Audio and Timecode

A dance piece ends on a drum hit and a shout from the whole company. The picture is sharp and the exposure holds. Play it back and the last two seconds crackle like a torn speaker. The sound arrived louder than the largest number the file could hold, so every sample above that ceiling was written as the ceiling, and no software can work out afterwards how far above it the real sound went.

Sound fails at the moment of recording, with far less warning on screen than a blown highlight gives. The Z6III offers a level meter, seven audio items, a microphone socket and a headphone socket. All of them act on one electrical signal, so that signal comes first, then the controls, the route I use (an Insta360 Mic Pro into the microphone socket), and timecode, the frame labels that let separate recordings be lined up.

01.Sound From First Principles

Pressure Becomes Voltage

Sound is a rapid rise and fall of air pressure. A note at 440 Hz pushes on your eardrum 440 times a second. A microphone holds a thin membrane, the diaphragm, in that moving air, and turns its movement into a voltage that rises and falls in the same pattern. That voltage is the audio signal.

The voltage is small, so the camera amplifies it first. The amount of amplification is the gain, which the Z6III calls sensitivity. The camera then measures the amplified voltage at fixed instants and stores each measurement as a number, a sample. A recording is a long list of those numbers.

Decibels

Sound levels span an enormous range, so engineers compare them as ratios on a logarithmic scale, the decibel (dB). Take two voltages, 0.1 V and 0.2 V. The second is twice the first, and in decibels the difference is

$$L = 20\log_{10}\frac{V}{V_{\text{ref}}} = 20\log_{10}\frac{0.2}{0.1} = 20\log_{10}2 \approx 6.02\ \text{dB}.$$

Here $V$ is the voltage being described, $V_{\text{ref}}$ the voltage it is compared with, and $\log_{10}$ asks what power of ten gives the ratio. Doubling a voltage adds about 6 dB and halving it subtracts about 6 dB. The factor is 20 rather than 10 because the decibel was defined for power, which grows as the square of voltage.

A decibel figure needs a reference. Against 1 volt the unit is dBV, which Nikon uses for the line input (p. 610), so +6 dBV is $10^{6/20} \approx 2.0$ V. Inside the file the reference is full scale, the largest number a sample can hold. Full scale is 0 dBFS, and everything the file can hold sits below it at a negative figure.

Levels, Headroom and Clipping

The level is how large the signal is at a given moment, which is what the meter shows. Headroom is the gap between the loudest peak and full scale. A signal with headroom is stored exactly. A signal without it clips.

Here is clipping in numbers. Suppose full scale is 1.0 and six successive samples of a loud passage would measure 0.6, 0.9, 1.3, 1.6, 1.3 and 0.9. The file stores 0.6, 0.9, 1.0, 1.0, 1.0 and 0.9. Three samples that differed are now identical. Nothing in the list says whether the true peak was 1.1 or 3.0, so lowering the volume later only lowers a flat-topped wave, which the ear hears as crackle.

The opposite mistake is quieter. Every microphone and amplifier adds a faint hiss, called noise, and a recording made far too low must be raised later, hiss and all. The aim is a level well clear of the noise with headroom left for the loudest moment. Nikon prints no scale for the meter in the Reference Guide, so I give no dBFS target. It does print the rule that matters: a sound level shown in red is too loud, and the fix is lower sensitivity (p. 611).

Sample Rate and Bit Depth

The sample rate is how many measurements are taken each second. The Z6III records at 48 kHz, 48,000 a second (p. 1004). A sampled signal can describe frequencies up to half its sample rate, so 48 kHz reaches 24 kHz, above the top of human hearing.

The bit depth is how many distinct values each sample can take. A 16-bit sample has $2^{16} = 65{,}536$ and a 24-bit sample $2^{24} = 16{,}777{,}216$. Each extra bit doubles the steps, about 6.02 dB, so 16 bits span roughly 96 dB and 24 bits roughly 144 dB in theory. Extra bits give finer steps near the noise. They do not raise the ceiling, and a 24-bit file clips at full scale like any other.

The file type decides which you get. N-RAW, ProRes RAW, ProRes 422 HQ and both H.265 types record Linear PCM, which stores every sample as measured. H.264 8-bit (MP4) records AAC, a compressed format that discards detail it judges inaudible (pp. 197-198). Linear PCM is 48 kHz and 24-bit, and AAC is 48 kHz and 16-bit (p. 1004). My everyday H.265 10-bit (MOV) files, chosen in Video Formats and Codecs, therefore carry 24-bit Linear PCM.

