Commissioning and maintaining a satellite earth station

1The station as a chain, and where each fault lives

The transmit path runs from the modem, along the inter-facility link, into the block upconverter whose local oscillator is locked to the station's 10 MHz reference, through waveguide to the orthomode transducer, the feed and the reflector. The receive path mirrors it through the low-noise block. Faults do not wander. Each class has an address on that line, and each address has a signature.

The reasonable first instinct is to skip all of this and let the modem tell you. It reports carrier-to-noise, it logs errors, and it is already installed. The trouble is that it reports one number for every fault on the chain: a wet connector, a drifting reference, a misaligned feed and an adjacent-satellite carrier all arrive as degraded carrier-to-noise, because the modem only sees what survived. Grant it per-carrier statistics and a year of logs and it still reports the symptom with nothing to say about the cause. Everything below exists to turn one alarm into an address.

Figure 1 is the half of the station that weather reaches and Figure 2 is the half a diagram can show. Table 1 maps Figure 2 onto the signatures each element produces when it fails.

Table 1. Fault classes by address in the chain. The last row surprises engineers: an out-of-band blocker compresses the low-noise block and every carrier drops together, with nothing to see inside the band a filtered analyzer is looking at.

Element	Fault class	What it looks like on a display
Inter-facility cable	loss slope, water ingress, intermittent contact, DC drop	level slope with frequency, standing-wave ripple, carrier jumping with vibration, a converter that resets
10 MHz reference	phase noise, harmonics, drift	widened carrier skirts after multiplication, spurs at multiples of 10 MHz
Block upconverter	LO leakage, compression, spectral regrowth	discrete spur at the LO, shoulders that grow faster than the carrier
Waveguide run and flanges	moisture, arcing, flange misalignment, corrosion	resonant suck-outs in a swept trace, loss that changes with weather
Feed and orthomode transducer	depolarization, wet radome, defocus	cross-polarization discrimination degraded; a flat or asymmetric peaking curve
Low-noise block	gain drift, LO unlock, saturation by an out-of-band blocker	whole-band level shift, or wideband desensitization with nothing visible in band

Interference follows allocation, so Table 1 reads differently in each band. C band pairs a 5.925 to 6.425 GHz uplink with a 3.7 to 4.2 GHz downlink, X pairs 7.9 to 8.4 with 7.25 to 7.75, Ku pairs 14.0 to 14.5 with 10.95 to 12.75, and Ka pairs 27.5 to 31.0 with 17.7 to 21.2, the operator's band plan governing at the edges. Two of those neighborhoods are crowded, the cleared 3.7 to 3.98 GHz segment in the United States and the 13.75 to 14.0 GHz uplink segment shared with radiolocation, and both return in section 9.

2Commissioning begins with the site survey

A survey made before the reflector arrives is the cheapest measurement in this paper and the one most often skipped. It establishes what the site hears before the station has transmitted a watt, and once a dispute exists it is the record that shows the emitter was there first. It is reported in field strength, because a coordination limit is written in field strength and a reading in dBm depends on which antenna and cable were used:

So a −60 dBm reading through that horn and cable is 47 dBµV at the connector and 79 dBµV/m in the air. The instrument does that point by point from an antenna factor table and a cable loss table, as Figure 3 shows.

Figure 3 is one sweep at one azimuth, and a survey is many of them. Then stop sweeping and start watching, because a sweep answers what is there now and the question that matters is what is there sometimes. Figure 4 is the form a survey record should take.

1. Fix the geometry first: planned azimuth and elevation, the horizon profile, and every known terrestrial transmitter within line of sight, in the file before any RF is measured.
2. Sweep the receive band and the segments beside it at the look angle, both polarizations, max hold for at least ten minutes each, with azimuth, elevation, antenna, factor table, cable loss, resolution bandwidth, detector and time recorded.
3. Sweep the horizon in azimuth steps of half the survey antenna's beamwidth, with max hold long enough to catch a rotating radar at least twice; section 9 derives how long.
4. Watch the band for an hour or more with a gap-free capture, so anything intermittent appears with its period and duty cycle instead of as a spike that may or may not return.
5. Archive everything with the instrument's model, serial, calibration date and the weather. A survey that cannot be reproduced cannot be compared with the one taken after the complaint arrives.

Most surveys file the clear-write trace of Figure 4. Max hold says something happened, the spectrogram says when, and the single sweep says only what was true for the milliseconds it took.

3The link budget, and which decibels you can measure

The satellite link equation is short, and the interesting thing about it is how few of its terms a field engineer owns:

At a representative geostationary slant range of 38,500 kilometers, free-space loss is 196.2 dB at 4 GHz, 205.7 dB at 12 GHz and 210.2 dB at 20 GHz. Ka pays fourteen decibels more than C over identical geometry, and the obvious conclusion is the wrong one: dish gain rises as the square of frequency, so the same reflector buys fourteen decibels at each end of the link. What Ka cannot recover is rain, which moves by tens of decibels in minutes, so Ka systems run adaptive coding. Table 2 works a downlink budget through line by line and marks the lines the station can measure.

Table 2. A worked Ku-band downlink budget. Every decibel in it is fixed by geometry or supplied by the operator except three, and those three, figure of merit, radiated power and interference, are exactly the commissioning measurements.

Line	Value	Where it comes from
Satellite EIRP toward the site	48.0 dBW	the operator's link budget
Free-space loss at 12 GHz	−205.7 dB	geometry, 38,500 km
Clear-sky atmosphere	−0.5 dB	gaseous plus pointing allowance
Earth-station G/T, 1.8 m at 150 K	+23.0 dB/K	measured on site, section 4
Boltzmann term	+228.6	physics
C/N0	93.4 dBHz
Carrier at 27.5 Msym/s	−74.4 dB	10 log of the symbol rate
Clear-sky C/N	19.0 dB	supports high-order modulation with margin

One noise-floor check decides whether the analyzer is ever the limit on Table 2. A 150 K system is −174 + 10 log(150/290) = −176.9 dBm/Hz at the LNB input and, with 60 dB of LNB gain, −116.9 dBm/Hz at the L-band output. An analyzer with a published displayed noise level of −159.9 dBm/Hz at 1 GHz, in 1 kHz resolution bandwidth, is 43 dB below that, so at the L-band port the instrument is never the limit; the one setting that can spoil a reading there is the reference level, and section 12 returns to it.

