Measurement of Ku-Band LO Frequencies : Hardware-Based
Introduction
At HawkRAO a variety of Ku-Band LNBFs are on hand for the purpose of observing the 12.178 GHz Methanol transition line. Critical to those observations is the local oscillator (LO) used to mix down the 12 GHz sky frequency (RF) to a lower intermediate (IF) frequency around 1400 - 1500 MHz. While it is possible to make modifications to various LNBF variants to increase the stability and correct offset, at HawkRAO it is thought desirable to not have to open up and modify the LNBFs. A basic LO frequency measurement setup (hardware-based) was implemented to measure the typical LO offset and drift for both DRO and PLL based LNBFs.
Test Setup
There are 3 uncertainties in the test setup - GSP730 spectrum frequency readout, Signal Generator frequency and LNBF Local Oscillator (LO) frequency. The first two uncertainties need to be resolved to arrive at the last - the LO frequency.
The 10 MHz output of a Rubidium Frequency Source (RFS) is fed to a comb marker generator. According to the specifications the output of the comb generator has harmonics up to at least 18 GHz - but the 10 MHz input of the RFS requires too much multiplication of the 10 MHz RFS input to reach 12 GHz with sufficient level. Instead a lower frequency common harmonic of the Comb Marker and Signal generator is used - 1500 MHz. The Signal Generator has sufficient harmonic output such as to produce a 12 GHz signal from a fundamental frequency setting of 100 MHz using the 122nd harmonic.
The harmonic 12.220 GHz of the Signal Generator radiates from a cable with a flying SMA chassis-mount socket at the end - with the short pin acting as an antenna when placed near the mouth of the LNBF.
Nominally - the 10.7 GHz LO will mix down the 12.220 GHz seen by the LNBF to 1500.000 MHz. The deviation from that 1500.000 MHz gives the offset error in LNBF LO frequency.
NOTE: The accuracy of the results below is limited by the spectral resolution of the GSP730 - therefore for actual radio astronomy observations of - say - cosmic masers, a more accurate method will have to be devised,
Test Procedure
Steps to perform the investigation...
Calibrate GSP730: Switch 3-pole coaxial switch to position 1. Set the GSP730 to 1.5 GHz centre frequency (150-th harmonic of 10 MHz from comb marker generator) with a span of 10 MHz (sufficient to cover any likely LO offset from 10.7 GHz). Note down the GSP730 frequency reading of the peak nearest 1.5 GHz. If the GSP730 was 'spot-on' the peak would be exactly 1500.000 MHz - but the GSP730 specifications indicate an accuracy of +/- 20 ppm (+/- 30 kHz at 1.5 GHz). The GSP730 has no readily accessible calibration function - so whatever the reading of the GSP730 is, that reading is to be taken as 1.5 GHz. For example, if the peak occurs on the GSP730 at 1.5000092 GHz, then other sources need to line up with reading (because it actually represents 1.5 GHz. Note that this should to be re-done if the scan bandwidth and/or the centre frequency is changed.
Adjust Signal Generator: Switch 3-pole coaxial switch to position 2. Adjust Signal Generator frequency starting from a nominal setting of 100 MHz such that the 15-th harmonic of its ~100 MHz signal is set to the same peak reading as in step 1 (i.e., the GSP730 frequency reading of the comb marker generator) as close as possible. The Signal Generator can only be changed in 1 kHz steps (therefore - 15 kHz steps at 1500 MHz) - so it is not possible in the setup at HawkRAO to get the Signal generator 'spot-on' to the 1500 GHz reading from the Comb Marker Generator. This meant a setting of 100.001 MHz for the Signal Generator to output the closest 1500 GHz harmonic. When this is done the 122-nd harmonic of the Signal Generator will be close to 12200.000 MHz.
Measure LO Offset and Drift: Switch 3-pole coaxial switch to position 3. Identify a peak which is the 12.2 GHz input to the LNBF down-converted to ~1500 MHz. Subtract that reading from 12.2 GHz to arrive at the LO frequency. Vary the temperature of the LNBF by some means - perhaps aggressively - but carefully - via a hot air gun, or just enclosing the LNBF in an sealed container to allow a slow increase in temperature due to self-heating. At HawkRAO the LNBF was placed on top of the RFS which is heated (and which can serve as a 'keep-warm' for coffee cups). Note the LO drift after some period - say - one hour. The actual temperature is not critical as the test is just to get ballpark values for drift.
Results
Two PLL LNBFs and ten DRO LNBFs were examined.
The PLL LNBF LO frequencies were found to be offset by -80 kHz and -64 kHz (-7.5 ppm and -6.1 ppm respectively). One PLL LNBF was given the 'RFS heat treatment' and was found to have drifted by -2 kHz (-0.2 ppm). That drift is equivalent to an apparent drift in LSR velocity of about 0.06 km/s
The DRO LNBF LO frequencies were found to be offset in a range covered by -1.34 MHz to + 0.55 MHz (-125 ppm and +51 ppm respectively)..
The DRO LNBF whose LO frequency was the closest to 10.7 GHz was also given the 'RFS heat treatment' and was found to have drifted by +160 kHz (15 ppm). That drift is equivalent to an apparent drift in LSR velocity of about 4.5 km/s.
Conclusion
The DRO based LO frequencies have drift which is significant and could cause velocity measurements error during an observation if there was a significant change in temperature - possibly even the difference between full sun and the occasional cloud. As each spectrum is recorded in the observation run a recalibration might be needed at a number of points in time.
The PLL based LO frequencies are much more stable (as expected) and there would probably only need to be a calibration at the beginning of the observation run. Further measurements in situ on a dish might reveal that once calibrated on the bench, no further calibration is required. This would be ideal !