> ## Content Index
> Fetch the complete content index at: https://abovethefader.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Two antennas and a compressor you can't turn off: how a wireless mic actually works
- URL: https://abovethefader.com/two-antennas-and-a-compressor-you-cant-turn-off-how-a-wireless-mic-actually-works/
- Published: 2026-11-01T18:32:27.000Z
- Updated: 2026-11-01T18:32:27.000Z
- Description: Your wireless system is an FM radio, a 2:1 compressor, a dice game with reflections, and a gate, all pretending to be a cable. Here's what each piece is doing, with the math, so the next dropout has a cause instead of a ghost.
- Author: Tal Kocen
- Tags: The Woodshed

*Your wireless system is an FM radio, a 2:1 compressor, a dice game with reflections, and a gate, all pretending to be a cable. Here's what each piece is doing, with the math, so the next dropout has a cause instead of a ghost.*

It's the second hour of a corporate keynote in a hotel ballroom, and the CEO has discovered the downstage-left corner. He likes it there. It's near the big LED wall, it lets him point at the slide, and every time he plants himself on that particular square of carpet, his lav goes from a person to a cough to nothing. Then he takes one step toward center and he's back, full and clean, as if nothing happened.

You've already done the obvious things. Fresh batteries an hour ago. The frequency scanned clean at load-in. The receiver's RF meter, which you are now staring at with the expression of a man reading a medical chart, shows both diversity LEDs flickering like a nervous tic whenever he's in the corner. The video engineer says it's the LED wall. The LED tech says it's not the LED wall. The client says nothing but looks at you in a way that means something.

Here's the thing nobody at that table will say out loud: your wireless system is not a cable. It's an FM radio station with a compressor bolted to the front of it, an expander bolted to the back of the receiver, a pair of antennas playing a probability game against the room, and a gate deciding when to let any of it through. Every one of those parts has a reason to exist and a way to fail. The corner of that stage is a specific failure with a specific cause, and once you know the parts, you can usually name it in about thirty seconds.

So let's walk the path from the capsule to the XLR out. There will be math, all of it the kind you can do in your head at FOH.

## The signal chain nobody draws

Most system drawings show a wireless mic as a box with a lightning bolt. Draw it properly, because the order matters. Inside an analog transmitter, the capsule feeds a preamp. The preamp feeds a pre-emphasis network that boosts the top end. That feeds a compressor, usually at a 2:1 ratio. The compressed, treble-boosted audio then wiggles the frequency of a carrier oscillator up and down around its center frequency. That's the FM. A small amplifier pushes the result, somewhere around 10 to 50 milliwatts, into an antenna.

At the receiver the order reverses, and the reversal has to be exact. Antenna, RF front end, filtering, a demodulator that turns frequency wiggles back into voltage, an expander that undoes the compressor, a de-emphasis network that undoes the treble boost, and finally a squelch circuit that decides whether any of this should reach the output at all. If any mirror is imperfect, you hear it.

Start with the radio part, because it sets every limit downstream. In FM, the audio's amplitude controls how far the carrier swings away from its center frequency. That swing is called deviation. Sennheiser's ew 500 G4 spec sheet, to pick a widely used analog system, lists a nominal deviation of ±24 kHz and a peak deviation of ±48 kHz. The receiver only cares how far the frequency moved, not how strong the carrier is, which is why FM shrugs off amplitude noise. It's also why there's a hard ceiling: past peak deviation the signal spills out of its channel, so the transmitter has to limit.

How much spectrum does that take up? The working engineer's tool is Carson's rule: bandwidth is roughly twice the sum of the peak deviation and the highest audio frequency. Written as the textbooks do, B ≈ 2(Δf + fm). With 48 kHz of peak deviation and audio out to about 18 kHz, that's 2 × (48 + 18) = 132 kHz. RF Venue's primer puts the practical figure for high-quality wireless audio at roughly 150 kHz and notes the FCC caps analog wideband FM wireless at ±100 kHz of deviation, a 200 kHz total. So one channel is a stripe 130 to 200 kHz wide, and every bit of quality is a trade between how wide it swings and how many neighbors fit in the band.

That trade is the root of everything that follows. Broadcast FM sounds great because of strong signals and a high modulation index. A wireless mic is a tiny transmitter crammed into a narrow stripe, sharing the stage with thirty others. AKG's white paper on its WMS61 system puts the usable signal-to-noise ratio of a link like this at about 60 dB and says flatly that isn't good enough for critical sound reinforcement. So the designers bolt on tools to cheat that 60 dB into something better.