The uplink chain, and the reading at the coupler

Radiated power is never measured directly. It is assembled. A 4 W block upconverter is +36.0 dBm, or +6.0 dBW. The waveguide and feed after it lose 0.3 dB, by the manufacturer's figure and the sweep of section 8. A 1.8 m reflector at 60 percent efficiency has 46.4 dBi of gain at 14.25 GHz. The EIRP is therefore +52.1 dBW. The one number a field analyzer owns outright is the upconverter output at a −30 dB directional coupler, and it has two traps. Arithmetic: the coupler and 1.5 dB of test cable put the display 31.5 dB below the BUC output. Safety: a 400 W amplifier at +56 dBm puts +26 dBm on the same coupler port, above a +23 dBm maximum input, so a high-power chain needs a pad and the pad goes into the correction too. An amplitude offset entered in the instrument folds both into every point of the trace, noise floor included, so the screen reads power at the flange and not at the connector. The offset value is part of the record. A trace saved without it is a different number, and the first question an operator's engineer asks about a disputed EIRP is which reference plane it was taken at.

4Figure of merit, measured with the sun

Figure of merit is the receive side: antenna gain divided by system noise temperature, measured in the field by pointing first at cold sky and then at the sun, at the same elevation, and taking the ratio of the two noise powers. That ratio, conventionally called Y, gives:

The eight is 4π from the aperture Gλ²/4π and a two because a single-polarization feed collects half of an unpolarized sun. The correction L is where the method most often goes wrong, so the derivation follows. For a Gaussian main beam of half-power beamwidth θ_hp and a uniform solar disc of diameter θ_s, the reciprocal of the fraction of the disc's power the beam collects is:

For a sun well inside the beam the expression reduces to 1 + 0.35x, the linear correction most handbooks print with a coefficient between 0.35 and 0.38, and Figure 5 shows where the two forms part company. Two idealizations are granted here. The sun at centimeter wavelengths is brighter at its limb than at its center, which puts more of its power where the beam is weaker and pushes the true correction up. A real main lobe is not Gaussian at its base and carries near sidelobes, which collect part of an overfilling disc the Gaussian ignores and pull it down. Both are small while the sun is inside the beam, and both are why handbook coefficients differ in the second decimal.

Work it once and it stops being algebra. A 1.8 m Ku station measured at 12 GHz has a half-power beamwidth near 70λ/D = 0.97 degrees, so the sun underfills it: x = (0.53/0.97)² = 0.30 and L = 1.11. With a measured rise of 11.5 dB between cold sky and the sun, Y = 14.1, and with 400 solar flux units at 12 GHz, equation (3) gives 3.470 × 10⁻²² × 13.1 × 1.11 / (6.241 × 10⁻⁴ × 4 × 10⁻²⁰) = 202, or 23.0 dB/K, the figure Table 2 assumed. Check it forward from the physics, not backward from the answer: a 1.8 m reflector at 60 percent efficiency has 44.9 dBi of gain at 12 GHz, and 44.9 dBi over 150 K is 23.1 dB/K. The two agree, and the sun raised the antenna temperature by about 2,000 K.

On a 9 m station the correction takes over, as Figure 5 warned. Its beam is 0.19 degrees, so the sun overfills it nearly three times over: x = 7.4 and L = 5.19. A measured rise of 18.6 dB gives Y = 72.4, and 3.470 × 10⁻²² × 71.4 × 5.19 / (2.496 × 10⁻²³) = 5,150, or 37.1 dB/K. Forward: 58.9 dBi over 150 K is 37.1 dB/K, and the antenna temperature on the sun was about 10,700 K, the order of the sun's brightness temperature at this wavelength.

Three practical notes carry more weight than the algebra. The flux must be at the measurement frequency: observatories report at a few fixed frequencies, 8.8 and 15.4 GHz being the nearest to Ku, and the 12 GHz value is interpolated on a log-log line; that is how 400 units was reached from 270 and 560. The measurement is made above about twenty degrees of elevation, so the ground stays out of the sidelobes, and repeated per polarization with the spread reported. And the Y-factor is written down with its date, because the rise measured on purpose is the depth of the sun outage of section 8.

5Line-up and carrier access, and what goes in the file

Commissioning is a sequence, each step produces a record, and each record has to survive being read by somebody hostile. The sequence runs: point and peak, confirm the satellite, null the cross-polarization under the operator's direction, step the power up, bring modulation up and measure it against the mask, then archive.

Pointing and peaking

Peaking is done on the satellite's beacon with the analyzer on the L-band monitor port, in a narrow span, by moving the antenna through a small angle in each axis. A max-hold trace held against the live trace makes it repeatable: the antenna is on axis when the live peak meets the max-hold peak and no further movement raises it. Zero span with the sweep time set to the beam crossing draws the peaking curve directly, in the form to file; a flat-topped or asymmetric curve indicates feed defocus or reflector distortion. Confirm the satellite by matching the observed carrier plan against the operator's, never by signal strength, because a strong beacon on the wrong satellite peaks just as well. Record the beacon level with the date and the weather as the pointing baseline.

Cross-polarization, power and modulation

Under the operator's direction, transmit a clean carrier at low power while they measure the co-polarized and cross-polarized levels, and rotate the feed to the null. Operator access standards state a minimum discrimination, commonly about 30 dB, and the figure in the file is the one the operator measured. Re-check after final torque-down, because the null moves when the feed is clamped. Then step to the assigned radiated power under supervision, watching the shoulders at the coupler at each step, bring modulation up, and measure the occupied bandwidth and the spectral shape against the operator's mask. Occupied bandwidth is the number the operator checks first, and its method is part of the record: a −3 dB width and a 99 percent power width of the same carrier are different numbers, and the operator's procedure says which it wants. Figure 6 is the mask record.

Figure 6 is what the operator's engineer will ask for first, so the sweep behind it wants its resolution bandwidth, video bandwidth, detector, reference level and offset recorded alongside it: a wider resolution bandwidth fills the shoulders in and a peak detector raises them. The third record is channel power at the coupler, integrated over the declared bandwidth with an RMS detector at the authorized drive, which is the station's EIRP record; a peak-detected maximum overstates a modulated carrier by several decibels.

The archive, and what it is for

Figure 7 is the file, row by row, against the accusation each row answers.

Figure 7 is the paper's argument in one picture, and it fixes what the instrument on site has to be: one whose readings carry a published accuracy with a stated condition and whose records carry their settings. A field analyzer that exports every record with its settings attached meets the second; whether its published figures meet the first is the question of sections 9 and 10.

