During a recent code review of a colleague's code, I noticed that for signals at frequencies > 3.2GHz and power > -60dBm, they issued a PRESEL CENTER command and re-read. I asked them about it (since my organization isn't doing it and it slows down reading time significantly) and they indicated that Agilent told them that this was a required setting (under those conditions). So I set my sig gen at 5GHz at a power of -50dBm and measured. Then I hit the PRESEL CENTER command and noticed a -4.6MHz adjustment and a higher reading by about .5dBm. It's interesting that the adjustment at 18GHz was only in the kHz off.
So my question is why is this setting optional and how can I find more commands/settings that have additional requirements on them to be accurate? Also, can someone please explain this command more clearly than the documentation? I'll need to explain to my design engineers and upper management that the data we've been taking for the past few years is not necessarily accurate.
Also noticed that we get different readings depending on the firmware version...the later the version (as late as A.11.20) we see less of a difference. An earlier version (A.11.10) gives a much more different reading.
The specifications for PSA and X-Series for frequencies in the preselected bands are guaranteed if Preselector Centering is applied before making a measurement. The preselector does have hysteresis and does have some drift, especially when tuning it over a wide range. The analyzers are calibrated with preselector centering applied. To center the preselector, the analyzer tunes the preselector up and down and looks to see where the signal drops by specified amount (about 5 dB). The preselector is then tuned to the mid point. There is some ripple in the pass band of the preselector, so if the preselector has not been centered, there will be some amplitude difference. The X-Series has improvements to its alignments that generally reduce the need for preselector centering. The X-Series has a built in noise source that it uses to preselector center. The PSA required an external CW source be used. Also, the X-Series does a two frequency calibration of the preselector whenever the full internal alignments are run. Additonally, the X-Series has a “Characterize preselector” alignment that can be done under the System -> Alignment menu to further calibrate the preselector’s default tuning. Consequently, for the PSA, if the preselector’s default tuning is found to be further than half the YTF’s bandwidth, then it is recommended to send in the instrument for calibration and request the service center TME software adjustment for the preselector be run to improve the default YTF tuning. The YTF’s bandwidth starts at about 30 MHz at 3 GHz and increases to 60 MHz at 26.5 GHz. Above 26.5 GHz for the mm-wave PSAs, the YTF bandwidth is about 300 MHz.
Is there any information on how much error would be in a measurement if the preselector centering is performed at only one single frequency during a measurement sweep over frequency vs every frequency being measured in a frequency sweep? What about over time? We are performing thousands of measurements a day and would be severely hampered if we are required to center the preselector for all measurements above 3GHz.
For many of our customers it is impractical to center the preselector before each measurement; however, our specifications are based on preselector centering applied. There are ways to mitigate the additional errors from not centering. Firstly, it is advantageous to limit the frequency range that you sweep over to reduce the effect of YTF hysteresis. For PSA one strategy is to first characterize the preselector yourself by going to each frequency of interest and centering the preselector and recording how far off it was at each frequency. Then, when you run your test and tune the analyzer’s center frequency, simply set the preselector offset to what it was when you ran you characterization procedure. For X-Series, if you simply run the Characterize preselector adjustment under the System menu, the default preselector tuning will be quite good. You may only see a few tenths of a dB difference in amplitude between not preselector centering and preselector centering. Practically, once you have run the characterization and have run the internal alignments, with X-Series, you should be able to achieve the amplitude specification in high band without preselector centering. Nevertheless, I don’t have any definitive amplitude difference in specifications when not preselector centering. App Note 1443 on page 10 has graphs to demonstrate preselector hysteresis for PSA.