## FM is quiet until it isn't

If you've ever walked a handheld out to the edge of range, you've heard the FM personality in full. It's clean, clean, clean, then there's a faint hiss, then a few clicks and spits, then it falls off a cliff. That shape comes straight from how an FM demodulator behaves.

The first piece is the threshold. Above a certain carrier-to-noise ratio at the receiver input, FM delivers a big improvement in audio signal-to-noise compared to what you'd get from the raw RF. Below it, the improvement collapses. The textbook numbers depend on whose textbook you open: All About Circuits' worked examples note that Haykin's *Communication Systems* puts the threshold around a carrier-to-noise ratio of 20, which is 13 dB, where occasional clicks appear, while Proakis recommends a less conservative 10 dB. Either way, the takeaway for the stage is the same. Above about 10 to 13 dB of carrier over noise, you're fine. A few dB below that, you're not, and there isn't much in between.

The clicks have a nice physical explanation in PA3FWM's note on FM noise. When the carrier and the noise are about the same size, a noise spike pointing the other way can swing the combined signal all the way around zero. The demodulator reads that as a huge instantaneous frequency jump and spits out a pop. As the carrier weakens, the pops multiply until they merge into hash. That spitting right before a dropout often isn't interference at all. It's a weak carrier falling off the threshold.

The second piece is capture. An FM receiver locks onto the strongest signal on its frequency and largely ignores weaker ones. RF Venue's primer puts the required margin at about 3 to 4 dB. That's wonderful when your transmitter is the strongest thing in the room and miserable when it isn't. If a rogue signal, an intermod product, or a neighbor's channel is within a few dB of your carrier in some corner of the stage, the receiver flips between them. That sounds like garbled audio, sometimes with a second voice bleeding through, and it's a different fault from a fade. When the RF meter says "strong signal" but the audio is trash, think capture before you think dropout.

The third piece is where the noise lives. FM hiss is not white. The noise coming out of an FM demodulator has a triangular spectrum: it rises with frequency. The demodulator measures frequency, which is the rate of change of phase, so slow phase wobbles from noise produce small errors and fast ones produce big errors. So the noise voltage at the output climbs with audio frequency, about 6 dB per octave. Most of the hiss you hear from an FM link lives in the top two octaves.

Why FM hiss is a treble problemDetector noise rises 6 dB/octave; a 75 µs de-emphasis (corner 2.1 kHz) cuts it back3020100\-10\-20\-30\-40\-5020501002005001k2k5k10k20kAudio frequency (Hz)Relative level (dB)2,122 Hzraw detector noise (+6 dB/oct)75 µs de-emphasiswhat you hear

Fig. 1: The raw FM detector noise (dashed) rises at 6 dB per octave. A 75 µs de-emphasis network rolls off above 2,122 Hz and flattens the noise you actually hear (teal). Pre-emphasis at the transmitter boosts the program by the same curve, so the audio comes out flat and the hiss doesn't.

That's what pre-emphasis and de-emphasis are for. The transmitter boosts the highs before modulation, and the receiver cuts them by the same amount after demodulation. The program comes out flat. The noise, which was added in between, only gets the cut. The classic broadcast time constant is 75 microseconds, which puts the corner at 1/(2π × 75 µs) ≈ 2,122 Hz, with a 6 dB per octave slope past it. Wireless manufacturers choose their own curves, and they don't usually publish them, but the principle is identical: you trade some of your high-frequency headroom for a quieter top end.

That trade has a consequence you've already met. Pre-emphasis means the transmitter hits peak deviation sooner on bright material. A sibilant vocalist or a tambourine on a handheld will run into the transmitter's limiter at a lower apparent level than a baritone saying "good evening." When someone says a particular wireless "sounds dull on cymbals" or "gets crunchy on Ss," that can be the limiter catching boosted treble, not the capsule.

Now the arithmetic, because it ties these pieces together. The thermal noise floor of anything at room temperature is −174 dBm per hertz of bandwidth. That's kTB, Boltzmann's constant times temperature times bandwidth, and RF Essentials lays out the standard form: noise floor in dBm = −174 + 10·log₁₀(B) + noise figure. For a 200 kHz channel, 10·log₁₀(200,000) is 53 dB, so the floor is −121 dBm before the receiver adds anything of its own. Give the front end a noise figure of about 5 dB and you're at roughly −116 dBm. Add 13 dB for the FM threshold and you land at about −103 dBm.