6Transmit chain health: back-off and intermodulation

Back-off is the cheapest fix in the chain, and it is not free. Every decibel of output back-off is a decibel off the link's carrier-to-noise, and Table 2 shows a 19 dB clear-sky margin that rain will take most of. Third-order products fall three decibels for every decibel of output back-off while the carrier falls one, so when the shoulders fail the mask one decibel of back-off usually costs less than anything else on the site. The two-tone test prices it: two equal tones produce third-order products at twice one frequency minus the other, and their level below the tones gives the intercept point:

The arithmetic in the caption of Figure 8 is the check to run on every intercept a display reports, and one further check belongs with it, because the analyzer is itself nonlinear. The published input intercept for this class of analyzer is +40.3 dBm at 1 GHz with the reference level at +20 dBm. That puts the instrument's own products at 2 × (40.3 + 6.0) = 92.6 dB below −6 dBm tones, about 20 dB under the −72.5 dBc measured. Added in phase they would raise the reading by 20 log(1 + 10^−20/20) = 0.8 dB at most. The capture was taken at a 0 dBm reference level, not +20 dBm, so the published intercept does not strictly apply and the contribution is bounded, not known; 10 dB of added input attenuation settles it, because products that fall by 20 dB were the analyzer's.

Harmonics matter for a reason the mask does not cover: the second harmonic of a C-band uplink near 6 GHz lands near 12 GHz, inside somebody's Ku downlink, and the feed does not filter what the waveguide will carry. In a harmonics table, a row whose frequency is not an integer multiple of the fundamental is a peak search returning the noise floor, recorded as a bound, not a value. A harmonic that was 60 dB down at commissioning and is 40 dB down a few years on says the amplifier has moved toward saturation, which the carrier power alone cannot show.

7Frequency references and phase noise

A Ku block upconverter locked to the station's 10 MHz reference multiplies it by 1,305 to reach a 13.05 GHz local oscillator, and phase noise rises with the square of the multiplication:

A mediocre reference that looks harmless at 10 MHz is fatal at 14 GHz, and the symptom is a modem that locks at low order modulation and refuses at high order while every level on the analyzer looks correct, because phase noise moves no power. The measurement to take before the block upconverter is blamed is the reference's own phase noise; Figure 9 shows one on a 1 GHz source.

Figure 9 needs one number beside it, and the two do not agree. The published phase noise for this family at 1 GHz and 10 kHz offset is −107.5 dBc/Hz, at 25 °C after a 10 minute warm-up, and the trace reads −124.9 at the same offset, 17 dB below the figure the instrument publishes for itself. A direct measurement reads the sum of the source and the analyzer, so the trace is either a conservative specification being beaten or a display whose method needs stating, and the screen does not say which. It is used here for its shape and for the arithmetic of multiplication, not as the source's absolute level, and the published figure is what to plan against. The rule it illustrates stands: a unit reading a source within about 10 dB of its published figure is measuring its own margin as much as the source, and any such claim wants a second method. The figure also shows how the penalty scales with where the chain starts. Carried from 1 GHz to 13.05 GHz the multiplication is 13.05 times, or 22.3 dB, so this source lands at −102.6 dBc/Hz at 10 kHz offset, which a high-order modem will live with. From a 10 MHz reference the multiplication is 1,305 times, or 62.3 dB, so a poor reference at −120 dBc/Hz at 10 kHz arrives at −57.7 dBc/Hz and nothing high-order will hold lock.

The law has a boundary, and it should be stated: a converter multiplies its reference inside a phase-locked loop, so the law holds inside the loop bandwidth, typically hundreds of hertz to a few kilohertz, and outside it the converter's own oscillator sets the floor. The reference governs the close-in noise; the converter's datasheet governs the rest. What a modem actually suffers is not the curve but its integral: a carrier-recovery loop tracks phase noise inside its bandwidth, and the noise outside it becomes a random rotation of every symbol:

Take the trace of Figure 9 at face value, which the paragraph above says is generous, and integrate it from 10 kHz to 1 MHz on log-linear segments between the tabled points: 4.9 × 10⁻⁸ rad², so φ_rms is 0.31 milliradians, or 0.018 degrees. Multiplied by 13.05 to the Ku local oscillator that is 0.23 degrees rms, about 0.4 percent of error vector magnitude, and harmless. Multiplied by 1,305 from a reference with the same curve, through a loop wide enough to pass it, it would be 23 degrees rms. A rotation of 7.7 degrees carries the outer corner of a 64-QAM constellation, at (7, 7) on its unit grid, across the decision boundary at 6, so the first case leaves more than thirty times its rms error in hand and the second, at three times the margin, leaves nothing. Figure 10 is what the first case looks like on a demodulator.

Figure 10 is a bench carrier; at an earth station the same measurement is made at the L-band monitor port on the station's own carrier, where the phase error column reads the whole chain's reference and converter contribution at once. The 710 parts per billion offset in the phase-noise capture is the other thing to read: at 13.05 GHz it is 9.3 kHz, and a receiver's acquisition range is a specification with a number in it. Judging it needs a timebase better than the thing judged: 1 ppm at 14.25 GHz is 14 kHz of uncertainty, which cannot measure a 10 kHz offset, while 0.15 ppm is 2.1 kHz, which can. That is a decision made on the order form, not on the day. A field analyzer such as the ICX-FieldHawk-R carries a TCXO under 1 ppm as standard and an OCXO under 0.15 ppm as option 01, both published for 25 °C after a 10 minute warm-up, so it is the option and not the base instrument that makes this measurement.

8Preventive maintenance: the calendar and the ladder

Maintenance is the commissioning file, repeated on a schedule, with each new record compared against the one before, and most of it is spent where Figure 11 is: at the joints. The cadence is set by how fast each part degrades and how much station time the measurement costs: a loop test takes a minute and runs before every pass, a figure of merit takes two hours and runs each quarter. Table 3 is the calendar, and the record column is the point of it.

Table 3. The maintenance calendar. Every row produces a record compared with the previous one, so the thresholds are changes from baseline, not absolute limits: a station that was never as good as its datasheet is a different problem from one that is degrading.