02.The Camera's Microphone and Jacks

The built-in stereo microphone is the pair of openings on the front of the viewfinder housing (p. 47), and Nikon asks you not to cover it while recording (p. 108). Being part of the camera, it hears the camera: Nikon warns that autofocus, vibration reduction and power aperture may be audible (p. 109). On my camera Fn1 and Fn2 drive power aperture in video, as Custom Controls sets out, so pressing them mid-take risks putting that noise on the built-in microphone's track. Nikon notes that an external microphone reduces that equipment noise (p. 972).

A cover on the side away from the grip protects the sound sockets (p. 48), shown in A Tour of the Body. The external microphone/line input is a 3.5 mm stereo mini-pin jack with line input and plug-in power, and the headphone output is the same size (p. 1006). Plug-in power is a small voltage the camera puts on the socket to run microphones that have no battery. Third-party microphones with 3.5 mm plugs work (p. 213), alongside Nikon's ME-1 stereo microphone and ME-W1 Bluetooth wireless microphone (p. 972).

The shooting display shows a sound level indicator and the sensitivity in the monitor and the viewfinder (pp. 57, 63). Slow motion records no sound (p. 203), and neither does the time-lapse movie in Silent Shooting, Timers and Time-Lapse.

03.The Audio Items

All seven audio items sit in the video recording menu, with Nikon's defaults on p. 579. Audio input sensitivity already holds slot 4 of the video i menu and Wind noise reduction slot 10, as The i Menu records, and Custom Setting g1 can add Attenuator, Frequency response and Headphone volume (p. 746).

Audio Input Selection

External mic, the default, is for a microphone. Line is for a mixer or similar device on a line cable (p. 610). A line signal has already been amplified far above a microphone's. Nikon is firm: with a line source always choose Line, because External mic may cause malfunctions, and keep the source at +6 dBV or less, about 2 V, since higher voltages may distort. Line also locks Frequency response to Wide range (p. 613) and cuts plug-in power (p. 615).

Audio Input Sensitivity

This is the gain. Auto, the default, adjusts it for you. Manual runs from 1 to 20, higher meaning more sensitive. Microphone off records no sound and marks the clip with an icon (p. 611). Nikon's checklist for a silent clip names Audio input selection as well as Microphone off and a low Manual value (p. 917).

I use Manual. A fixed number cannot change during a take, so loud and quiet passages keep their relationship. Nikon does not say how Auto decides, and I prefer a setting I can read off the screen.

Attenuator

ON lowers the input sensitivity to prevent distortion in loud surroundings (p. 612), and the default is OFF. Nikon gives no amount. I treat it as the next step when the meter still reaches red at a low Manual number, such as the built-in microphone near a loudspeaker.

Frequency Response

This sets which pitches the built-in and external microphones record. Wide range, the default, covers everything from music to street noise, and Vocal range suits human voices (p. 613). Nikon gives no frequencies. Dance carries music, so I keep Wide range and would choose Vocal range only for an interview.

Wind Noise Reduction

Wind across a microphone makes a low rumble. ON applies a low-cut filter, which removes the lowest frequencies, to the built-in microphone, and other sounds may be affected. It has no effect on optional stereo microphones (p. 614). The default is OFF. Nikon names no wireless receiver, but the filter is described for the built-in microphone and the Mic Pro arrives through the socket those optional microphones use, so I infer the item does nothing for it. I keep it OFF with the Mic Pro and turn it ON from slot 10 for the built-in microphone in a breeze.

Mic Jack Plug-in Power

ON, the default, supplies plug-in power, OFF supplies none, and Line cuts it either way. Nikon recommends OFF for a microphone that needs no power, to avoid noise from the supply, and suggests asking the microphone's maker (p. 615). Firmware 1.10 fixed noise from such microphones at manual gain, as Firmware History records.

Headphone Volume

This sets how loud headphones in the headphone socket play. Press up or down to change it (p. 616). The default is 15 (p. 579) and Nikon prints no range. Firmware History lists a 2.00 change that lets it be adjusted from the i menu while recording.

I own no headphones, so I watch the meters. A reader should consider a pair, because a meter shows how loud the sound is and nothing about what it is. Hum, clothing rubbing a transmitter and wind all register as level, and only ears catch them during the take. Headphones in the camera's socket hear what the camera receives. The Mic Pro receiver's own black 3.5 mm monitoring jack, Insta360 says, carries what the transmitters pick up, so my inference is that it cannot reveal a camera-side problem such as the wrong sensitivity.

04.The Insta360 Mic Pro Route

Insta360's box list for the Two Transmitters, One Receiver kit names two transmitters, a receiver, a charging case, a 3.5mm Camera Audio Cable, two windshields, back clips and button magnets. The transmitters clip to the performers and the receiver sits on the camera.