Now look at the Sennheiser spec sheet. It quotes receiver sensitivity as under 1.6 µV for 52 dBA of effective audio SNR with HDX. Treat that as volts across a 50-ohm input and the power is (1.6 × 10⁻⁶)² ÷ 50, which is about 5.1 × 10⁻¹⁴ watts, or −103 dBm. The noise figure in my version is an assumption, not a published number, but the point stands: the spec sheet isn't magic. It's kTB, plus a reasonable front end, plus the FM threshold. When you know that, the RF meter on your receiver stops being a vague bar graph and becomes a distance from a cliff.

## The compressor you can't turn off

Here's where most of the "sounds like a wireless" quality comes from. To squeeze a 100-plus dB program through about 60 dB of usable SNR, analog systems use companding. Per Shure's explainer, lower-tier systems compress at a fixed 2:1 at the transmitter and expand at 1:2 in the receiver to restore the original dynamics. Shure's companion piece on its Audio Reference Companding says pro wireless needs companding to get dynamic range above 100 dB.

It looks like a free lunch. A 2:1 compressor turns a 100 dB input range into 50 dB on the air, which fits above a 60 dB noise floor, and the expander stretches it back to 100.

But look at what happens to the noise. The expander doesn't know what's signal and what's noise. It sets its gain from the level it detects, and that gain applies to everything that arrived over the air, including the channel hiss. When the program is loud, the expander is near unity and the noise stays 60 dB down. When the program gets quiet, the expander turns its gain down hard, and it drags the noise down with it. The noise floor rides up and down with the music.

You can do the math in your head. Call the input level L in dB below full scale. After a 2:1 compressor, it's on the air at L/2\. Channel noise sits at −60 dB on the air regardless. The expander restores L/2 to L, which means it's applying a gain of L/2\. That same gain hits the noise, so the noise comes out at −60 + L/2\. The signal-to-noise ratio at any moment is L − (−60 + L/2), which simplifies to 60 + L/2.

Plug in some real moments. A vocal peak at −10 dB: SNR is 55 dB, and the music masks the hiss completely. A quiet spoken passage at −40 dB: SNR is 40 dB. A soft reverb tail decaying through −90 dB: SNR is 15 dB, and the noise is sitting right under the tail, following it down.

Companding: the noise floor follows the music2:1 compressor into an FM link with 60 dB channel SNR, 1:2 expander at the receiver0\-20\-40\-60\-80\-100\-120\-100\-80\-60\-40\-200Input level at transmitter (dB re full scale)Level at receiver output (dB)40 dB SNR at a -40 dB passage15 dB at -90signalnoise, no companding (-60 flat)noise with companding\= input/2 - 60

Fig. 2: With 2:1 companding over a link with 60 dB of channel SNR, the output noise (rust) equals input/2 − 60 dB. The noise falls as the signal falls, which is why the measured "dynamic range" looks huge in silence, but at any moment the hiss sits only 60 + L/2 dB below the program.

That's the breathing. Shure's explainer names it: with fixed-ratio companding, the noise floor becomes audible at low levels as breathing or pumping. You'll hear it most on exactly the material you'd expect from the math: sparse acoustic guitar, a solo voice with long gaps, a lav on a presenter who pauses for effect, any slow decay. The dynamic range spec, by the way, is measured with no signal at all, where the expander has the noise pushed to the floor. It's honest. It's just measuring the one moment that the breathing can't happen.

This is also why the higher-end analog systems moved to level-dependent schemes. Shure's Audio Reference Companding only compands at high audio levels, where the artifacts are masked, and leaves the quiet stuff alone. Shure says this avoids the low-level breathing of fixed-ratio designs. Sennheiser has its own HDX compander on the G4 series. The details are proprietary, but the idea is the same everywhere: stop running the expander hard on the quiet material where you can hear it work.

Two practical consequences. First, the compander and the expander have to be mirror images. If they're not, the dynamics come back wrong. That's one more reason an analog transmitter from one brand, or even one product family, isn't a drop-in match for another's receiver, even on the same frequency. The FM part might demodulate fine. The expander will put the dynamics back in the wrong shape.

