One of the most useful pieces of test gear to become generally available to the average broadcast engineer in the last 20 years or so is the spectrum analyzer. This electronic tool, while still rather pricey, is in many cases the best way to find the sources of problems of interference and to make proper adjustment of such things as satellite receivers and HD Radio transmission systems.
As with any complex anything, it is axiomatic that the learning curve for a spectrum analyzer is not short; and even for those who “know the ropes” of operating it, a great deal of care is required to make certain that what a spectrum analyzer is telling you is the truth. Such a situation recently presented itself at one of my client stations. Even given my own experience, I wasn’t quite prepared for what I found when I needed to use one to find the source of a problem. Since I’ve not seen much written on the subject of spectrum analyzer pitfalls for the relative newcomer, I felt that what I learned should be shared with those who’ve yet to deal much with spectrum analyzers.
This story started when the local manager at one of my clients, a 50 kW noncommercial FM, called to tell me that the station had received an interference complaint from a technician at a nearby cell tower site, within a half mile of the station’s transmitter. The ninth harmonic of the station frequency, 91.9 MHz, allegedly was interfering with one of the cell site’s receive frequencies. I contacted the group director of engineering, informing him of the situation. He had the company’s analyzer sent right out.
When it arrived, we set it up with the connection scheme shown in Fig. 1. From the transmitter monitor output, a T-connector was added, with one output connected to the station modulation monitor, the other output to the spectrum analyzer via a pad and a notch filter. The modulation monitor had to stay in the line to send audio back to network control, so that the mother ship would know that all was well.
As it turned out, that connection scheme was a mistake.
I should mention for the benefit of the first-time user (and for the spectrum analyzer) that no direct connection should be made from a transmitter to a spectrum analyzer without including a resistive pad in the line to protect the spectrum analyzer from being over-driven and even damaged by the excessive RF input. While it is preferable to use a variable pad, a fixed one of known attenuation will work. Try 40–70 dB of resistive padding, depending on the output level of the monitor port.
Also, a tuned band-stop filter, to attenuate the transmitter frequency, is required to prevent erroneous spurious readings on the spectrum analyzer, again from overdriving. The filter should be tuned broadly enough to reduce the level of input in the broadcast FM band, while at the same time allowing little or no attenuation above that. However, the first measurement (as seen in Fig. 2) should be with the band-stop filter out of the line and the transmitter completely unmodulated (including turning the stereo pilot off). You could call it the “dead carrier reference level.” This will establish a base measurement against which all your other analyzer’s measurements would be compared. You may use either the internal or external RF pad to help establish that base point and then don’t touch it. Insert the fundamental frequency filter and, if you wish, take another measurement to establish its actual attenuation. (See Fig 3.) Then you may get on with the rest of the measurements.
Once set up, our first look at the spectrum analyzer appeared to reveal a sizable problem (Fig. 4). The harmonics looked to be well over the –80 dBc maximum level as specified in FCC Rules 73.317(c)(d) for harmonic radiation.
We were astounded.
Understand that the transmitter used at this station is a Continental 816-R2C, with output at 17.5 kW. For a while, we thought that it was the problem. It wasn’t. I’ve cared for Continental transmitters for more than 20 years. I’ve never encountered such a problem with them, and I know that those rigs are designed with harmonic suppression in mind.
What could have gone so wrong here? We tried everything — replacing the final tube, checking and re-checking the tuning and the efficiency control — with no change in the result. At this point, we called Continental; they clued us in on a couple of things. First, regarding that tenth harmonic “spike”: The BNC connection at the transmitter monitor output utilizes a small pickup loop (about a half-inch of wire between the BNC connector center pin and ground) inside the RF output coax. The wire length is fine, considering how little RF output is needed at the monitor port to drive a modulation monitor properly, especially if the transmitter’s output power is less than its rated maximum. However, a half-inch length of wire from output port to ground is approaching one-tenth wavelength in air at 900 MHz. That’s enough to affect the response of the port in that frequency range. Continental doesn’t print this in the 816 FM transmitter series manuals, but they actually have a response curve available for that monitor output pickup loop. It shows an increase of roughly 6 dB per octave up through the 900 MHz region (Fig. 5), which is predictable. My contacts with other transmitter manufacturers yielded about the same answers.