Cadence	Measurement	Instrument	Record, and the change that acts
Before every pass	RF loop test: injected carrier at the LNA coupler, read at the monitor port	Signal Generator, analyzer	level and date; act on 1 dB from baseline or a level that moves when the cable is flexed
Daily	Beacon level, C/N per carrier, converter supply voltage and current	monitor port, telemetry	trend against baseline; act on a step, not a slope
Weekly	Full-band receive spectrum, both polarizations, max hold ten minutes	analyzer, monitor port	trace with settings; act on a new carrier or a floor rise over 1 dB
Monthly	Transmit spectrum against the mask; channel power; harmonics	analyzer, coupler	shoulders in dBc and EIRP; act on 1 dB of shoulder growth at fixed drive
Quarterly	Figure of merit by the sun; cross-polarization with the operator; reference phase noise	analyzer, sun, operator	G/T, discrimination, phase noise at 10 kHz; act on 0.5 dB of G/T or 2 dB of discrimination
Annually	Scalar sweep of every IFL run and waveguide; DC at the far end; ground resistance; flange inspection	Signal Generator, analyzer, meter	insertion loss and ripple per run; act on ripple that was not in last year's sweep
After an event	Peaking curve in both axes; cross-polarization; sweeps of any disturbed run	analyzer, operator	compare with commissioning; a storm, a lightning strike, a lift on the structure or a feed change all count

The receive-side rows of Table 3 are organized by a ladder of four tests that share one injection point and each prove one more thing than the rung below. Figure 12 draws it. Its right-hand column is the useful one: a test that cannot prove an element is a test that sends an engineer up the tower for nothing.

Cable and waveguide, by scalar sweep

The bottom rung of Figure 12 fails most often in the cable, and the annual sweep is how a cable is caught before it fails. Insertion loss should rise smoothly with the square root of frequency; a wet connector adds disproportionate loss at the high end and, more usefully, puts periodic ripple on the trace. That ripple is a tape measure:

The stimulus is a Signal Generator swept across the band into one end of the run, with the analyzer at the other end in max hold. A reference sweep through the test cables alone is subtracted. Return loss needs a directional bridge or coupler at the source end, and in dispersive waveguide the distance uses the group velocity at the measurement frequency. The numbers to carry: a return loss of 20 dB is a reflection coefficient of 0.10, a VSWR of 1.22 and a mismatch loss of 0.04 dB, a healthy field joint; 14 dB is 0.20, 1.50 and 0.18 dB, marginal; and below 9.5 dB, where the VSWR passes 2:1, something is mechanically wrong.

The annual visit carries three checks the sweep does not. The DC check: the inter-facility cable usually carries the converters' supply, corrosion raises its loop resistance over years, and a converter that resets under load is a cable fault no sweep shows, so measure the voltage at the far end under load. The flange inspection: every bolted joint, gasket, pressure window and drain, because a corrosion track is what a sweep will not find. And the installation discipline: cables labeled at both ends, signal separated from power, cables on supports so water runs beneath them, and one grounding grid that everything reaches.

Rain fade and sun outage

Two losses are scheduled by physics and not by the calendar, and both belong in the plan so they are not reported as faults. Rain attenuation at Ku and Ka is predicted from the site's rain rate and answered by adaptive coding and modulation; the record is the fallback log read against the rain gauge, and a fade that does not match the weather is a fault.

The sun outage can be predicted to the minute. Twice a year, near the equinoxes, the sun passes behind the satellite, and for a few minutes on each of a few consecutive days its noise fills the beam. The depth of the outage is exactly the Y-factor of section 4, the same geometry seen from the other side. The 1.8 m station that measured an 11.5 dB rise on the sun loses 11.5 dB of carrier-to-noise at the center of its outage, which takes the 19.0 dB of Table 2 to 7.5 and drops every high-order carrier. The sun crosses the sky at a quarter of a degree a minute, so the transit through a 0.97 degree beam plus the sun's 0.53 lasts about six minutes, and the 9 m station's 0.19 degree beam is crossed in under three. The operator publishes the dates; the station's job is to have them in the log before the alarm.

9Interference, and why a short survey proves nothing

Most interference is continuous, and continuous interference is easy: adjacent-satellite carriers, cross-polarization, a terrestrial link in the band, the station's own intermodulation products. Point an analyzer at it and it is there. The cases that consume months are the intermittent ones, and against those a short swept survey is close to worthless. Every candidate in Figure 13 is in Table 4, sorted by mechanism, and the arithmetic that follows shows why the last two rows cost the most.

Table 4. Seven interference classes. The first five are continuous and a sweep finds them. The last two are intermittent, a sweep misses them in proportion to their duty cycle, and the derivation that follows says by how much.

Class	Mechanism	Signature	What settles it
Adjacent-satellite	a 2 or 3 degree neighbor inside the station's sidelobes, or a neighbor's uplink mis-pointed	steady carrier on a neighbor's frequency plan	pointing check against the beacon; the operator names the carrier
Cross-polarization	feed rotation off null, wet radome, depolarization in rain	the opposite polarization's carriers appear, rising with rain	discrimination re-measured with the operator
Terrestrial, in band	a fixed link or base station sharing the allocation, far stronger than any satellite	steady, often wideband, azimuth-dependent	horizon survey in field strength; bearing by the survey antenna
Terrestrial, out of band	a strong emitter outside the band compresses the LNB	every carrier drops together; nothing visible in band	wideband sweep at the L-band port; a filter ahead of the LNB restores the level
Self-interference	intermodulation in the HPA, a reference harmonic, an LO leak	products at arithmetic combinations of the station's own carriers	two-tone test at the coupler; the products vanish when one carrier is muted
Radar and pulsed	rotating maritime or weather radar in the shared segment; altimeters beside C band	periodic bursts, milliseconds long, seconds apart	gap-free capture long enough to show the period, then the bearing
Intermittent mechanical	an arcing connector, a loose flange, a cable that moves in wind	irregular bursts that correlate with weather or vibration	gap-free capture; flex the cable; the loop test with the cable disturbed

The fourth row of Table 4 deserves its arithmetic, because its mechanism is the least intuitive in the table. A full Ku transponder from a satellite at +48 dBW arrives at the output of a 1.8 m antenna near −83 dBm by the budget of Table 2: 48 − 205.7 − 0.5 + 44.9 is −113.3 dBW, or the 93.4 dBHz carrier-to-noise density over a −176.9 dBm/Hz floor, and the two agree. Twenty such transponders total about −70 dBm at the LNB input. A commercial LNB with 60 dB of gain and an output compression point near +10 dBm compresses at about −50 dBm of total input and misbehaves ten decibels earlier. That leaves it ten decibels of headroom above the whole satellite load. A terrestrial emitter delivering −50 dBm to the LNB, twenty decibels above everything the satellite sends, drags every carrier down together while a filtered analyzer sees a clean band. The compression figures are illustrative and yours are on the LNB's datasheet; the ratio is not, and it is why a filter ahead of the LNB has become routine wherever the band next door was reallocated.