Second, the transmitter gain knob matters more than people think. The compressor and the transmitter's limiter are working on whatever level you send them. Too hot, and the limiter is clipping deviation on every peak, which sounds like grit on the attack. Too cold, and you're living in the low end of the curve, where the expander is working hardest and the noise is closest to the program. Set the transmitter gain so the peaks hit the receiver's audio meter near the top on the loudest thing the talent will actually do, then leave it.

## Half a wavelength from silence

Back to the CEO in the corner. His signal isn't too weak on average. Fifty meters of ballroom is nothing for a 10 mW transmitter, as we'll see. What's happening is that his signal is arriving at the receive antenna by more than one path, and in that particular spot, the paths are cancelling.

Radio waves bounce off anything conductive, and a hotel ballroom is a mirror box at UHF: ceiling steel, truss, the LED wall's metal frame. At the antenna, the direct signal and the reflected copies add together as vectors. If they arrive in phase, you get more signal than the direct path alone. If they arrive out of phase, they cancel, and you can lose a lot of signal in a very small space.

How small? The wavelength in meters is 300 divided by the frequency in megahertz. At 600 MHz, that's half a meter. A reflection off a wall behind the receiver sets up a standing pattern where peaks and nulls repeat every half wavelength: every 25 cm. Move the transmitter or the antenna a hand's width and you can go from a peak to a null.

Half a wavelength is a pattern, not a cureDirect wave plus one reflection (80% amplitude) at 600 MHz: nulls repeat every 25 cm50\-5\-10\-15\-20\-25\-300255075100125150Antenna position (cm)Received level vs. direct path (dB)AB½λ apart: both in a nullAB¼λ apart: one wins

Fig. 3: One direct wave and one reflection at 80% amplitude, 600 MHz. The received level swings from +5 dB to −14 dB and back every 25 cm. Two antennas a half wavelength apart (rust) land in matching nulls. Two antennas a quarter wavelength apart (black) can't both be in a null of this pattern.

That's the cleanest case, one reflection, and it already swings 19 dB across less than a forearm. Real rooms have dozens of reflections, and the ripple becomes a lumpy 3D field of hot and dead spots. Shure's antenna best-practice article ran a walk test in a 9 m × 8 m room with ULX-D and measured about 40 dB of RF level variation for a handheld and 50 dB for a lav, just from walking around.

There's a standard statistical model for this kind of mess. When there are lots of reflections and no single path dominates, the signal envelope follows a Rayleigh distribution. Its cumulative form is short enough to keep in your head: the probability that the received power is more than X below its average is 1 − e^(−x), where x is the fade depth as a linear power ratio. Andrea Goldsmith's *Wireless Communications* uses this form in its diversity chapter, and it's the standard starting point for diversity design.

Do it at FOH. A 10 dB fade is a power ratio of 0.1\. One minus e to the minus 0.1 is about 0.095, so in a strongly multipath room, roughly 9.5% of positions will be at least 10 dB below the average. A 20 dB fade is a ratio of 0.01, giving about 1%. A 30 dB fade is 0.1%. That 1% is the number to remember. In a busy reflective room, about one spot in a hundred is a hole 20 dB deep. Your CEO found one, and he likes it there.

This is also why "just turn the transmitter up" rarely fixes a corner dropout. Going from 10 mW to 50 mW is 7 dB. A Rayleigh null can be 20, 30, 40 dB deep. You can't out-shout a cancellation, because the reflection gets louder by exactly the same amount. You have to change the geometry, or you have to hedge your bets. That's what diversity is.

## Two antennas, three ways to use them

The logic of diversity is just the Rayleigh math, multiplied. If two antennas see fades that are statistically independent, the chance both are in a 20 dB hole at the same moment is 1% times 1%, which is 0.01%, one in ten thousand. Four independent antennas take it to one in a hundred million. Diversity doesn't make your signal stronger on average. It makes the deep holes rare.

Two antennas turn 1-in-100 into 1-in-10,000Probability a Rayleigh-faded signal sits deeper than X dB below its average, independent branches, selection100%10%1%0.1%0.01%1e-51e-61e-71e-80510152025303540Fade depth below average (dB)Probability (log scale)1 antenna2 antennas4 antennas20 dB fade: 1.0%same fade: 0.0099%

Fig. 4: Probability that a Rayleigh-faded signal is deeper than a given fade, for one, two, and four independent antennas with selection. At 20 dB, one antenna sits there 1% of the time; two antennas, 0.0099% of the time. Computed from P = (1 − e^(−x))^M.

