressing the turns on the associated coils. Alternatively, variable capacitors could be used as a portion of the required capacitance. The MSA1105 MMIC was on hand but other devices should be suitable. MSA1105 is a 3.5 dB noise figure device with 12 dB gain and a 1 dB compression point of +18 dBm.
The proof of concept circuit was built 'dead bug' style on a small piece of scrap circuit board and located close to the shielded compartment that houses the crystal or CR-282 TCXO. The 30 MHz crystal was removed and three 1" wires connect the tripler to the R-75; wire 1 to the base of the oscillator transistor (a CR-282 connection), wire 2 to ground and wire 3 to +13.6 VDC (a CR-282 connection). The R-75 fired right up and was indeed 'locked to GPS'.
And this should be the end of a happy story ... right? Well, not exactly. With the receiver successfully locked to GPS one would expect the frequency to be 'dead on'. Using WWVB as a signal source, frequency errors of .03 Hz on SSB and .09 Hz on CW were noted. Further checks with WWV and a GPS referenced signal generator confirmed the problem. The frequency errors were consistent thoughout the tuning range of the receiver. Adjustment of the passband tuning controls, however, changed the errors by amounts up to a couple tenths of a Hz. Careful adjustment of both passband tuning controls could 'zero' the frequency but this is hardly a viable solution to remove the errors.
A review of the R-75 manual points to the likely source of the problem - there are at least a couple internally generated DDS signals. With any DDS there are rounding errors and the fewer the bits the bigger the errors. Unfortunately there's no practical way to make changes to the basic scheme of the receiver so one is stuck with the errors.
The 'GPS lock' setup is still a worthwhile improvement over the stock 30 Mhz crystal or TCXO in terms of accuracy and long term stability. It's those small frequency errors that degrade the overall usefulness of this modification and prevent 'military grade' frequency accuracy.
Had the results worked out as originally expected, the plan was to design a circuit board with the tripler, the original 30 MHz crystal and an automatic switch to select 'GPS lock' when the 10 MHz reference input was present.