The detection-probability argument

A swept analyzer covering span S at resolution bandwidth B listens to any one frequency for the fraction B/S of each sweep. For an interferer of burst length τ and period T_p, the probability of catching it on one sweep is the overlap of the dwell and the burst, and detection over a window follows the usual geometric law:

Take a rotating maritime radar in the shared Ku uplink segment: twenty revolutions per minute, so a three second period, and a 1.2 degree beam, so ten milliseconds of illumination each turn. Against a 500 MHz survey at 300 kHz resolution sweeping every 150 milliseconds, the dwell in the affected cell is 90 microseconds and the single-sweep probability works out at three and four tenths parts in a thousand. One simplification is made here, and it is generous: the ten milliseconds is treated as a continuous burst, and a radar fills it with microsecond pulses at a kilohertz or so, so the dwell must also coincide with a pulse. The times that follow are floors.

Figure 14 makes the case for engineering the survey instead of scheduling it, under three conditions. The revisit period is the lever. A fast FFT-based sweep crosses 500 MHz in under a millisecond: a real-time analyzer such as the ICX-FieldHawk publishes 577.5 GHz/s at 250 kHz resolution bandwidth, so 500 MHz takes 0.87 ms. A sweep that revisits faster than the burst lasts catches every burst that is continuous for its ten milliseconds, so the minutes above belong to the slow sweeper they were worked for. The simplification above comes back here with its cost: a radar fills those ten milliseconds with microsecond pulses, and a sweep that passes through the cell eleven times per burst still has to land a pulse inside the short time it spends there, the same coincidence at a faster clock. The fast sweep shortens the wait; it does not remove the coincidence, and only a receiver that holds the cell for the whole burst does. The derivation assumes independence between sweeps, and if the revisit is commensurate with the interferer's period the sweep can alias and never see it. And max-hold across many sweeps raises the catch probability exactly as the formula says, but destroys the timing: it tells you that something happened, never when, and cannot separate two intermittent sources. The fast sweep shrinks the first problem and leaves the other two. Figure 15 answers them.

Figure 15 is the record an intermittent case is built on. The spectrogram gives the burst its period and its duration, the two numbers that classify it: three seconds and ten milliseconds is a rotating radar, thirty seconds and forty milliseconds is something on the site, and a burst that arrives when the wind rises is a cable. The density gives the duty cycle without arithmetic, and a gap-free capture holds the timestamped IQ record of the burst itself, which a swept trace cannot supply.

Sidelobes, and the neighbor two degrees away

The first row of the class table is quantified by a design objective, not a duty cycle. Satellites on the arc are two or three degrees apart, so the station's transmit sidelobes at those angles interfere with the neighbor's uplink. For an antenna of fifty wavelengths or more, which a 1.8 m Ku reflector is at 86, the design objective is a sidelobe envelope of 29 − 25 log₁₀ θ dBi from one degree, or 100λ/D degrees where that is larger, out to twenty degrees off axis. The coordination reference pattern is looser, 32 − 25 log₁₀ θ. Figure 16 draws both against two main beams.

Figure 16 is why a 2 degree slot forces low sidelobes and why pointing error is the commonest adjacent-satellite cause, though not by the route most engineers picture. A 1.8 m beam is 0.8 degrees wide at Ku and its main lobe ends at a null near one degree, so half a degree of mis-pointing does not put the neighbor in the main lobe: on the chart's own model the beam is near 6 dBi at 1.5 degrees, well under the envelope. What it does is cost the wanted link about 4.5 dB and move the neighbor from two degrees off axis to one and a half, where the envelope relaxes by only 3 dB, from 21.5 to 24.6 dBi, and where a real pattern's first sidelobe falls. The error is paid twice, once on the wanted carrier and once on the neighbor's. Figure 17 puts the classes of Table 4 in the order a field engineer meets them: three questions reach every class.

The requirements, stated so any instrument can be scored

One more number belongs here, because the last row of the decision flow asks for a watch that runs when nobody is on site. A continuous IQ record of a 25 MHz segment runs at 31.25 MSPS, and at four bytes per complex sample, taken as a working figure, that is 125 Mbyte per second, or 10.8 Tbyte a day, which no station ships to a network operations center. A decision with a ten millisecond IQ snippet attached is 1.25 Mbyte, and a station that sees the radar above every three seconds would send twelve hundred an hour if it reported every revolution, or one if it reported the classification. The watch that is worth having raises a decision with its evidence, not a stream, and what counts as a decision is the station's own rule, not a vendor's.

Sections 2 to this one are a criterion, and Table 5 states it as requirements a reader can score any instrument against. None is a sensitivity figure, because sensitivity is not what fails a ground-segment measurement. What fails it is a reading that cannot be reproduced, a timebase worse than the thing it judges, a survey that was too short, or a record nobody can defend.

Table 5. Nine requirements. R2 is the row most often left unstated: Ku and Ka both lie above 9.5 GHz, so a station working either carries the wider published bound and its file has to say which. R3 and R4 are the rows to read twice: the requirement is on the station's reference, and what the instrument supplies is a timebase and a phase noise good enough to judge it.