The word "independent" is carrying the whole argument, and that's where spacing comes in. But first, there are three broad ways receivers use two or more antennas, and they're not equal.

**Antenna switching.** One receiver, two antennas, and a switch that picks which antenna feeds the single RF front end. It's cheap and it works, but the receiver only ever hears one antenna, so it has to guess when to switch, and it typically switches only after the current antenna's signal has already degraded. Done well, you don't hear it. Done slowly, the switch itself can be the click. Naming isn't standardized either (Sennheiser's EK 500 G4 portable receiver lists "adaptive diversity"), so judge any receiver by walk test, not by the label.

**True diversity, or selection.** Two complete RF front ends, one per antenna, each demodulating its own copy of the signal. The receiver continuously picks the better one, or blends toward it. This is the classic pro receiver, and its two LEDs show which branch is winning right now. Those LEDs flickering in one spot on stage is the signature of multipath: one branch falls into a hole and the other catches it. That's diversity doing its job, and a warning that you're near the edge there.

**Combining.** Instead of choosing, add the branches together with the right phase alignment and weighting. The math here is nice: with maximal-ratio combining, the combined SNR is the sum of the branch SNRs, per Goldsmith's chapter. Two equal branches give you 3 dB more average SNR on top of the fade protection. Shure's Axient Digital goes further, with Quadversity assigning four antenna inputs to a single channel, and a frequency diversity mode where a transmitter sends the same audio on two independent frequencies and the receiver uses or combines them. Frequency diversity attacks the same problem from a different side: a null at one frequency is usually not a null at another, because the path lengths that cancel at 550 MHz don't cancel at 590.

### Spacing, and the half-wavelength myth

Now the most repeated rule of thumb in wireless: space your diversity antennas a half wavelength apart. Shure's own guidance is a little more careful: no less than a quarter wavelength apart, with a half to one full wavelength ideal. At 600 MHz that's 12.5 cm minimum, 25 to 50 cm ideal. Shure also notes that spacing much wider than a wavelength is common, and it can help range, but performance then starts to look more like a non-diversity system with more reach.

RF Venue published a piece calling half-wave spacing a myth. Its argument: nulls in a real room are messy, three-dimensional, and close together, so there's no single correct spacing. Half a wavelength helps compared to one antenna, but it isn't a magic number. The comments under that piece include Robert Weller, formerly of the FCC's Office of Engineering and Technology, pointing out that diversity only helps where there's multipath fading. In line-of-sight conditions, which describes a lot of wireless audio use, the improvement is modest at any spacing.

Figure 3 shows why "half wavelength" is a slippery rule. In the single-reflection case, nulls repeat every half wavelength. Two antennas exactly a half wavelength apart along that axis land in matching nulls. A quarter wavelength apart, one is always on a slope or a peak when the other is in the hole. In a real room with many reflections, that neat periodicity dissolves, and what you actually want is for the two antennas to be far enough apart that their fades are uncorrelated. That's a statistical property of the room, not a fixed distance. A half wavelength or more is a decent bet. Exactly a half wavelength isn't special.

Two other habits matter as much as spacing.

**Polarization.** A transmitter antenna and a receive antenna at right angles can lose a lot of signal. Shure's article puts the theoretical worst case for a 90° offset at up to 20 dB, and handhelds and bodypacks get tilted every possible way during a show. Shure's recommended fix is to mount the two receive antennas in a V, at 45° and 315°, which caps the theoretical polarization loss at about 3 dB whatever angle the transmitter is at. In that same ULX-D walk test, switching to a V reduced the fade range by about 10 dB.

**Height and surroundings.** Shure recommends at least 2 m of antenna height where possible, mainly to get over the people, and keeping receive antennas at least one wavelength from any reflective surface. At 600 MHz that's half a meter. Half-wave whips screwed directly onto a receiver in a rack, behind a steel rack rail, next to the amp racks, are breaking both rules at once. Then there's the body, which is mostly salt water. A study for Switzerland's federal communications office (OFCOM), run by the IT'IS Foundation, simulated wireless mics on realistic body models and concluded the body plays a very important role in wireless mic transmission. A bodypack on the small of someone's back, with the performer facing away from the antennas, is shooting through them.

