How do you measure a building's RF noise floor, and what reading means the site needs filtering rather than more signal?

You measure it by comparing the receiver's effective sensitivity with the antenna connected against the same receiver terminated into a 50-ohm dummy load. The difference, in dB, is what the building's own noise is costing you. At 400-470 MHz a degradation of 3 dB or less is normal and the site is thermally limited, so more signal helps. Between 3 and 10 dB, filtering and hunting the source recovers more usable range than any amount of extra gain. Above 10 dB the site is externally noise-limited, and adding a preamplifier or a higher-power repeater makes things worse, not better.

That last threshold is the one that gets missed, and it is why buildings buy the wrong fix.

What should the noise floor read in the first place?

Start from thermal noise, because every other reading is judged against it. Thermal noise is −174 dBm per hertz at room temperature, and a 12.5 kHz DMR channel is 41 dB wider than a hertz, so the thermal floor in that channel is about −133 dBm. A commercial UHF receiver adds a noise figure of typically 6 to 8 dB, putting its effective noise floor at roughly −125 dBm. A published 12 dB SINAD sensitivity of −119 dBm then makes sense: it sits about 6 dB above that.

A genuinely quiet site, measured in-band at the repeater's receive port, should therefore read close to −125 dBm. I have measured plant rooms in occupied Singapore buildings at −105 dBm in the same bandwidth — 20 dB of noise the design never budgeted for, deducted from the uplink before the first portable keys up.

How do you measure it on site?

The substitution method is the one that matters commercially, because it reports the answer directly in dB of lost sensitivity. Inject a signal generator into the receiver through a lossy tee, find the level that produces 12 dB SINAD with the antenna connected, then repeat with the antenna replaced by a 50-ohm dummy load. The difference is the site noise degradation. It needs no assumptions about analyser settings and it is defensible at handover, because it is measured on the receiver actually in service.

A spectrum analyser then tells you what the noise is, which is what you need in order to fix it. Set resolution bandwidth to the channel bandwidth and read the in-band floor with your own transmitter keyed and unkeyed. If the floor rises when you key, that is self-desense — an isolation or duplexer problem, not the building. If it is elevated with your transmitter off, walk the analyser and a directional antenna until the level peaks. The source is almost always the building's own services: LED drivers, variable-frequency drives on lift motors and air-handling units, or basement EV chargers.

Hard numbers worth keeping to hand:

• Thermal noise in a 12.5 kHz channel: about −133 dBm (−174 dBm/Hz plus 41 dB)

• Effective receiver noise floor with a 6-8 dB noise figure: about −125 dBm

• Typical sensitivity at 12 dB SINAD: −119 dBm (verify against the datasheet)

• Acceptable site noise degradation at UHF: 3 dB or less

• Repeater self-desense from its own transmitter: under 2 dB

• Design threshold for reliable DMR voice at the portable: −95 dBm, with 10 dB fade margin commercially and 15 dB where the system is mission-critical

What reading means filtering rather than more signal?

Below 3 dB of degradation the receiver is still hearing its own thermal noise, not the building's. A low-noise preamplifier or a receiver multicoupler genuinely improves the uplink here, because the limiting noise is internal.

Between 3 and 10 dB, gain is a poor investment. A preamplifier raises the wanted signal and the interference by the same amount, so the signal-to-noise ratio at the demodulator is unchanged — you have made the meter move without making the radio work. What helps is a cavity or band-pass filter ahead of the receiver if the source is out of band, and finding and fixing the source if it is in band.

Above 10 dB the site is externally noise-limited and gain is actively harmful: the preamplifier amplifies strong nearby signals too, and an overloaded front end generates intermodulation products on frequencies where nothing is transmitting. At this point the correct answer stops being a receiver-side fix at all. If you cannot lower the noise, you must raise the wanted signal relative to it, and the only way to do that is to put the receive antenna inside the area you are trying to cover — which is a distributed antenna system. This is a two-way radio conclusion, not a cellular one; most published repeater-versus-DAS comparisons are written about cellular boosters and do not transfer. We set out the receiver-side options in multicoupler and preamplifier versus a DAS.

What does an elevated noise floor cost in coverage?

Put it in floors, because that is how a building owner buys it. A reinforced-concrete floor slab costs 15 to 25 dB, an RC core or lift shaft 25 to 35 dB. If site noise has degraded the repeater's effective sensitivity from −119 dBm to −104 dBm, you have given away 15 dB of uplink — roughly one basement slab. A guard who could talk from B2 now cannot talk from B1, and nothing about the radios, repeater or antennas has changed. It is also why a downlink-only survey passes a system that fails in service: the portable hears the repeater perfectly, because the downlink never sees the building's noise. We cover that failure mode in why your repeater cannot hear your portables.

Related questions

Should a noise floor measurement be part of the acceptance test? It should. Record the degradation in dB at handover alongside the coverage grid, so that when the building adds plant in three years there is a baseline to argue from.

Does a DAS fix a noise problem or just move it? A DAS does not lower the noise. It raises the wanted signal by putting the antenna inside the coverage area, so the signal-to-noise ratio improves because the path loss to the user collapses.

Can a system pass a −95 dBm survey and still fail on noise? Yes, routinely. The survey measures downlink level and says nothing about what the repeater hears.

How Suneast handles this

We measure the noise floor before quoting hardware, not after the system underperforms. The substitution measurement takes under an hour at the repeater and tells us whether the building needs filtering, a receiver-side fix, or a distributed antenna system — three solutions at very different prices. Where the source is the building's own plant we say so, because a filter costing a few hundred dollars sometimes does what a DAS quotation was asked to do. Where the site is genuinely externally noise-limited, we design the DAS, run both uplink and downlink budgets, and handle the IMDA licensing for the repeater.