	Requirement	The number	Why this number
R1	Continuous, gap-free observation matched to the suspected duty cycle	one emitter period to catch a 40 ms burst every 30 s, a few periods to fix its duty cycle, against 5.6 min swept; one transponder, 36 to 72 MHz with its guard bands, held without a sweep	equation (9) with the receiver's blindness removed: only the emitter's period is waited out, and a burst that falls between sweeps is a burst that was never there
R2	Bounded amplitude accuracy across the bands in use, with the condition stated	a published bound with its condition, at every frequency in the plan; tighter at the IF than at RF	an EIRP and a mask margin are quoted against a limit, and a reading without its uncertainty is not a reading
R3	A timebase better than the reference it judges	0.15 ppm or better; GNSS-disciplined time for multi-site records	710 ppb was the worked carrier offset; 1 ppm cannot measure it and 0.15 ppm can
R4	Instrument phase noise 10 dB under the reference as it arrives at the measurement port	a −120 dBc/Hz reference at 10 kHz offset, carried to a 1 GHz carrier by a Signal Generator or a modem locked to it, arrives there at −80, so the instrument needs −90; a −140 reference needs −110	equation (6) carries the reference up by 20 log 100 = 40 dB inside the locked source's loop, which is the one port where the instrument's own figure is published and the reference is still what is being read; through the station's LNB the multiplication is about 1,075, or 61 dB, and at 10 kHz offset the converter's loop has usually handed over to its own oscillator, so the L-band port judges the converter, not the reference
R5	Frequency coverage through the highest band the station uses	to 40 GHz for Ka; to 15 GHz for Ku; to 13 GHz for C with its harmonic	the uplink is the top of the plan and the harmonic check reaches twice it
R6	An input that survives the coupler port and a reference plane that can be moved	+23 dBm or better; an amplitude offset per trace	a 400 W chain on a 30 dB coupler is +26 dBm; the file reads power at the flange
R7	The commissioning measurement set included at purchase, with any option named before the order	channel power, OBW, mask, two-tone, harmonics, phase noise, antenna factor and amplitude offset included; demodulation as a named option	every row of the file is a different function, and a function bought later is a record that does not exist on commissioning day
R8	A recorded baseline that outlives the engineer who took it	state, trace, IQ and conditions together, exportable	section 5; a screenshot is not a record
R9	A watch that runs when nobody is on site and raises a decision with its evidence, not a stream	under 2 Mbyte per event at the antenna; a 25 MHz IQ stream is 125 Mbyte per second	10.8 Tbyte a day cannot be shipped to an operations center; a ten millisecond IQ snippet with a classification can

10Meeting the requirements: the ICX-FieldHawk and the ICX-400R

An instrument that satisfies Table 5 has to be a real-time analyzer with a published transform relation, a disciplined timebase, coverage through the highest band on the site, and a measurement set on the van's instrument, not on a price list. The ICX-FieldHawk line is built that way, in three enclosures sharing one signal chain: a handheld, a module presenting over USB or LAN, and the ICX-FieldHawk-R, an IP68 tablet. For an earth station carry the tablet, and the 40 GHz model, the ICX-400R, covers every band in section 1 and carries the onboard AI option described below. Table 6 is the nine requirements against the published values, arranged with the weak entries in plain view.

Table 6. Published values against the derived requirements. R2 is met with the wider bound above 9.5 GHz, which every Ku and Ka station carries into its file, R4 at Ku is bracketed, not published, and R7 is met only in part.

Requirement	Published value	Condition	Met
R1 gap-free observation	100 MHz analysis bandwidth; FPGA FFT with no missing samples; 61,035 frames/s and 32.768 µs POI at 2048 points; 3,906,250 frames/s and 0.512 µs at 32 points	the two published relations, worked below	Yes
R2 amplitude accuracy	±2.0 dB, 9 kHz to 9.5 GHz; ±3.0 dB, 9.5 to 40 GHz; IF flatness ±2.0 dB	25 °C, 10 minute warm-up, spur reject on; VSWR < 2.0:1 to 16 GHz and < 3.0:1 above	Yes; ±2.0 dB at the L-band IF, ±3.0 dB at Ku and Ka, both published with their conditions
R3 timebase	TCXO < 1 ppm standard; OCXO < 0.15 ppm with option 01; GNSS 1PPS ±100 ns standard, ±75 or ±50 ns with options	option 01 is the one to order for a station	Yes, with option 01
R4 instrument phase noise	−107.5 dBc/Hz at 1 GHz, 10 kHz offset (40 GHz family); −85.7 dBc/Hz at a 40 GHz carrier	no Ku carrier figure is published; the two points bracket it	Yes for a reference at −120 dBc/Hz or worse; 2.5 dB short for a −140 reference, which wants a second method
R5 coverage	9 kHz to 40 GHz (ICX-400R); 9 kHz to 9.5 GHz (ICX-090R)	image rejection > 90 dB typical to 33 GHz; IF rejection > 68 dB between 8.2 and 21.75 GHz	Yes with the ICX-400R; the ICX-090R covers C and X without the harmonic check
R6 input and reference plane	+23 dBm maximum CW input, 50 MHz and above with the preamplifier off; +10 dBm below 50 MHz or with it on; amplitude offset and correction tables	display range to +20 dBm typical on the 40 GHz family	Yes
R7 measurement set	channel power, OBW, x dB bandwidth, ACPR, IM3, SEM, harmonics, phase noise, antenna factor, amplitude offset, signal track, peak table, record and playback included; digital demodulation option 71; pulse detection option 72	options 71 and 72 are ordered with the instrument, not licensed later	Partly: the file's eight functions are included; demodulation and pulse detection are options
R8 recorded baseline	IQ capture to 128 Mbyte, 100 MHz burst, 25 MHz continuous, 125 MSPS with decimation 1 to 4096; data record and playback; API in C, C++, C#, Python, MATLAB, Qt, LabVIEW; SCPI	a continuous record is bounded at 25 MHz; a 100 MHz burst is bounded by memory	Yes
R9 unattended watch	onboard NVIDIA Jetson Orin NX module, option 08, on the ICX-FieldHawk-R only; up to 117 TOPS (sparse INT8) in a module mode drawing up to 40 W; runs models the user loads through the same API	the tablet's typical power is 10 to 16 W, so sustained throughput depends on the power mode the enclosure runs; no classifier or accuracy is published	Yes, as architecture; performance is measured on site
Instrument values from the ICX-FieldHawk datasheets, all at 25 °C after a 10 minute warm-up with spur reject standard on.

The real-time engine, worked

The R1 row of Table 6 is the only one a reader cannot check from a datasheet number alone, so here are the two published relations. The engine transforms the incoming stream in consecutive blocks of N samples at an 8 ns sample period, decimated by D, with no gap between blocks:

At 2048 points with no decimation: 10⁹ / 16,384 = 61,035 frames per second and a probability of intercept of 32.768 microseconds. At 32 points: 10⁹ / 256 = 3,906,250 frames per second and 0.512 microseconds. Both reproduce the published operating points exactly, and that agreement is the check on the relation. A radar burst of ten milliseconds is 610 frames of 16.384 microseconds at the first setting, and the microsecond pulses that fill it, which section 9 set aside, are the second setting's job: at 32 points the engine measures a 0.512 microsecond event at its true amplitude. Nothing in the bottom row of Figure 17, as section 9 modeled it, falls below the second published point; a short-pulse radar mode at a tenth of a microsecond is detected at that setting but not measured at its true amplitude, and either way the survey time of R1 is spent on the emitter's period, as Figure 14 says, not on the receiver's. The published amplitude accuracy changes at 9.5 GHz, inside Ku, which makes a rule: measure at the L-band IF whatever can be measured there, and at RF only the coupler reading, the regrowth at the amplifier and the harmonic check.