Insta360's guide to digital cameras says to fix the receiver's back clip in the hot shoe and run the 3.5mm TRS cable from the receiver's output, marked with a red circle, to the camera's audio input. TRS names the plug's three contacts, tip, ring and sleeve. Insta360 adds Sync Camera Power, on by default, which switches the receiver on and off with a camera connected this way, provided the camera is in video mode.

Three Volume Controls in a Row

The sound now passes three gain stages: each transmitter's TX Gain, the receiver's RX Gain and the camera's sensitivity. Insta360 gives TX Gain and RX Gain the same span, $-12$ to $+12$ dB, and says RX Gain sets the level sent to the camera so that the camera's input does not clip. In voltage that span runs from $10^{-12/20} \approx 0.25$ to $10^{12/20} \approx 3.98$ times, a factor of about 16. Insta360 advises keeping TX Gain away from its extremes, where it raises hiss or clips, and also offers TX Auto Gain.

Insta360's camera guide gives settings only for Canon and Sony: set the camera's level to manual and low, then use the receiver's gain as the main volume control. Nikon is not mentioned. My judgement is that the pattern suits the Z6III: Audio input sensitivity on Manual at a low number, TX Gain near the middle, and RX Gain set against the camera's meter in a rehearsal. Each stage amplifies the hiss of the stages before it, so turning them all up piles hiss on hiss and loud passages reach red sooner.

I keep Audio input selection on External mic. Insta360 calls the camera's socket a microphone input and gives no output level for the receiver, so treating it as a microphone is my inference. Nikon's plug-in power advice points to OFF for a source with its own power, and Insta360 lists a 410 mAh receiver battery, but it does not say whether the receiver draws plug-in power or how Sync Camera Power detects the camera. So I would try OFF and check that the meter still moves and the receiver still follows the camera on and off. If either fails, go back to ON.

Channels, Filters and a Backup

Insta360 lists four channel modes. Stereo puts transmitter 1 on the left and transmitter 2 on the right. Mono sends the same mix to both. Quadraphonic gives four transmitters a channel each, but a stereo cable carries two, and Insta360 ties four-channel camera recording to its Sony hot-shoe adapter. Safety Track is Mono with the right channel 6 dB lower, as a backup against clipping. If a shout peaks 4 dB over full scale on the left, the right copy peaks at $4 - 6 = -2$ dB, two below full scale, and survives. With two performers wearing transmitters I would choose Stereo, and with one, Safety Track.

Each transmitter's Low Cut, Insta360 says, filters sound at 100 Hz and below. With the windshields, that is the wind control on this route.

The transmitters can also record WAV files to their own storage, which Insta360 presents as a backup if the wireless link drops, in 24-bit or in 32-bit float, the second a format with room far above the usual full scale. Insta360 states that 32-bit float comes only from the transmitter, never the camera.

05.Timecode From First Principles

Timecode labels each frame with hours, minutes, seconds and frames. The label 01:23:45:12 means frame 12 of second 45 of minute 23 of hour 1. When two devices stamp the same label on the same instant, their recordings can be lined up by matching labels instead of by eye.

That needs clocks that agree, and no two clocks run at quite the same speed. As an illustration, a clock off by 20 parts per million gains $3600 \times 0.00002 = 0.072$ s an hour, a little over two frames at 30p. A timecode generator supplies one shared count to every device, and setting a device's count from it is called jam sync.

Drop Frame at 29.97

The Z6III's 30p actually runs at 29.97 frames a second (p. 201), which is $30/1.001$. A timecode label still counts 30 frames to each labelled second, so the labels fall behind the wall clock. In one real hour the camera records

$$29.97 \times 3600 = 107{,}892 \text{ frames},$$

which at 30 frames per labelled second reach $107{,}892 / 30 = 3{,}596.4$ seconds. After a full hour the label reads 00:59:56:12, about 3.6 seconds short.

Drop-frame timecode skips frame numbers, never frames. At the start of every minute it leaves out labels 00 and 01, except in every tenth minute. Six of the 60 minutes (0, 10, 20, 30, 40 and 50) keep their labels, so $54 \times 2 = 108$ labels are skipped, and $108 / 30 = 3.6$ seconds. At the exact rate of $30/1.001$ the shortfall is 3.5964 seconds, so drop frame matches it to within about a tenth of a frame an hour. Nikon describes the Drop frame setting for 30 and 60 frames a second (p. 617), and 60p runs at 59.94 (p. 201).

06.The Timecode Item

With timecode recording on, every frame gets a label, though not in H.264 8-bit (MP4) (p. 617), and slow motion blocks timecode output (p. 204). It holds four settings.

