70 cm Repeater Desense: Isolation, Filtering and Site Tests

Two 70 cm systems can meet their individual specifications and still fail when installed together. A 50 W transmitter only 125–500 kHz from a neighbouring receiver is not a normal weak-signal situation: the wanted signal may be near the receiver threshold while the unwanted signal at its input is millions or billions of times stronger. The cure is not one magic number or one magic box. It is a measured isolation budget.

Begin With the Level at the Victim Receiver

At 50 W, transmitter power is approximately +47 dBm. If the complete path from the offending transmitter output to the victim receiver input has 62 dB of isolation, the coupled carrier is about −15 dBm before any additional feeder or filter loss:

That arithmetic is exact enough for planning, but the 62 dB must come from a defensible measurement or model. It cannot be inferred from antenna spacing alone and then treated as a guaranteed site value. A receiver may be expected to recover a signal around −110 or −120 dBm in a narrow channel, but such figures are meaningless without the modulation, bandwidth and criterion—SINAD, quieting, BER, packet success or another stated threshold.

The important quantity is the unwanted level at the receiver reference plane, not merely transmitter output power or free-space distance. Every item in the path matters: transmitter feeder, antenna coupling, support structure, victim antenna, receiver feeder, multicoupler, cavity filters, connectors and any unintended conductive route.

What “Desense” Can Mean

Desensitization is the observable loss of wanted-signal performance when the unwanted transmitter is active. It is an outcome, not a single circuit mechanism. At a shared site, investigate at least these cases:

ETSI EN 300 086 defines adjacent-channel selectivity through a controlled wanted-plus-unwanted measurement and a stated degradation criterion. Blocking, intermodulation and transmitter unwanted emissions are separate tests. That distinction matters: a radio that passes one laboratory number does not automatically pass every co-site scenario.

Why a Duplexer May Not Protect the Neighbour

A duplexer is designed around its own transmitter and receiver frequencies. It provides transmitter-to-receiver isolation for that duplex pair while allowing both branches to share one antenna. It is not automatically a narrow guard filter for a second system whose carrier sits only a few hundred kilohertz from the first receiver.

If the neighbouring transmitter lies inside the receive branch’s useful passband, that branch may provide little attenuation. Conversely, the transmit branch can suppress some out-of-band transmitter energy only according to its actual transfer curve at the victim frequency. “There is a duplexer” is therefore not a protection specification. Record the insertion loss and rejection of the complete assembled network at every relevant frequency, including connector and jumper effects.

Antenna Separation: Useful, but Measure It

Vertical stacking of omnidirectional antennas is often valuable because each antenna can sit near a null in the other antenna’s vertical pattern. Unlike a frequency-selective filter, physical isolation attenuates the carrier, transmitter noise and mixing drive across the coupled path. It is therefore usually the first and most broadband part of the solution.

A familiar planning estimate for idealized vertically separated dipoles is:

At about 435 MHz, λ is roughly 0.69 m; inserting d = 5 m gives approximately 62 dB. This is a useful order-of-magnitude comparison for the assumed geometry. It is not a universal law for two unspecified “3 dBd” collinear antennas. Real isolation depends on each antenna’s vertical pattern and electrical length, overlap of active regions, mounting brackets, mast currents, feeder routing, mutual coupling, tower steel, nearby roofs and reflections.

Horizontal spacing is sometimes compared with the far-field free-space equation:

Using two 3 dBd antennas—about 5.15 dBi each—and solving for roughly 62 dB of net isolation produces a distance around 230 m. That is a free-space, far-field, unobstructed boresight comparison. It is not a tower-site prediction, and it does not prove that 5 m vertically is always equivalent to hundreds of metres horizontally. Near structures and long collinears, measure S₂₁ or received power in the installed configuration.

ITU-R Report M.2141 specifically addresses isolation between nearby VHF land-mobile antennas. Broader ITU co-location studies likewise show strong dependence on antenna pattern, height, azimuth, downtilt and site geometry. The engineering lesson is simple: use formulas to choose a candidate arrangement, then verify it at the actual site.

Filtering: Put Selectivity Before the Vulnerable Stage

If a strong unwanted signal overloads the first amplifier or mixer, selectivity later in the receiver cannot undo the damage. A receive preselector must therefore sit ahead of the stage that is being compressed. At 125 kHz separation around 435 MHz, the ratio f/Δf is about 3480. That number illustrates how demanding the frequency discrimination is, but it is not the required resonator Q by itself.