The point is that the response change must be taken into account when using the monitor port for transmitter harmonic measurements.
MORE PIECES OF THE PUZZLE
With all that said, the pickup loop response still didn’t account for the excessive levels observed on the second through seventh harmonics. But then we got another piece of the puzzle. With a little more advice from Continental, we kept our setup the same, then put the backup transmitter on the air and shut the Continental down. The second through sixth “harmonics” were still there! (See Fig. 6.)
Wait a minute! Where could this be coming from? Then we realized: The modulation monitor was still attached to the spectrum analyzer through the T-connection.
Really? Could the modulation monitor actually produce that much “RF output” through its input? The answer proved to be an unqualified “yes!” Just ask anyone who remembers the theory of operation of a superheterodyne receiver: The local oscillator in a superhet can leak output back through the mixer and right out to the antenna input. Proving that was easy — just disconnect the modulation monitor from the system. With that done, the “harmonics” disappeared, as can be seen in Fig. 7 (and with the loss of the audio feed that the monitor provides, the mother ship freaked out!).
At that point, we decided to check the actual frequencies of the “emissions” observed. Moving the cursor of the spectrum analyzer to each of the “harmonic” frequencies proved them not to be harmonics of the transmitter at all, but rather of the modulation monitor’s local oscillator, “leaking” out the antenna input. What fooled us was that the monitor’s harmonic frequencies happened to be very close to the transmitter’s harmonic frequencies. A call to the monitor’s manufacturer’s customer service department at first yielded a response of “Are you sure?” and a promise to check it out. They called back shortly to report that their findings matched ours.
RF ROAD TRIP
At that point, the decision was made to take the spectrum analyzer “on the road” for a more unbiased look at the station’s signal. It helps that our Anritsu MS-2721A has provision for using a car battery as its power source. Most spectrum analyzers made today have that provision.
The Continental was put back on line, a location was selected about a mile from the tower in a vacant parking lot, the spectrum analyzer was set up there, using an antenna of almost-known gain, and a new set of measurements were taken. These results were more in line with what we expected (Fig. 8). The frequency scale was expanded around each of the station harmonic frequencies, a couple of which are shown here (Figs. 9 and 10). In particular, the second through seventh and the tenth harmonics show no noticeable emission from the transmitter, even though we at first expected to see something, even at that distance, from the modulation monitor’s local oscillator.
Lesson learned: When in doubt about what your spectrum analyzer is telling you, it’s time to take that show on the road, and measure away from the site. Assuming that your spectrum analyzer is battery-operable, take your readings in the field, and leave it at that.
A final note: When we later contacted that cell technician, he was still complaining that our ninth harmonic was an issue at his site. I asked him to measure that harmonic from our peak signal, and this time, he found –82.3 dBc … 2.3 dB below the FCC legal 73.317 (d) maximum limit! We’ve since learned that, with the new 4G LTE system for smartphone communications being offered by virtually all the cell providers, the cell company techs have been mandated to open up their receiver RF front ends to maximum sensitivity, in order to guarantee happy customers for their cellular broadband services, thus making them more susceptible to interference from the already-extremely-low harmonic radiation from locally operating broadcast transmitters.
The lesson: If a local cell site tech calls you or your station manager with such a complaint, ask them to verify the level of your harmonic output at the frequency in question, as referenced to your station’s peak carrier level. Chances are, per FCC rules for broadcast FM, they have nothing to complain about.
My thanks to my assistant Roger Gardner and to Virginia Beehn and Jeff Zimmer for their support in this project.
Arthur Reis, CPBE, CBNT, AMD, DRB, has been in broadcasting since 1968, 44 of those years as an engineer and 19 (full- and part-time) as chief engineer of Crawford Broadcasting’s Chicago operations. Now a consulting/contract engineer, he is owner of RadioArt Enterprises LLC, New Lenox, Ill., and a member of the SBE Education Committee.