Measurement of Ku-Band LO Frequencies: Software-Based

Introduction

Although a basic measurement of Ku-Band LNBF local oscillator (LO) frequency offsets can be done using a hardware setup of test equipment, when attempting to measure the drift over time a software-based approach is needed. While the GSP730 Spectrum Analyser is fine for simple offset measurements, it is not suitable for use in a software setup.  This is because the return values from queries through its USB interface do not have enough significant digits to be useful. The activities here are designed to explore software options for measuring and tracking LO offsets and drift. The results of these activities may be used as the basis for an automated cosmic maser observation system.

For Methanol (12.2 GHz) maser observations the readily available PLL-based Ku-Band LNBFs will probably be used at HawkRAO.  However - looking to the future - for Water maser observations (22 GHz) the Ka-Band PLL-based LNBFs available are prohibitively expensive, and so DRO-based Ka-Band LNBFs will most likely be used.  For this reason, not only will PLL-based Ku-Band LNBFs be examples, but also examples of DRO-based Ku-Band LNBFs. In fact - as DRO-based LNBFs are worse case with respect to frequency offset and drift - these will be examined first. In the spirit of trying to achieve goals with the minimum investment in equipment, a relatively inexpensive RTL-SDR 'dongle' variant will be used as the backend.

Software Overview

Initially, advantage will be taken of the pre-existing basic applications which come with installation of the RTL-SDR drivers - specifically 'rtl_power' - which can be commanded to perform a Fast Fourier Transform (FFT) of the radio frequency (RF) input to the dongle. For this investigation that application will be called by a Widows C# GUI application (for quick development).  The results for single spot LO offset measurements are graphed within the Windows application, but the results for drift measurements (nominally over one hour) will be exported as *.CSV files and plotted in Excel.

LO Frequency Drift Results

While spot LO frequency offsets are useful for getting an idea of the typical accuracy of LNBF LOs, of more importance is the drift behaviour over time. The first set of these drift measurements are done at room temperature (~ 21°C)

RTL-SDR Drift Measurement

The 'Log Drift' function of the Windows C# GUI application was used to measure the drift of the RTL-SDR sample clock against the RFS frequency standard. The measurement used 10 seconds of signal (resulting in about 12 seconds between measurements due to result processing) over a period of about 1 hour.

In the plot the initial drift over the first 5 minutes can be seen as it heats up from the increased dissipation when recording data. After that it settles as the temperature profile across the PCB stabilises. This was for the aluminium cased Nooelec RTL-SDR without an external heatsink and plugged into a USB port on the front of the PC (so hanging in air).

The drift under these conditions is about +/- 0.045 ppm from cold start to an hour's run. If the RTL-SDR was preconditioned by doing a 5 minute dummy run before the actual observation, that range is roughly halved (+/-0.016 ppm). This is +/-25 Hz @ 1530 MHz.

Assuming the LNBF LO is perfectly stable at 10.7 GHz, a 12.178 GHz signal would be read with an error of +/-25 Hz - (+/-0.002 ppm) - or a velocity error of +/-0.0006 km/s due to RTL-SDR drift under the conditions given above. Even starting from cold the error would be about +/-0.002 km/s.

This appears to indicate that heatsinking this particular model of dongle - while a good idea - is not mandatory if the dongle is run inside.

Signal Generator Drift Measurement

Once again the 'Log Drift' function is used to measure the drift of the signal generator over about 1 hour. Note that this is the 'apparent frequency' - that is, the RTL-SDR sample clock offset is not applied. This is not needed for a relative drift type observation.

Ignoring the initial drift (part of which would be the RTL-SDR dongle drift) the drift is about 0.065 ppm or about 80 Hz. As the 8-th harmonic is used as the 12GHz test signal, this translates to 640 Hz @ 12.2295 GHz - or about 0.016 km/s. If the initial drift from cold is included that's still only about 0.032 km/s.

I think it's worth noting that this and the RTL-SDR drift - in terms of velocity -  is nearly an order of magnitude less than the likely accuracy of the calculation of VLSR.

Example DRO-Based LNBF LO Drift Measurements (room temperature)

Although the above measurements of the test gear show drift, the magnitude of that drift - in terms of sky velocity - is much less than the accuracy required for useful observations of Methanol masers. As a rule of thumb accuracies of +/- 0.1 km/s would seem adequate.

NOTE: While the above drift measurements for the test gear are considered sufficient evidence for test measurements indoors, the LNBFs will be placed outdoors - and so subjected to a wider range of temperatures. Indoor measurements of LNBF drift can only imperfectly simulate conditions to be encountered outdoors.

The DRO (Dielectric Resonator Oscillator) is an inexpensive way of generating the 10.7 GHz LO signal. Unfortunately they are not very stable and have considerable drift over time and temperature.

After the initial 'jumping about' for 3 minutes or so from a cold start, the LO begins a slow drift upwards (the LO is below the signal frequency - so a drift down in frequency of the IF signal means a drift up in LO frequency). If the start frequency is taken to be 1529.585000 MHz, after 1 hour the frequency has drifted by about 1529.585000 - 1529.581500 = 0.0035 MHz = 3.5 kHz.  This equates to about 0.33 ppm - or about 0.1 km/s. This seems surprisingly good given the anecdotal reputation of DRO-based LNBF w.r.t. to frequency drift. However, it should be noted that this result was obtained under the conditions of draft-free indoors. Exposure to outside temperature excursions will be a different ballgame.

NOTE: the LBNF IF frequency axis range of the DRO drift plot below right is 6.54 ppm.

Example PLL-Based LNBF LO Drift Measurements (room temperature)

The PLL (Phase Locked Loop) is a more expensive way of generating the 10.7 GHz LO signal. The advantage of this type of LO is lower offset (presumably) and lower drift with temperature. The reference clock is a 25 MHz crystal oscillator phase-locking a voltage controlled oscillator (VC) designed to produced the 10.7 GHz LO signal.

After the initial wandering about for about 20 minutes or so from a cold start as the temperatures stabilise, the LO begins a slow drift downwards (the LO is below the signal frequency - so a drift up in frequency of the IF signal means a drift down in LO frequency). If the start frequency is taken to be 1529.584400 MHz, after 1 hour the frequency has drifted by about 1529.584400 - 1529.584850 = -0.00045 MHz = 450 Hz.  This equates to about 0.042 ppm - or about 0.013 km/s. This is roughly an order of magnitude better than the tested DRO-based LNBF. However - once again - it should be noted that this result was obtained under the conditions of draft-free indoors. Exposure to outside temperature excursions will again be a different ballgame.

NOTE: This result is of the same order as the drift from the signal generator test signal - but in the opposite direction.  The signal generator signal drifts lower over time - which means the 12.2 GHz test signal drifts lower over time.  That should mean the (assuming no drift of the 10.7 GHz LO) that the trace below should drift down over time. In fact it drifts higher in frequency - so that indicates the actual LO drift is (450 + 640) = 1090 Hz. This is about 0.1 ppm or about 0.03 km/s.

To get accurate measurements of PLL-based LNBF LOs it is clear that a better signal generator 12.2 GHz test signal is required. I will attempt to modify one of the signal generators on hand to accept either a 25 MHz RFS-derived clock reference input, or - more conveniently - a 10 MHz clock reference input.

NOTE: the LBNF IF frequency axis range of the PLL drift plot below right is 0.523 ppm (a factor of over 12 times smaller than for the DRO LNBF plot).