## Spending the link budget

How much signal do you actually have to play with? Start with the free-space path loss. The standard form, with distance in kilometers and frequency in megahertz, is FSPL = 20·log₁₀(d) + 20·log₁₀(f) + 32.44 dB. For 50 meters at 600 MHz: 20·log₁₀(0.05) is −26.0, 20·log₁₀(600) is 55.6, add 32.44, and you get about 62 dB. A 10 mW transmitter is +10 dBm. So in empty air, at 50 meters, you'd see about −52 dBm at the antenna.

Against a receiver sensitivity around −103 dBm, that's 51 dB of margin. Sounds enormous. Now spend it.

Spending a 10 mW link budget600 MHz, 50 m line of sight. \*Body loss is an illustrative placeholder; measure your own10\-10\-30\-50\-70\-90\-110\-130Level (dBm)TX power+10 dBm (10 mW)Free space, 50 m\-62 dBBody\-10 dB\*Polarization (V)\-3 dBFade reserve\-20 dB (1%)\-85 dBm leftAt the receiverreceiver sensitivity -103 dBmthermal floor, 200 kHz (kTB) -121 dBm18 dB of margin left

Fig. 5: A 10 mW, 600 MHz transmitter at 50 m. Free-space loss takes 62 dB; a V antenna pair limits polarization loss to about 3 dB; holding a 20 dB fade reserve covers all but about 1% of Rayleigh positions. The 10 dB body loss is a placeholder for illustration, not a measured figure. What's left is about 18 dB above sensitivity.

Polarization in a V configuration, about 3 dB. A fade reserve to cover the 1% Rayleigh hole, 20 dB. Then the body, which depends on where the pack sits and which way the talent faces; the 10 dB in the figure is a stand-in, so walk-test your own. That leaves about 18 dB before cable loss, distribution, a mis-aimed antenna, or an LED wall raising the noise floor in one corner. Not so enormous now.

Doubling the distance costs 6 dB. RF Venue gives a tidy example: 1,600 nanowatts at 25 feet becomes 400 at 50 feet and 100 at 100 feet. So moving the receive antennas from the back of the room to the side of the stage is often worth more than any amount of transmitter power, because the distance term is logarithmic and the first few meters cost the most.

Antenna gain is cheap dB. RF Venue's primer lists a typical log-periodic paddle at about 5 dBd with a 55° beamwidth and a circularly polarized helical beam around 9 dBd. That gain is real, and it also narrows where the antenna listens. Point a paddle at the stage and you've traded the back of the room for the front.

And then there's the opposite problem. Shure recommends at least 3 m between a 10 mW transmitter and a passive omni receive antenna, so you don't overload the receiver's front end. With an active directional antenna at +12 dB and a UHF-R transmitter, its example stretches that to as much as 15 m. A presenter who walks right up to the antenna mast can drop out from too much signal, which is a fun one to explain to a client.

### The gate at the end: squelch

The last stage in the receiver is squelch, and it's the reason a dead link goes silent instead of roaring white noise into the PA. RF Venue's explainer describes three common approaches. RF level squelch mutes when signal strength drops under a threshold; simple, but noise above that threshold opens it. Audio frequency squelch watches the demodulated audio for energy up above roughly 16 to 20 kHz, where a microphone shouldn't have any, and mutes when it sees hash there. Tone key, or pilot tone, squelch has the transmitter add an inaudible tone, and the receiver stays muted unless it hears that tone. Modern receivers usually combine more than one method.

The squelch threshold is a real tradeoff, not a set-and-forget. Raise it and the receiver mutes earlier and more reliably when the link gets marginal, at the cost of range. Lower it and you get more range, plus a better chance of hearing the spits and clicks of a near-threshold carrier. Adding directional antennas or a distribution system changes system gain, so the squelch may need resetting. And if you ever have a receiver chirping noise into an open channel with the transmitter off, check whether someone has turned off the pilot tone. Sennheiser's spec sheet notes the G4's pilot tone squelch can be switched off, which is useful when mixing generations and a liability when it's done by accident.

So, the CEO. Signal strong on average, both diversity LEDs flickering in one spot, audio dropping out rather than going garbled. That's multipath, not capture and not weak signal. The fix, in order: get the antennas off the rack and up high, at least a wavelength away from that LED wall's steel frame; set them in a V; move them closer to the stage so the average signal is higher and the fade reserve is bigger; and if you can, put the bodypack on his hip on the side facing the antennas. Start with the first two. They cost nothing but a ladder.

