Microphones, mixers and signal flow

The first piece of AV equipment Maya wires on her own is the paging microphone for Riverside's main office: a gooseneck microphone on a desk base with a push-to-talk button, cabled to the rack in the communications room two floors up. It works on the first try, but the page sounds thin and a little noisy. Wendell shows her why: the microphone input on the mixer is set for a line-level source, the cable shield is connected at both ends to two different grounds, and the gain on the DSP has been turned up to make up for it. "Every hum, hiss and squeal in a sound system starts in the first few meters of the signal chain," he says. This lesson covers microphone types and pickup patterns, signal levels, balanced and unbalanced connections, gain structure, mixers and DSPs, and the signal flow drawings that tie it all together.

How microphones work

A microphone turns sound pressure into a small electrical signal. Two types cover nearly everything an installer meets. A dynamic microphone moves a coil attached to a diaphragm inside a magnet's field, generating a voltage the way a small generator does. Dynamic microphones are rugged, need no power and handle loud sounds well; many paging and handheld microphones are dynamic. A condenser microphone uses a very thin diaphragm as one plate of a capacitor; sound moves the diaphragm and changes the capacitance, and a small amplifier inside the microphone turns that into a signal. Condensers are more sensitive and capture high frequencies better, which suits ceiling, boundary and gooseneck microphones for speech, but they need power.

That power usually comes from the mixer as phantom power: 48 volts DC sent down both signal conductors of the microphone cable equally, with the shield as the return. Because it is the same on both signal wires, it does not disturb the audio. Phantom power is switched on per input at the mixer or DSP. It does no harm to a balanced dynamic microphone, but it should be off for inputs connected to line-level equipment or to some older and unbalanced devices. Network microphones, which lesson 7 covers, are powered by PoE instead.

Pickup patterns

Microphones differ in which directions they hear. An omnidirectional microphone picks up equally from all around; it suits a microphone in the middle of a table where people sit on every side, but it also picks up room noise and loudspeaker sound from every direction. A cardioid microphone, named for its heart-shaped pattern, picks up mainly from the front and rejects sound from the rear; it is the usual choice for lecterns, paging consoles and handhelds. Supercardioid and hypercardioid patterns are narrower at the front with a small rear lobe. Beamforming ceiling microphones, common in meeting rooms, use arrays of small capsules and processing to form several narrow pickup lobes aimed at the seats, which lesson 10 covers.

Pickup pattern matters most for feedback. Feedback is the howl that happens when a loudspeaker's sound re-enters a microphone, is amplified again, and builds up at the frequency where the loop gain is highest. A directional microphone aimed at the talker and away from the loudspeakers, loudspeakers aimed at the audience and away from the microphone, and a short distance from mouth to microphone all increase the gain available before feedback. Distance matters a lot: halving the talker's distance to the microphone raises their level at the microphone by about 6 dB, which is 6 dB more gain the system can use before it howls.

Signal levels

Audio signals travel at very different levels at different points in a system, and connecting the wrong level to an input is the first cause of poor sound. Microphone level is tiny, a few millivolts, and needs a preamplifier with high gain. Line level is the standard level between equipment: professional gear works at about +4 dBu, about 1.23 volts, and consumer gear at about −10 dBV, about 0.32 volts. Speaker level is the output of a power amplifier: tens of volts and several amperes for low-impedance loudspeakers, or up to 70 or 100 volts on distributed lines. Lesson 4 covers speaker level.

The units look confusing but follow the same pattern as SPL: dBu is relative to 0.775 volts, and dBV is relative to 1 volt. Plugging a line-level source into a microphone input overloads the preamplifier and distorts; plugging a microphone into a line input leaves the signal far too weak, so someone turns up the gain later in the chain and raises the noise with it. The office paging microphone at Riverside had the second problem.

# Convert between audio level units and check a connection
import math
def dbu_to_v(dbu): return 0.775 * 10 ** (dbu / 20)
def dbv_to_v(dbv): return 1.0 * 10 ** (dbv / 20)
def v_to_dbu(v): return 20 * math.log10(v / 0.775)
print(f"+4 dBu  = {dbu_to_v(4):.3f} V")
print(f"-10 dBV = {dbv_to_v(-10):.3f} V = {v_to_dbu(dbv_to_v(-10)):.1f} dBu")
mic_mv = 5.0
print(f"A 5 mV microphone signal is {v_to_dbu(mic_mv / 1000):.1f} dBu")
gap = 4 - v_to_dbu(mic_mv / 1000)
print(f"Gain needed to bring it to +4 dBu line level: about {gap:.0f} dB")

+4 dBu  = 1.228 V
-10 dBV = 0.316 V = -7.8 dBu
A 5 mV microphone signal is -43.8 dBu
Gain needed to bring it to +4 dBu line level: about 48 dB

About 50 dB of gain is needed to bring a speech microphone up to line level, which is what a microphone preamplifier provides. If that gain is not applied at the preamplifier, where the signal is cleanest, it ends up being applied later, where it amplifies every bit of noise picked up along the way.

Balanced and unbalanced connections

A balanced connection uses two signal conductors and a shield. The signal is carried on the two conductors in opposite polarity, and the receiving input subtracts one from the other. Hum or interference picked up along the cable appears equally on both conductors, so the subtraction cancels it, while the signal, which is opposite on each, adds. This is called common-mode rejection, and it is why professional audio uses balanced lines for anything longer than a few meters. Microphone cables, XLR connectors and the three-pin Euroblock, or Phoenix, terminals on most installed equipment are balanced.

An unbalanced connection uses one signal conductor and a shield that also carries the return. It has no protection against interference and suits only short connections between consumer devices, such as a media player and a nearby input. Where an unbalanced source must reach a balanced input over any distance, a balancing transformer or an active converter is used at the source end. On Euroblock terminals, the pins are usually marked plus, minus and ground; for a balanced connection both signal pins are used, and getting plus and minus swapped inverts the polarity, which can cause cancellation when two microphones or speakers are combined.

