Separation, Desense, and the Harmonics Nobody Checks

Introduction

This is the volume that determines the layout. Everything else in this dive is inventory and hardware; this is the physics that says where things can and cannot go, and it contains the one finding that is genuinely surprising: on this particular combination of equipment, every band the ham radio transmits on has a low-order harmonic landing inside a cellular band the gateway is trying to receive. That is not bad luck. It is an arithmetic property of the amateur allocations and the cellular allocations, and it is completely invisible unless someone works it out.

3.1 The two mechanisms, which are often confused

Interference between antennas on one vehicle has two distinct causes, and they call for different fixes.

Front-end overload, also called desense. A receiver’s first amplifier has a finite dynamic range. Present it with a signal enormously larger than anything it was designed for and it stops behaving linearly — it compresses, generates spurious products internally, and its sensitivity to the wanted signal collapses. Crucially, this does not require the interfering signal to be in band. A 50 W transmitter at 440 MHz will desense a cellular receiver at 700 MHz simply by being too loud, because the low-noise amplifier sees it before any filtering that might have rejected it. This is the dominant mechanism, and the only real cure is isolation — distance, orientation, and where necessary filtering.

Harmonic and spurious emission. Every transmitter emits at integer multiples of its operating frequency. A well-designed radio suppresses these substantially, typically 60 dB or more below the carrier, and that is normally sufficient because the harmonic lands somewhere nobody is listening. But 60 dB below 50 W is still 50 microwatts, and 50 microwatts radiated a metre from a receiving antenna trying to resolve a ten-billionth of a watt is not a small signal. When the harmonic lands squarely on the band being received, it is not interference — it is a local transmitter on the wanted frequency.

The first mechanism is unavoidable and is managed by geometry. The second is avoidable only by knowing where the harmonics land, which is why the table below is the most valuable thing in this dive.

3.2 Where the harmonics land

Amateur and GMRS fundamentals against the cellular bands that matter in the rural Upper Peninsula — the low bands, because those are the ones that will actually be carrying the connection.

Table 1 — Where the harmonics land

Transmit band	Harmonic	Falls at	Lands in
2 m, 144–148 MHz	5th	720–740 MHz	LTE B12 downlink (729–746) — AT&T and Verizon 700 MHz
2 m, 144–148 MHz	6th	864–888 MHz	LTE/NR B5/n26 downlink (869–894) — cellular 850
1.25 m, 222–225 MHz	3rd	666–675 MHz	n71 uplink (663–698) — T-Mobile 600 MHz
1.25 m, 222–225 MHz	4th	888–900 MHz	B5 downlink (869–894)
70 cm, 420–450 MHz	2nd	840–900 MHz	B5 uplink and downlink — straight through the cellular 850 band
GMRS, 462/467 MHz	2nd	925–935 MHz	Just above the cellular 850 downlink, in the 900 MHz SMR range

Read the 70 cm row twice. The second harmonic of the 70 cm band covers the entire US cellular 850 allocation. A second harmonic is the least suppressed of all harmonics, 70 cm is the band a mobile tri-bander is most likely to be used on for local repeaters, and the result lands directly on one of the two or three bands most likely to be carrying service in the Keweenaw.

The 1.25 m third harmonic landing in the n71 uplink is the second-worst case, because n71 is T-Mobile’s 600 MHz low band — precisely the allocation identified in the Cellular Gateway dive as the most likely to reach into forested terrain.

GMRS comes out comparatively well: its second harmonic clears the cellular downlink. That is a genuine, if modest, argument for GMRS over 70 cm for routine convoy communication on this particular rig.

None of this means the equipment is faulty or that the combination is unworkable. It means that transmitting and streaming are activities that will compete, that the effect will be worst on exactly the bands that matter most, and that separation is worth real effort rather than being an afterthought.

3.3 How much separation, and in which direction

The conventional guideline is at least a quarter wavelength at the lowest frequency in use. At 2 m that is about 20 inches; the received wisdom in mobile installations is 19 to 20 inches minimum, and more is better. On a 60-inch rack with one long side occupied by an awning, 20 inches is achievable but it is not generous, and it is a floor rather than a target.