## What digital fixed and what it charged you

Digital wireless changes the middle of the chain. Instead of wiggling a carrier with analog audio, the transmitter digitizes the capsule, encodes it, and transmits data. The receiver decodes it back to audio. Shure's companding explainer notes digital systems don't need companding at all, which can mean a less colored sound. The breathing is gone, and so is the pre-emphasis trade. Axient Digital's published dynamic range is 120 dB(A), and the receiver can carry AES-256 encryption, which matters a lot more to a corporate client than to a rock band.

Digital also changed spectral efficiency. Axient Digital's High Density mode fits up to 47 active channels in a 6 MHz TV channel, or 63 in 8 MHz. If you've stared at a coordination plot of thirty analog systems dodging their own intermod, that reads like a dream.

What did it charge you? Two things.

**Latency.** Conversion and processing take time. Shure's latency explainer gives the ULX-D figure as about 2.9 ms from audio input to the receiver's analog output. Axient Digital's product page claims 2 ms from transducer to analog output, rising to 2.9 ms in High Density mode. Shure's own guidance on monitoring is that 5 to 10 ms is generally fine for wedges, more than 10 ms gets noticeable, and for in-ears 5 ms or less is recommended, because a singer hears themselves through bone conduction and any delay in the ear shows up as comb filtering against it. A 2.9 ms mic fits that on its own. Add a digital console, a network hop, and a digital IEM transmitter and you can blow past 5 ms without any one box looking guilty.

**The cliff.** FM degrades with a warning. You hear hiss, then spits, then nothing. A digital link stays perfect until the error correction runs out, then it's gone, sometimes with a mute, sometimes with a short artifact. Your ears lose the early warning, so the meter is the warning now. Keep a real fade reserve.

There's one more wrinkle worth knowing on the monitoring side. Analog stereo in-ear systems usually transmit stereo the way FM radio does, with a multiplex: a mono sum, a 38 kHz difference signal, and a 19 kHz pilot to tell the receiver to decode stereo. A ProSoundWeb first look at a digital IEM system walked through the costs: the pilot and encoding force a steep roll-off above about 15 kHz, and the asymmetrical multiplex is more susceptible to noise and multipath than mono. So if a marginal analog IEM link offers a mono mode, try it. You give up the difference channel for the robustness of the sum.

## What you can take back to the desk

You don't need an RF degree to run wireless well. You need the shape of the chain and a handful of napkin numbers.

**Read the receiver like a doctor reads a chart.** Strong RF and garbled audio: think capture, a stronger rogue signal or intermod within a few dB. RF that swings wildly while the diversity LEDs flicker: multipath. RF steadily low across the whole stage: distance, antenna placement, or the body. Audio fine but noise breathing on quiet passages: that's the compander doing exactly what it was designed to do.

**Know where your cliff is.** Thermal noise in a 200 kHz channel is −121 dBm. A decent receiver plus the FM threshold puts the useful floor around −103 dBm, which matches a typical spec of about 1.6 µV. A 10 mW transmitter starts at +10 dBm. Everything between is budget.

**Keep a 20 dB fade reserve.** In a reflective room, a 20 dB hole shows up about 1% of the time with one antenna. Diversity drops that to about one in ten thousand, if the antennas are far enough apart to fade independently.

**Space and angle the antennas.** At least a quarter wavelength, ideally a half to a full wavelength apart. At 600 MHz that's 25 to 50 cm. Mount them in a V. Get them up at least 2 m and a wavelength away from steel. Don't leave them on the back of the rack because that's where they came.

**Move the antennas before you raise the power.** Doubling distance is 6 dB. Going from 10 to 50 mW is 7 dB. A Rayleigh null can be 20 to 40 dB. Geometry beats wattage almost every time.

**Set transmitter gain once, and set it right.** Peaks near the top on the loudest real moment. Too hot and the limiter grits up the attacks, especially on bright sources because of pre-emphasis. Too cold and you're living in the part of the curve where the noise rides closest to the program: SNR is roughly 60 + L/2.

**Budget latency for the whole chain.** A 2 to 3 ms digital mic is fine on its own. The in-ear target is 5 ms total. Add up every converter and network hop before the singer finds it for you.

