Why your BDA squeals or shuts itself down, and how much antenna isolation you actually need

This is one of the most common calls I get after a building hands over. The off-air booster was commissioned, coverage looked fine on the day, and three weeks later the unit is alarming, cycling its gain down, or putting a howl across the channel. Nobody touched the equipment. What changed was usually the donor signal, or the weather, or someone moved a service antenna during a fit-out. The underlying cause is almost always the same: not enough isolation between the donor antenna and the indoor distribution.

What is actually happening

An off-air bi-directional amplifier is a feedback loop with a very high gain block in it. The donor antenna on the roof picks up the outside signal, the BDA amplifies it by 60 to 80 dB, and the indoor antennas radiate it. Some of that radiated energy leaks back to the donor antenna. If the leakage path attenuates the signal by less than the amplifier's gain, the loop gain exceeds unity and the amplifier oscillates. That is the squeal.

Modern BDAs do not usually squeal for long — they detect the oscillation and throttle gain or shut the branch down. Which is worse, in a way, because the failure is silent. You do not hear a howl, you just have no coverage in the basement and nobody knows why.

The design rule I work to is straightforward: isolation must exceed the BDA's gain by at least 15 dB. A 70 dB gain unit needs 85 dB of isolation, minimum. Some vendors will quote 10 dB and some very conservative specifiers ask for 20 dB. I use 15 dB because 10 dB gives you nothing when the donor signal fades and the automatic level control pushes gain back up, and 20 dB is often unbuildable on a compact rooftop.

Where the isolation comes from

Isolation is not one number you buy, it is three things stacked together.

Physical separation. Free-space path loss at 450 MHz over 25 m is roughly 53 dB. Over 12 m it is only 47 dB. The curve is flat — doubling the distance buys you 6 dB, and no more. Separation alone will never get you to 85 dB on a normal rooftop.

Structure. A reinforced concrete roof slab between the donor antenna and the topmost service antenna is worth 15 to 20 dB, and it is the most reliable contribution in the stack because it does not change with weather or fit-out. Use it. Put the donor on the roof and keep the highest indoor antenna below the slab, not in the lift motor room.

Antenna pattern discrimination. A donor Yagi with a 20 dB front-to-back ratio, aimed at the donor site and pointed away from the building, contributes its full F/B figure if the leakage path is genuinely behind it. Downtilt and a parapet edge help. An omni donor antenna contributes nothing here, which is why I rarely specify one for off-air work.

Stack those on a typical job — 25 m vertical separation, one RC slab, a Yagi with 20 dB F/B — and you get roughly 53 + 18 + 20 = 91 dB. Against a 70 dB BDA that is 21 dB of headroom. That design will be stable. Shorten the separation to 12 m and lose the slab because someone put a service antenna in the roof plant room, and you are at 67 dB against 70 dB of gain. That system oscillates on day one.

Measure it, do not calculate it

The arithmetic above is for design. Before you energise the BDA, measure the real number. Disconnect the amplifier, feed a CW carrier at a known level — I use 0 dBm at the channel frequency — into the donor feeder at the BDA position, and read the level appearing at the service feeder port with a spectrum analyser. Isolation is the difference. Reading 90 dB below 0 dBm means resolving −90 dBm, which needs a narrow resolution bandwidth (300 Hz or less) and often a preamp, so plan for that. Measure uplink and downlink paths separately if the system uses separate donor antennas.

Record the figure in the commissioning report with the BDA's actual set gain next to it. That single line is what lets whoever inherits the system diagnose it in five minutes instead of five site visits.

The failure mode nobody mentions

Even a BDA that is comfortably stable can cause a different problem: uplink noise rise at the donor repeater. A booster with 70 dB of uplink gain amplifies its own thermal noise and radiates it back at the donor site. One unit is usually tolerable. Six of them in a row of buildings along the same street desensitises the repeater and the network operator starts asking questions. Set uplink gain to the minimum that closes the portable-to-repeater path, not to the maximum the box supports. Uplink is very often the limiting path anyway — a 4 W portable behind a body loss of 3 dB has far less to give than a 25 to 40 W repeater — so pushing uplink gain past what the link budget needs buys nothing but noise.

When to stop fighting it

If you cannot get 15 dB above the amplifier gain on the rooftop you have, stop redesigning the isolation and change the architecture. An on-site repeater on the building owner's own licensed channel removes the feedback loop entirely, gives you a stable and predictable downlink source, and takes the donor network out of your risk register. It costs more in hardware and it needs a licence, but it does not fail in August because a neighbouring tower changed tilt. In Singapore, note also that off-air repeating of another party's licensed channel requires the licensee's consent as well as your own IMDA station licence — verify both before you commit the design.

How Suneast handles this

Suneast designs and installs in-building radio coverage systems for buildings in Singapore, primarily UHF DMR and analog two-way radio. On off-air BDA jobs we measure donor-to-service isolation before energising the amplifier and record the figure against the set gain in the commissioning documentation, so the margin is auditable rather than assumed. Where the rooftop cannot deliver the isolation the gain requires, we say so at design stage and price the on-site repeater alternative instead of building a system that will throttle itself six weeks after handover.