Why can our staff hear the repeater clearly, but the repeater cannot hear them from inside the building?

Because the two directions are not equally strong. A 40 W repeater feeding a rooftop antenna radiates roughly 47 dBm EIRP, while a 4 W portable held at the hip radiates about 31 dBm - a 16 dB deficit before you account for anything else. Add rooftop receiver noise and feeder loss on the uplink and the real imbalance is commonly 25 to 30 dB. Your staff hear the repeater because the downlink has power to spare; the repeater cannot hear them because the uplink ran out of budget several floors ago.

How big is the imbalance, exactly?

Work the two paths separately, because they are two different radio links that happen to share a building.

Downlink: the repeater transmits 40 W (+46 dBm), loses about 2 dB in the duplexer, 1.8 dB in 40 m of 1/2 inch coax at 450 MHz and 1 dB in jumpers and connectors, then gains 6 dBi at a rooftop collinear. Net EIRP is about +47 dBm.

Uplink: the portable transmits 4 W (+36 dBm) into a helical antenna with roughly -2 dBi gain, then loses about 3 dB to body absorption when it is worn on a belt or held at the chest. Net radiated power is about +31 dBm.

• Raw TX power difference (40 W vs 4 W): 10 dB

• Antenna gain difference (+6 dBi rooftop vs -2 dBi helical): 8 dB

• Body loss, uplink only, portable at hip: 3 dB

• Typical rooftop receiver desensitisation from site noise: 10-20 dB

• Design threshold for reliable DMR voice: -95 dBm at the portable

• Fade margin: 10 dB minimum commercial, 15 dB minimum mission-critical

With an indoor path-loss exponent of about 3.5, a 20 dB one-way deficit does not cost you 20 per cent of your range - it cuts the usable talk-in distance to roughly a third of the talk-out distance. That is why a guard can stand in a basement lift lobby, hear the console perfectly, key up, and get nothing back.

Why the repeater's receiver is the weakest link

A portable sitting in a corridor is a reasonably quiet receiver. A rooftop repeater receiver is not. It sits next to LED floodlight drivers, lift machine-room variable-frequency drives, telco cabinets and other people's antennas. Thermal noise in a 12.5 kHz DMR channel is about -133 dBm, and a receiver with a 6 dB noise figure sees an effective floor near -127 dBm, which is where the -119 dBm sensitivity figure on the datasheet comes from. Rooftop site noise routinely lifts that floor by 10 to 20 dB, so the receiver specified at -119 dBm is actually hearing nothing below -105 dBm.

Feeder loss compounds it. Every dB of coax loss between the antenna and the receiver comes straight off the uplink sensitivity, and unlike the downlink there is no transmitter power to absorb it. A 60 m run of 1/2 inch cable costs 2.7 dB in both directions; the downlink shrugs it off and the uplink does not.

Will turning the repeater up fix it?

No, and it will make the diagnosis harder. Going from 40 W to 100 W adds 4 dB to the downlink and exactly nothing to the uplink. Your staff will hear the repeater more clearly in more places, keep pressing PTT in those places, and keep not getting through - while the coverage map looks better than the system actually performs. The same applies to fitting a higher-gain rooftop antenna: passive gain does help both directions, but only if the pattern still illuminates the floors you care about, and a higher-gain collinear has a narrower vertical beamwidth that often makes the lower floors worse.

What actually fixes an uplink-limited system

Fix the receive path first. A receiver multicoupler with a low-noise preamplifier at the antenna recovers most of the feeder loss and typically buys back 3 to 6 dB. Then attack the noise floor itself - LED drivers and VFDs are the usual offenders in Singapore buildings, and a spectrum sweep at the rooftop before and after isolating circuits will tell you within an hour whether you are fighting a noise problem or a coverage problem.

We covered that diagnosis in detail in our post on radio interference from LED lighting and variable-frequency drives.

If the noise floor is clean and the uplink still fails, the geometry is wrong and no amount of receiver improvement will save it. Distributed antennas are the only fix that works symmetrically: bringing the antenna to within 15 to 25 m of the user cuts path loss identically in both directions, so the uplink improves by exactly as much as the downlink. Satellite receivers with a voting comparator are the alternative where a full DAS is not viable - several receive-only sites, each closer to the users than the main repeater, with the comparator selecting the best copy.

Related questions

How do we prove the problem is the uplink and not the handsets? Run the grid survey in both directions. Measure downlink received signal at the portable and, separately, have the portable transmit from each grid point while a monitor at the repeater records the received level. A floor that passes outbound at -80 dBm and fails inbound is uplink-limited, definitively. Our grid survey and acceptance testing guide sets out the method.

Does Singapore require the test to be bidirectional? Singapore has no IFC 510-equivalent code mandating it, so it comes down to what your specification says. US practice under IFC 510 requires both inbound and outbound signal levels to meet -95 dBm, and any competent Singapore specification should copy that requirement - a one-direction acceptance test proves nothing about talk-back.

Would higher-power portables solve it? Marginally. Moving from 4 W to 5 W buys about 1 dB, which is inside measurement error. A body-worn remote speaker microphone with the antenna at the shoulder recovers more - typically 3 to 5 dB of the body loss - for far less money, and it is worth doing regardless.

How Suneast approaches this

We survey both paths before quoting. Every site survey we run in Singapore records inbound and outbound levels at the same grid points, plus a rooftop noise-floor sweep across the licensed channel, so the report says which direction is limiting and by how many dB. Where the uplink is the constraint we specify the receive-path fix, the antenna distribution, or both, and we state the resulting margin against the -95 dBm threshold rather than a coverage percentage on its own. Designs are built around a 10 dB commercial or 15 dB mission-critical fade margin, and IMDA localised private network licensing for the repeater is handled as part of the works.