One engine, three ways in

The eight functions of R7 are included, not licensed, and every function is driven three ways. The first is the touch display through the supplied application, SpectraCore. The second is a host over the open API, in C, C++, C#, Python, MATLAB, Qt or LabVIEW, with SCPI as standard, on Windows or Linux. The third, on the ICX-FieldHawk-R with option 08, is an onboard processor running the station's own code. Figure 18 draws the three as taps on one engine, and the shared engine makes a number taken at the flange comparable with one taken in the rack a year later.

The onboard AI module, and what it changes operationally

R9 is the row the third tap in Figure 18 answers, and it changes the maintenance calendar. The ICX-FieldHawk-R takes an optional onboard NVIDIA Jetson Orin NX module, option 08, rated by its maker at up to 117 TOPS (sparse INT8) in the module's highest power mode, which draws up to 40 W. The tablet's published typical power is 10 to 16 W, so the sustained throughput inside the sealed enclosure depends on the power mode the enclosure allows, and that is the condition to carry with the number. The module runs classification and anomaly models the user loads through the open API, on IQ and spectra at the antenna, so the station logs a decision and an IQ snippet instead of a raw stream. This paper claims no classifier, accuracy, latency or number of signal classes, because none is published; the claim is architectural, and the arithmetic behind it was done in section 9.

11Honest limits

• Scalar is not vector, and the analyzer has no tracking generator. The ICX-FieldHawk datasheets list no tracking generator, so the scalar sweep of section 8 uses an external Signal Generator and, for return loss, a bridge or coupler. Scalar gives magnitude only: no phase, no group delay, no vector distance to fault. Waveguide qualification and any dispute headed for a formal report want a vector network analyzer.
• An analyzer does not demodulate the link. The demodulation option measures error vector magnitude on a generic single-carrier modulation at the IF. Energy per bit, error rate and adaptive coding behavior belong to the modem. The analyzer proves the RF chain; the modem proves the service.
• Above 9.5 GHz the amplitude figure is ±3.0 dB, and the input match loosens above 16 GHz. A VSWR of 3.0:1 at a Ka coupler port against a source match of 0.2 adds about ±0.9 dB of mismatch uncertainty on top of the published figure. A 6 dB pad at the input improves the effective return loss by 12 dB; its value goes in the offset.
• The instrument's own phase noise at a Ku carrier is not published. The two published points, −107.5 dBc/Hz at 1 GHz and −85.7 dBc/Hz at 40 GHz, both at 10 kHz offset, bracket it. Plan against the conservative end, and judge the reference where the instrument's figure is published and the reference is still what is being read: on a 1 GHz carrier locked to it, where equation (6) adds 40 dB.
• Sustained AI throughput is a power-mode question. The 117 TOPS (sparse INT8) rating belongs to a module mode drawing up to 40 W, against the tablet's published 10 to 16 W typical. Size a model for the mode the enclosure runs, and measure it.
• Off-site sources can be characterized, not resolved. A field instrument at the victim site establishes an interferer's frequency, bandwidth, timing, polarization and bearing, which opens a case; finding a distant or on-orbit source is multi-site geolocation work with the operator and the regulator.
• You may not need this instrument. A teleport with a carrier monitoring system on every feed and a bench analyzer in the rack already has the weekly and monthly rows of the calendar covered; what it lacks is the flange and the horizon. A C-band-only station that checks its uplink harmonic by other means is covered by the 9.5 GHz ICX-090R, and paying for 40 GHz to measure a 6.4 GHz uplink is the wrong trade; the selection table of section 13 gives two reasons to pay it, the harmonic at 12.85 GHz and Ka in the site's future.

12Measurement practice, and the workload the instrument is chosen for

The calendar of Table 3 makes one thing plain. Commissioning is one day with the operator on the line and the good reference in the rack. The calendar after it is fifty-two weekly sweeps, twelve mask checks, four sun measurements and one cable sweep a year, every one of them compared with the file, most of them in weather with whatever is in the van. An instrument chosen for the commissioning day is chosen for the wrong workload. The following holds for any analyzer, and most of it costs nothing.

• Set the reference level from the strongest signal at the port, not the one of interest. At the L-band monitor port the strongest carrier may be 30 dB above the one under test, and an attenuator set for the weak one compresses on the strong one with no warning. Confirm with 10 dB of added attenuation: readings that move were compressed.
• Let the instrument warm up for ten minutes before quoting an amplitude, because the published accuracy carries that condition, and enter the reference plane as an offset and record the offset with the trace: coupling factor, pad, test cable, every time.
• Use an RMS detector for power and a peak detector for masks and surveys, and write the resolution bandwidth beside every shoulder reading, because a mask margin changes with it. Survey for the interferer you suspect, not for the time you have: take the period and burst length from the class table, put them through the detection equation, and book that long.
• Never clear a trace without saving it, and never save a trace without its settings. Model, serial, calibration date, firmware, timestamp, weather.

13Summary and selection

Table 7 collects the governing relations and published values with their conditions, so they can be lifted whole into an evaluation matrix.

Table 7. Governing relations and published values with their conditions. The instrument rows are the ICX-FieldHawk published figures, the only lines that depend on which analyzer is carried up the tower. Link and budget rows are worked examples, not a specific operator's numbers.