**Walk the room.** The ULX-D walk test found 40 to 50 dB of variation in a single small room. A spectrum scan won't find the CEO's corner. Someone walking the stage with a transmitter while you watch the meters, before doors, will.

Your wireless is not a cable. It's a small radio station, a compressor, an expander, a coin toss with the walls, and a gate. Treat it like that, and the next dropout has a cause instead of a ghost.

*Above the Fader*

## Sauce

- [Shure: Wireless system companding explained](https://www.shure.com/en-GB/performance-production/louder/wireless-system-companding-explained?ref=abovethefader.com)
- [Shure: Audio Reference Companding, the basics](https://www.shure.com/pt-BR/shows-e-producoes/louder/audio-reference-companding-the-basics?ref=abovethefader.com)
- [Shure: All about wireless antenna positioning, best practices](https://www.shure.com/en-US/insights/all-about-wireless-antenna-positioning-best-practices?ref=abovethefader.com)
- [Shure: Digital wireless latency explained](https://www.shure.com/en-GB/insights/digital-wireless-latency-explained?ref=abovethefader.com)
- [Shure: Axient Digital product page](https://shure.com/axientdigital?ref=abovethefader.com)
- [Sennheiser: ew 500 G4 product specification (PDF)](https://www.sennheiser.com/globalassets/digizuite/41579-en-sp%5F1156%5Fv2.0%5Few%5F512p%5Fg4%5Fproduct%5Fspecification%5Fen.pdf?ref=abovethefader.com)
- [RF Venue: Three essential concepts in wireless audio (PDF)](https://www.rfvenue.com/hubfs/316692/Three%5FEssential%5FConcepts%5Fin%5FWireless%5FAudio.pdf?ref=abovethefader.com)
- [RF Venue: The myth of half-wave diversity antenna placement](https://www.rfvenue.com/blog/2014/12/15/the-myth-of-half-wave-diversity-antenna-placemen?ref=abovethefader.com)
- [RF Venue: What is squelch?](https://rfvenue.com/blog/2015/03/20/what-is-squelch?ref=abovethefader.com)
- [AKG WMS61 white paper: noise considerations (via ManualsLib)](https://www.manualslib.com/manual/3778/Akg-Wms61.html?page=4&ref=abovethefader.com)
- [All About Circuits: Exploring the threshold constraints of FM demodulation](https://www.allaboutcircuits.com/technical-articles/exploring-the-threshold-constraints-of-fm-demodulation-through-solved-examples/?ref=abovethefader.com)
- [PA3FWM: Noise in FM reception](https://pa3fwm.nl/technotes/tn24-fm-noise.html?ref=abovethefader.com)
- [Wikipedia: Preemphasis improvement](https://en.wikipedia.org/wiki/Preemphasis%5Fimprovement?ref=abovethefader.com)
- [RF Wireless World: Pre-emphasis vs. de-emphasis](https://rfwireless-world.com/terminology/rf-basics/pre-emphasis-vs-de-emphasis?ref=abovethefader.com)
- [RF Essentials: Noise floor](https://rfessentials.com/resources/rf-glossary/noise-floor/?ref=abovethefader.com)
- [RF Essentials: Free space path loss](https://rfessentials.com/resources/rf-glossary/free-space-path-loss/?ref=abovethefader.com)
- [Goldsmith, Wireless Communications, ch. 7: Diversity](https://personal-wiki-ddd.pages.dev/wireless/signal-processing/WirelessComm%5FChp1-16%5FMarch32020/ch-07-diversity?ref=abovethefader.com)
- [OFCOM (BAKOM): The effect of the human body on wireless microphone transmission](https://www.bakom.admin.ch/bakom/en/home/frequencies-and-antennas/facts-and-figures/the-effect-of-the-human-body-on-wireless-microphone-transmission.html?ref=abovethefader.com)
- [ProSoundWeb: First look, the Lectrosonics Quadra (FM multiplex in IEMs)](https://www.prosoundweb.com/first-look-the-lectrosonics-quadra/2/?ref=abovethefader.com)

*Photo: [Philipp M.](https://unsplash.com/@luftschnitzel?utm%5Fsource=above%5Fthe%5Ffader&utm%5Fmedium=referral) on [Unsplash](https://unsplash.com/photos/black-wireless-microphone-daSwn4uD8Q0?utm%5Fsource=above%5Fthe%5Ffader&utm%5Fmedium=referral).*