Skip navigation
All Places > Keysight Blogs > Better Measurements: The RF Test Blog > Blog > 2018 > January

  Fortunately, we can make things better—for your signal analyzer


Note from Ben Zarlingo: This guest post comes to us from Bill Scharf, a Keysight engineer with long experience in microwave signal analysis.


If you use even a gently aged spectrum analyzer, you may sometimes wonder why its amplitude accuracy above about 4 GHz is slightly worse than when it was new or after it has been freshly calibrated. Personally, I sometimes wonder why I cannot do the things I did when I was 20 years old.


In both cases aging is occurring. Although nothing is technically broken, we can make things better without magically locating a certain DeLorean car equipped with a flux capacitor and then driving 88 mph, hoping for a lightning strike, and traveling back in time.


What is a preselector, and why would it drift?

If we assume the instrument is a Keysight X-Series signal analyzer, what has probably happened is the preselector, sometimes called a YIG-tuned filter or tracking filter, has drifted a bit—but not enough to cause an out-of-specification situation. In an X-Series analyzer, the preselector is located in the signal path between the input attenuator and the first mixer, and it is used only at tuned frequencies of 3.6 GHz and higher.


The filter bandwidth should be wide enough to measure the desired signal, yet narrow enough to reject image frequencies and undesired signals (which may overload the first mixer). Depending on the tuned frequency, the bandwidth of the filter ranges from about 40 MHz to 75 MHz. Filter shape and ripple across the passband also vary with tuned frequency. As the analyzer tunes, the preselector filter tracks the change and provides a “centered” passband at the current frequency, as shown below.

Frequency response or gain/attenuation parameters of a YIG-tuned filter, used as a preselector in signal analzyers to remove undesired image or out-of-band signals and the spurious responses they would create in the signal analysis results

Typical passband response of a YIG-tuned preselector

Instrument software automatically handles most of this preselector tuning; however, careful adjustment of the instrument will help deal with the rest.


Ensuring better performance

As the instrument ages, especially its preselector assembly, the filter bandpass will drift. As a result, the signal being measured might fall in an area of passband ripple or on a steeper portion of the filter response. Here are three tips to help ensure the best performance:

  •  For the absolute best amplitude accuracy, the Preselector Center function (accessible via the front panel or SCPI) uses internal calibration signals to vary the preselector filter tuning in real time and obtains the best possible tuning. Be forewarned that this routine is time-consuming. If you need the very best amplitude accuracy using the preselector, then re-center the preselector at each measurement frequency.
  • Every three to six months, apply the Characterize Preselector routine. This performs “preselector centering” at various pre-determined frequencies up to the maximum frequency range of your analyzer. The analyzer stores the tuning values and automatically uses them the next time the analyzer is tuned to those frequencies. One advantage: after this routine runs, you may not need to rely on the slower Preselector Center routine (above). No external equipment is required: simply press System, Alignments and Advanced then select Characterize Preselector.
  • Bypass the preselector filter. If your instrument contains option MPB, microwave preselector bypass, you can select the bypass path and remove the preselector from the signal path. The downside: the instrument is no longer filtering the input signals (i.e., it isn’t “preselected”). Depending on the span setting, you may see image frequencies that are not being rejected by the preselector and so appear at the first mixer. The advantage: the bandwidth is about 800 MHz at the first mixer, preselector drift is no longer an issue, and measurement speed may increase because the instrument is no longer trying to avoid oversweeping the preselector filter.

More detail is available in our preselector tuning application note.

Wrapping up

Three closing comments: The “Y” in the YIG-tuned filter, when inverted, is almost the same schematic symbol as the flux capacitor. If you are more than 20 years old, use a knee brace when running marathons, thereby avoiding future trips to the hospital. Those of you that have an X-Series analyzer can use the Characterize Preselector routine to optimize accuracy between periodic calibrations.

  Adapting to future circumstances instead of expecting to anticipate them

 There’s nothing inherently wrong with trying to predict the future, whether that of technology or any other area. It’s easy to understand why “trying” is an attractive pastime, but expecting consistent success is where engineers and others may run into trouble.

Instead, I suggest that engineers use their super-powers of creative adaptation.

My jaded attitude toward predictions comes from work I did a couple of decades ago, forecasting the future sales (0 to 18 months out) of about two dozen measurement products. I put my analytical skills to work with some modest success, but a little honest self-appraisal left me doubting that I’d added real value. Sometimes I was just lucky, and it was hard to take much satisfaction from that.

Some research into the general landscape of prognostication left me wondering if maybe the universe was actually hostile to the whole enterprise of predicting the future. Or if not actively hostile, then resistant in a passive and maddening way.

Back then, greater minds than mine had repeatedly come to grief in such prediction efforts, including a group of brilliant academics and bureaucrats in Japan. They were economists and mathematical modelers, and despite their dedication and diligence, they were no more successful than I was.

In this situation the obvious question was to ask what kind of approaches, if any, were effective in somehow handling the important unknowns the future held. If you accept that you can’t reliably predict the future, what can you do?

In short, you can adapt as the future arrives. To be more successful than others in your field, you can work to adapt faster and better than they do. In fact, you may be able to speed up the process by pre-adapting using techniques such as scenario planning. In scenario planning, multiple possible futures are considered, and steps are taken in advance to outline carefully considered responses to the ones judged most likely to occur.

Scenario planning is usually thought of as a large-scale strategic activity, but you may already be doing it with a narrower view. For example, your designs may be anticipating a clear price/performance trend in either digital signal processing or analog semiconductors such that your product will be ready for the new leading edge. Tactically, this may mean implementing a modular design that lets you drop in the new elements as soon as they’re available in quantity.

As much as I’ve been disappointed in our collective inability to accurately predict the future, I have been repeatedly impressed by the ability of designers to adapt as technologies and markets evolve. Take wireless networking as one example.

Crowd of over 100,000 people at Michigan Stadium.  An example of the demands of large numbers of wireless networking and cellular data users in close proximity

The original designers of Michigan Stadium anticipated that it would need to hold more than 100,000 people, and designed its footings accordingly. However, they had no conception of a future where the vast majority of the fans would be carrying wireless telephones and would expect mobile network or Wi-Fi access. (photo from Wikipedia)

The definitive Wi-Fi standard, 802.11b, emerged more than 20 years ago, in 1997. 3G telecom networks began appearing perhaps a year later. In the years since, growth in all dimensions—users, connected devices, infrastructure, and data rates—has been enormous and continuous. It’s likely to continue at a similar pace for years to come.

While the original standards couldn’t handle these demands, creative engineers were—and still are—constantly working to adapt and expand. They continue to succeed beyond the expectations of many, including me.

They’re also making it clear that predicting the future is less important than creatively responding to shifting demands, expectations, and technologies. That multi-dimensional creativity has included OFDM, OFDMA, MIMO (including multi-user MIMO), beamforming, carrier aggregation, and manufacturing techniques that make microwave and millimeter devices practical and affordable.

Twenty years after my forecasting adventures, the underlying lesson—and my suspicion about nature’s hostility to prediction—remain the same: count on relentless change, and rely on your adaptability and creativity. It’s OK to burn a few mental cycles speculating about what’s over the horizon, but our real power lies in our ability to solve problems and optimize designs when the future becomes the present.