The more useful finding is about direction, and it is the single highest-leverage piece of geometry available here:

• Antennas separated horizontally, side by side, gain isolation at roughly 20 dB per decade of distance.
• Antennas separated vertically, one above the other, gain isolation at roughly 40 dB per decade — because a vertical antenna’s pattern has a deep null straight up and straight down, so stacked antennas present their nulls to each other.

The practical rule that falls out of this is dramatic: each foot of vertical separation is worth roughly ten feet of horizontal separation.

For a rack layout this reframes the whole problem. The instinct is to spread antennas across the available deck, which is the weak axis. The far better move is to exploit height differences — put the cellular array low and flat on the rack deck, and the transmitting whip up on a mast or an upright well above it. A tri-band whip whose feedpoint sits two feet above a low-profile cellular panel has, in isolation terms, roughly the equivalent of twenty feet of horizontal separation, which is more than the entire truck could provide.

This also happens to be the arrangement that is best for everything else. The whip wants height for its own performance; the cellular panel is low-profile by design and gains little from being high; and the tall thing being removable for garage clearance is exactly the item that should be on top.

3.4 Orientation and polarization

Two smaller effects worth using, because they are free:

Cross-polarization buys 15 to 20 dB of isolation between two antennas whose elements are at right angles. Most cellular MIMO panels already use cross-polarized elements internally for exactly this reason. A vertical ham whip beside a panel with slant-polarized elements is already partly decoupled; this is a reason to prefer a panel over four vertical whips.

Pattern nulls work in the horizontal plane too, though far less strongly than vertically. Where a directional or panel antenna is used, aiming its back toward the transmitter rather than its face is worth doing when nothing else costs anything.

3.5 What to do when geometry is not enough

In descending order of preference:

1. More separation, preferentially vertical. Free, and always the first move.
2. Operational discipline. Do not transmit at 50 W on 70 cm while relying on a marginal cellular link. This sounds like a cop-out and is in fact the correct answer most of the time — the two activities are rarely both urgent, and a person who understands why will simply not do both at once.
3. Reduce power. A mobile rig at 5 or 10 W into a repeater a few miles away works as well as one at 50 W, and the interference scales directly. Most mobile transmissions do not need full power, and running lower by default is the cheapest single improvement available.
4. Filtering. A band-pass filter on the transmitter output, or a notch at the offending harmonic, addresses the harmonic mechanism directly. This is real engineering with real insertion loss and real cost, and it should be a response to a measured problem rather than a precaution.
5. Accept it. Some desense while transmitting is normal in every multi-service vehicle installation, including professional ones. The goal is that the cellular link recovers when the key is released, not that it is unaffected.

3.6 Measuring rather than guessing

The Cellular Gateway dive noted that the GL.iNet interface exposes live RSRP, RSRQ, and SINR for the serving cell. That turns this entire volume from theory into a fifteen-minute experiment, and it should be run before the layout is called finished:

1. Park somewhere with a moderate, stable cellular signal — not strong, because a strong signal hides the problem, and not absent, because there is then nothing to degrade.
2. Record RSRP and SINR with nothing transmitting.
3. Key the ham rig on each band in turn, at full power, into a real load, and record the readings again while transmitting.
4. Repeat at reduced power.
5. Repeat after moving the whip higher.

The result is a small table specific to this truck that settles arguments no amount of reading can. SINR is the number to watch — desense shows up there before it shows up in RSRP, because RSRP measures the wanted signal and SINR measures whether it can be heard over everything else.

The nanovna, analyzers-vnas, and power-swr-field-strength dives in the Antennas project cover the complementary measurements — that each antenna is actually resonant and matched where it is mounted, which on a rack with an improvised ground plane should never be assumed.

3.7 A safety note that is not optional

Several of these are transmitters at meaningful power, mounted at head height beside where people stand, sit, and sleep. The FCC publishes guidance specifically on human exposure to RF fields from vehicle-mounted antennas, and it exists because vehicle installations put radiating elements unusually close to people.

The practical points for this build: the HF setup at 100 W is the one that warrants real attention; nobody should be standing beside a transmitting whip at arm’s length; and an antenna mounted on a bed rack is at roughly the height of a person’s head when they are standing next to the truck, which is worse than a roof mount rather than better. This is another argument for the HF antenna living at the rear bumper or, better, in a tree at camp.