Grounding and shields cause most hum. The standard practice for installed audio is to connect cable shields at one end only where equipment is in different racks or rooms on different power circuits, usually at the receiving end or as the manufacturer's manual directs, and to keep all the AV equipment in a rack on the same power circuit and ground. A ground loop, which happens when two pieces of equipment are connected both by the shield and through separate building grounds, lets a small current flow in the shield and adds a 60 Hz hum and its harmonics. Isolation transformers break the loop where it cannot be avoided. Lifting the safety ground on equipment power cords is never the fix.

Mixers and DSPs

A mixer combines several inputs into one or more outputs, with a level control for each. Older paging systems used a small mixer amplifier that did everything in one box: microphone and music inputs, tone controls, priority muting and a power amplifier. Most installed systems now use a digital signal processor, a DSP, a rack unit with microphone and line inputs, outputs and a processing engine configured from a laptop. Inside the DSP, the designer or programmer builds the system from software blocks: input gain, high-pass filters to remove rumble, equalizers, automatic gain control, feedback suppression, compressors and limiters, mixers, matrix routers, delays and output levels.

A DSP can do in software what used to need a rack full of hardware, and it can be changed without rewiring. It can also hide mistakes: a programmer can make up for a wrongly set preamplifier with gain three blocks later, and the system will appear to work until it is examined. Cobalt Line's practice is that the installer checks the physical connections and the input settings, the programmer builds the processing from the design, and both check gain structure together at commissioning.

Gain structure

Gain structure means setting the level at each stage so the signal stays well above the noise floor and well below the clipping point everywhere along the chain. The rule is to set the gain at the first stage, the microphone preamplifier, so that normal speech reaches the design's nominal level, often around 0 dBFS minus 18 to 20 dB on a digital meter, with loud speech still leaving headroom below clipping. Every later stage should then be set close to unity, passing the signal on at about the same level, and the final level in the room set at the amplifier or the DSP's output stage. A chain built this way has the least noise and the most headroom.

Digital equipment uses dBFS, decibels relative to full scale: 0 dBFS is the largest value the converter can represent, and every signal is a negative number below it. Nothing passes above 0 dBFS; trying to produces harsh clipping. Analog and digital levels meet at the converters, where the DSP's setup states which analog level corresponds to which dBFS value.

# Check a gain structure: levels through each stage, against noise and clip points
stages = [  # name, gain dB, noise floor at stage output dBu, clip point dBu
    ("mic preamp",  48, -80, 24),
    ("high-pass",    0, -95, 24),
    ("equalizer",   -2, -95, 24),
    ("compressor",   0, -95, 24),
    ("mixer",        0, -95, 24),
    ("output gain",  2, -90, 24),
]
def run(start_dbu, name):
    level = start_dbu
    print(name)
    for st, g, noise, clip in stages:
        level += g
        flag = "CLIP" if level > clip else ("near noise" if level - noise < 40 else "ok")
        print(f"  {st:12s} {level:6.1f} dBu  headroom {clip - level:5.1f} dB  above noise {level - noise:5.1f} dB  {flag}")
run(-46, "Normal speech, gain set at the preamp:")
stages[0] = ("mic preamp", 28, -80, 24); stages[5] = ("output gain", 22, -90, 24)
run(-46, "Same speech, preamp too low and made up at the output:")

Normal speech, gain set at the preamp:
  mic preamp      2.0 dBu  headroom  22.0 dB  above noise  82.0 dB  ok
  high-pass       2.0 dBu  headroom  22.0 dB  above noise  97.0 dB  ok
  equalizer       0.0 dBu  headroom  24.0 dB  above noise  95.0 dB  ok
  compressor      0.0 dBu  headroom  24.0 dB  above noise  95.0 dB  ok
  mixer           0.0 dBu  headroom  24.0 dB  above noise  95.0 dB  ok
  output gain     2.0 dBu  headroom  22.0 dB  above noise  92.0 dB  ok
Same speech, preamp too low and made up at the output:
  mic preamp    -18.0 dBu  headroom  42.0 dB  above noise  62.0 dB  ok
  high-pass     -18.0 dBu  headroom  42.0 dB  above noise  77.0 dB  ok
  equalizer     -20.0 dBu  headroom  44.0 dB  above noise  75.0 dB  ok
  compressor    -20.0 dBu  headroom  44.0 dB  above noise  75.0 dB  ok
  mixer         -20.0 dBu  headroom  44.0 dB  above noise  75.0 dB  ok
  output gain     2.0 dBu  headroom  22.0 dB  above noise  92.0 dB  ok

The second run shows the problem Wendell found at Riverside: with the preamplifier 20 dB too low, the signal through the middle of the chain sits 20 dB closer to the processing noise, and the extra 20 dB of output gain lifts that noise along with the voice. The page leaves the DSP at the same level as before, but about 75 dB above the noise instead of about 92: some 17 to 20 dB noisier, which is easily heard as hiss in a quiet room.

Signal flow drawings

Every AV design includes signal flow drawings, also called line diagrams or schematic diagrams. They show each device as a block with its inputs and outputs, each cable as a line with its identifier and type, and where each cable runs. The installer uses them to pull and label cable, terminate connectors, connect the rack, and later troubleshoot. Cobalt Line's labeling follows the design's scheme: each cable carries a number that appears on the drawing, the cable schedule and the labels at both ends, and each connector panel in the rack is labeled with what lands on it. When the installer makes a change in the field, such as using a spare cable because one was damaged, the change is marked on the drawing that day so the as-built record is correct at handover.