Record timecodes offers On, which records the code and shows it in the display, On (with HDMI output), which also passes it to an HDMI recorder, and Off, the default (pp. 579, 617). Nikon cautions that On (with HDMI output) may disrupt footage output to HDMI devices. My F700 is a monitor, not a recorder, so I have no reason to risk it, and HDMI and External Monitors covers that link.

Count-up method chooses Record run, the default, counting only while recording, or Free run, counting continuously even with the camera off (p. 617). Record run gives one unbroken count across a day's clips. Free run behaves like a clock, which suits several devices set to the same time.

Timecode origin sets the start. Reset returns the count to zero, Enter manually takes a typed value, and Current time copies the camera clock, so check the time zone and date first (p. 617). Reset using remote lets one WR-T10 remote reset several cameras together, each fitted with a paired WR-R11b or WR-R10 (p. 618).

Drop frame is ON by default (p. 579).

I leave Record timecodes Off. With one camera and no timecode source it can read, the labels would match nothing. The Mic Pro receiver keeps its own timecode clock, but it sends the code as sound, and Nikon describes no way to read timecode from the microphone socket.

07.ATOMOS AirGlu BT Options

This is for a reader with an Atomos UltraSync BLUE, a timecode generator that sends its count over Bluetooth to several cameras and recorders, other makers' included (p. 898). Connect to ATOMOS AirGlu BT opens the link to an UltraSync BLUE paired earlier, and defaults to OFF (p. 871). Save ATOMOS AirGlu BT pairing info pairs the two, with the pairing started from the UltraSync BLUE's own controls. Despite its name, Nikon describes Delete ATOMOS AirGlu BT pairing info as ending the wireless connection (p. 898). Camera sets the name the UltraSync BLUE lists the camera under, NCZ6_3 by default (pp. 871, 898).

After connecting, pick either of the On choices under Record timecodes. The UltraSync BLUE's frame rate must match the camera's: 29.97 fps or 29.97 fps DF for 120p, 60p and 30p, 25 fps for 100p, 50p and 25p, and 23.98 fps for 24p, with 29.97 fps DF for drop frame. While the link is up, the camera locks Drop frame, Timecode origin and Count-up method, and MP4 records no code (p. 899).

The TC indicator turns red when no code arrives or the link has ended, and footage started before any code arrives records none. Once code has been received, the camera keeps showing the value after the link drops (p. 898). Firmware History notes that not all of Nikon's 2.00 sources list this behaviour. Switching the camera off ends the link, and so does the standby timer running out, so Nikon suggests No limit for the standby timer in c3 Power off delay (p. 899).

08.Lining Up Separately Recorded Sound

Sound recorded on a separate device, such as a field recorder or a Mic Pro transmitter, starts at a different moment from the camera's file. Putting them together needs one event both captured.

The oldest answer is a clap. On a waveform, a graph of level against time, two hands meeting make a tall spike in both recordings, and the picture shows the frame where the hands meet. Slide the separate sound until its spike sits on that frame. A frame at 30p lasts $1/29.97 \approx 0.033$ s, which is why a clap beats a spoken word: its start is sharp to well within a frame.

Editing software can also slide one file along until the two waveforms match, which is a reason to record sound in the camera even when a better copy exists elsewhere. Over a long take the clocks drift as worked out above, so a second clap at the end, or shared timecode, keeps the far end aligned. Insta360 says its receiver can send timecode as audio on the left channel, sound on the right. I infer the camera would store that code as an ordinary audio track, costing a channel, so I leave it off.

09.Recommended Settings

10.Beginner Mistakes

Turning everything up. High TX Gain, high RX Gain and a high Manual number stack three amplifiers, raising the hiss and clipping the loud passages. Keep the camera low, TX Gain moderate, and drive the level with RX Gain.

Feeding a mixer into External mic. Nikon warns that a line signal on External mic may cause malfunctions. Choose Line and keep the source at +6 dBV or less (p. 610).

Recording silence. Microphone off, a Manual value near the bottom or the wrong Audio input selection can each give a silent clip (p. 917), and Nikon describes a marking icon for the first only (p. 611). Check that the meter moves before every take.

Expecting timecode in MP4. H.264 8-bit (MP4) records none (p. 617), so a shoot that relies on timecode needs a MOV or NEV file type.

11.Related Chapters

• A Tour of the Body: where the microphone and headphone sockets sit.
• Viewfinder and Monitor Displays: the shooting display and its sound level indicator.
• The i Menu: the video i menu, which already holds two audio items.
• Custom Controls: the power aperture roles the built-in microphone can hear.
• Firmware History: the audio and AirGlu changes in versions 1.10 and 2.00.
• Video Formats and Codecs: the file types that decide between Linear PCM and AAC.
• HDMI and External Monitors: the HDMI link that On (with HDMI output) uses.