Achievable rejection and insertion loss depend on filter order, resonator unloaded Q, loaded Q, coupling, bandwidth, temperature drift and tuning accuracy. A single very high-Q cavity may not provide the necessary skirt; several coupled resonators may be required. Extra cavities also add receive-path loss, and every decibel ahead of the receiver raises system noise figure when external noise is low.

For the transmitter, measure both the carrier and the noise density at the victim frequency. A cavity or dedicated low-noise transmit filter can suppress out-of-band energy, but only if its skirt provides rejection at that precise offset and can handle the power, voltage, temperature and duty cycle. A clean spectrum-analyzer trace also requires a test setup whose own dynamic range and phase noise are better than the quantity being measured.

What a Circulator or Isolator Can—and Cannot—Do

A circulator routes forward and reverse energy between ports; with a suitable load it can present a transmitter with a more controlled termination and can reduce intermodulation generated inside a transmitter by incoming site signals. It is not a frequency-selective filter and does not, by itself, remove broadband noise already generated by that transmitter. It belongs in a system design with the correct load, cavities, power rating and fault analysis.

Build a Pair-by-Pair Isolation Budget

For every transmitter A and receiver B, document the path rather than quoting one “site isolation” number:

• TX-path filtering: loss or rejection from transmitter output to transmit antenna.
• Antenna isolation: installed coupling between the two antenna ports, with normal feeders and structures connected.
• RX-path filtering: attenuation from victim antenna to receiver input at the unwanted frequency.
• Margin: allowance for temperature, tuning tolerance, aging, additional transmitters and changing site configuration.

Do not simply add catalog maxima obtained under unrelated conditions. Measure at consistent reference planes with calibrated cables, attenuators and fixtures. Confirm that no test instrument is being overloaded and that leakage between bench cables is not the apparent isolation limit.

A Field-Test Sequence That Identifies the Cause

1. Define the wanted criterion. Feed a calibrated wanted signal into the receiver and choose a repeatable endpoint: 12 dB SINAD, stated quieting, BER, packet success or another service-appropriate metric.
2. Measure baseline sensitivity. Record the wanted level needed to meet that criterion with the site transmitters off.
3. Apply the unwanted signal. Operate the suspected transmitter in its real mode and duty cycle, initially at a controlled level if necessary, while protecting generators, analyzers and receivers with adequate isolation.
4. Measure degradation. Increase the wanted signal until the original criterion returns. The increase in decibels is the desensitization for that condition.
5. Separate carrier blocking from transmitter noise. Repeat with a laboratory source whose noise at the victim frequency is demonstrably lower than the transmitter’s. If the real transmitter is worse at the same coupled carrier level, source noise or transmitter-created products are implicated.
6. Test selective remedies one at a time. Add receive preselection, transmit cleanup or extra antenna isolation and record what changes. A remedy that changes nothing is evidence, not a reason to guess harder.
7. Exercise the whole site. Test all relevant transmitter combinations, power levels, channels and antenna states. Two-tone products may appear only when another service transmits.
8. Recheck over time. Temperature, moisture, loose hardware, corrosion and cavity drift can alter a previously clean site.

A Practical Design Order

A robust installation normally develops in this order:

1. Choose a frequency plan with as much separation as the coordinated service permits.
2. Use transmitters with low noise and receivers with verified close-in blocking, selectivity and phase-noise performance.
3. Maximize measured antenna isolation through position, pattern, polarization and feedline routing.
4. Add receive preselection before the vulnerable active stage.
5. Add transmitter cleanup where measured noise or products require it.
6. Control incoming signals and reverse energy with circulators, isolators or combiners where the full system analysis supports them.
7. Validate every pair and relevant multi-transmitter combination at operating power.

A well-designed vertical stack may contribute tens of decibels, but the correct acceptance criterion is the measured end-to-end performance. Conversely, even excellent cavities cannot compensate for a noisy transmitter whose energy already falls inside the victim passband, a nonlinear rusty junction on the tower or a receiver whose first stage is driven beyond its linear range.

The Practical Verdict

Repeater desense at a shared 70 cm site is not solved by declaring that five metres gives 62 dB, by pointing to a duplexer or by buying the highest-Q cavity available. The system must suppress two different quantities: the unwanted carrier at the receiver and the offending transmitter’s energy at the wanted frequency. It must also avoid creating new products in receivers, transmitters and passive junctions.

Questions, antenna-factor records or height trials to share? Contact RF.Guru.