Quantity	Relation or value	Condition
Link equation	EIRP − Lfs − Latm + G/T + 228.6	C/N0 in dBHz; Lfs 196.2 / 205.7 / 210.2 dB at 4 / 12 / 20 GHz, 38,500 km
Figure of merit	8πk(Y − 1)L / λ2F; L = x ln 2 / (1 − 2−x)	sun above 20 degrees elevation, flux at the measurement frequency; L = 1.11 (1.8 m), 5.19 (9 m) at 12 GHz
Reference multiplication	20 log10(M)	+62.3 dB for a 13.05 GHz LO from 10 MHz
Third-order intercept	Ptone + ΔIM3/2	equal tones; 30.30 dBm recomputed from the capture; regrowth about 3 dB per dB
Detection probability	1 − (1 − P1)T/Tr	independent sweeps at a 0.15 s swept-tuned revisit; 5.6 min for 40 ms every 30 s at 95 percent; one emitter period gap-free
Real-time engine	109/(N D 8) frames/s; POI = 2 N D 8 ns	published relations; 61,035 frames/s and 32.768 µs at N = 2048
Amplitude accuracy	±2.0 dB to 9.5 GHz; ±3.0 dB above	25 °C, 10 min warm-up, spur reject on
Phase noise, 1 GHz at 10 kHz	−107.5 dBc/Hz	ICX-400 family, 25 °C after a 10 minute warm-up; −85.7 at a 40 GHz carrier
Onboard AI compute	up to 117 TOPS (sparse INT8), option 08	ICX-FieldHawk-R only; module rating at up to 40 W; tablet typical 10 to 16 W

Table 8 is the selection guide, and the frequency plan decides most of it. A C-band-only station is covered by the 9.5 GHz tier, because its uplink stops at 6.425 GHz; checking that uplink's second harmonic near 12.85 GHz needs the 40 GHz tier, as does any Ku or Ka station.

Table 8. Selection by station. The tablet is the field instrument and the module is the rack instrument; both run the same engine, so a record from either is comparable with the other.

Station	Model	Options worth ordering	Why
C or X band only, no Ku or Ka plans, harmonic check by other means	ICX-090R	01 OCXO; 71 demodulation	9.5 GHz covers the uplink; the lower tier has the lower published noise level
C band with its harmonic check, Ku, or any site with Ka in its future	ICX-400R	01 OCXO; 71 demodulation; 40 or 41 temperature class for an exposed site; a 6 dB pad for Ka	the uplink or its harmonic lies above 9.5 GHz; the pad answers the looser input match above 16 GHz
Unattended watch at a remote site	ICX-400R or ICX-090R with option 08	08 onboard AI module; a GNSS option for timestamps	the station's own model runs at the antenna and raises a decision with its IQ snippet
Rack monitoring at the L-band IF	ICX-090U module	01 OCXO; the 100 MHz analysis bandwidth option	the IF is below 2.2 GHz and the module runs on the API; the module's standard 50 MHz holds a 36 MHz transponder but not a 72 MHz one, and R1 asks for the whole of it

14Definitions, symbols and acronyms

Amplitude offset and reference plane

A constant the analyzer adds to every point of a trace so the display reads power at a chosen plane, the coupler's main line or the feed flange, instead of at its own connector. The offset carries the coupling factor, any pad and the test-cable loss, and it is part of the record.

Antenna factor

The ratio of field strength at a survey antenna to the voltage it delivers, in dB/m. Adding it and the cable loss to a reading in dBµV gives dBµV/m, which is the unit a coordination limit is written in.

Beam-size correction

The factor by which the sun-method figure of merit is raised to account for the part of the solar disc the main beam does not collect. Near 1 for a small antenna and large for a big one.

Cross-polarization discrimination

The ratio of co-polarized to cross-polarized received field from the same transmission, in dB.

Figure of merit

Receive antenna gain divided by system noise temperature, in decibels per kelvin. The single number that describes a receive station.

Gap-free capture

An acquisition in which every input sample contributes to an output, so no interval exists in which an arriving burst goes unrecorded. A fast sweep still revisits.

Max hold

A trace that keeps the highest value seen in each frequency cell since it was cleared. It proves that something happened, never when.

Probability of intercept

The minimum signal duration for which an analyzer measures the amplitude accurately.

Spectral regrowth

Shoulders raised beside a modulated carrier by amplifier nonlinearity, judged against the operator's mask.

Y-factor

The ratio of noise power with the antenna on a hot source to noise power on cold sky, linear or in dB. The measured quantity in the sun method.

Symbols used in this document.

Symbol	Meaning	Units
E, P, AF, Lc	Field strength; power at the connector; antenna factor; cable loss	dBµV/m; dBm; dB/m; dB
EIRP, Lfs, Latm	Radiated power; free-space path loss; atmospheric loss	dBW; dB; dB
C, N0, G/T, Tsys	Carrier power; noise density; figure of merit; system noise temperature	dBm or dBW; dBm/Hz; dB/K; K
Y, F	Sun to cold-sky noise power ratio; solar flux density	linear; W m−2 Hz−1
k, λ	Boltzmann's constant; wavelength	J/K; m
L, x	Beam-size correction; squared ratio of sun diameter to beamwidth	dimensionless
θs, θhp	Sun's angular diameter; half-power beamwidth	degrees
Ptone, ΔIM3	Tone power; third-order product below the tone	dBm; dB
ℒ(f), φrms	Single-sideband phase noise at offset f; integrated rms phase error	dBc/Hz; rad
ℓ, vf, c, Δf	Distance to a reflection; velocity factor; speed of light; spacing between adjacent ripple peaks	m; dimensionless; m/s; Hz
S, B	Swept span; resolution bandwidth	Hz
τ, Tp	Interferer burst length, period	s
td, Tr, T, P1, Pdet	Dwell per resolution cell; sweep revisit period; observation window; single-sweep catch probability; detection probability over the window	s; s; s; dimensionless; dimensionless
M	Frequency multiplication ratio of the reference	dimensionless
N, D, fframe	Transform size; decimation factor; frame rate	points; dimensionless; frames/s
POI	Probability of intercept, amplitude accurate	s

15Further reading and one route onward

• Recommendations ITU-R S.465 and S.580, the sidelobe envelopes of section 9; S.733, G/T determination for earth stations, for the sun method and its corrections.
• Recommendations ITU-R S.524 and S.728, off-axis radiated power density, with 47 CFR 25.209 and 25.218 for the United States; P.618 and P.838, the rain attenuation model behind the fade margin.
• Report ITU-R S.2199, broadband wireless access against the fixed-satellite service in 3400 to 4200 MHz, for the LNB compression question.
• ETSI EN 301 428 and EN 301 443, harmonized standards for Ku and C band terminals; the ICX-FieldHawk-R datasheet for the values in section 10.
• BNC-AN-102, Wide-area spectrum monitoring with automated classification, and BNC-AN-105, RF record and playback, for the unattended watch and the IQ record.

One route onward

If you are writing an acceptance procedure or preparing an interference case and want the survey duration engineered against the interferer you suspect, or the sun measurement worked for your aperture, our application engineers would be glad to help.