Best Practices in Q-SYS Gain Structure (Part 1)

Achieving good gain structure is a key component

to getting the best audio performance out of your Q-SYS system.

Unfortunately, there is a lot of misinformation out there about how to achieve it

so we are going to start from scratch and correct some of those bad habits.

So let’s start off by establishing some basic vocabulary.

“Clipping” is the point at which you can no longer reproduce a signal

that looks or sounds like what is usable.

The “noise floor” is the base electronic noise of the circuit.

The signal level between those two points is what’s called “dynamic range”.

The typical pro audio device has a clipping point of about +24.

Both Q-SYS and most other DSPs on the market will clip +24 dBu.

For example, If you look at the Q-SYS CIML4 input card,

the noise floor around -81 dBu,

so that means your “dynamic range” is greater than 105 dB assuming clipping at +24.

If the program material in the audio system is dynamic,

there could be a lot of variation in the ‘average’ signal level and ‘peak signal level’.

We want to make sure that peaks will not clip the audio input.

The space between the average signal level and clipping is called the ‘headroom’.

Now remember, our ears are not linear.

In general, it takes about a 10 dB increase in SPL to be perceived as twice as loud.

If we allow for program peaks to be four times the average signal level,

which is the industry standard, this means a headroom of 20 dB.

While we want to leave headroom to avoid clipping,

we also want to make sure the signal into the device isn’t noisy.

The amount of signal vs. the amount of base noise in the circuit

is called the Signal to Noise Ratio, or SNR.

So in summary, what we want is a signal that is sufficiently above the noise floor,

but leaves the right amount of headroom before clipping.

We refer to this as ‘nominal’ level.

This is the ‘sweet spot’ between those two considerations.

In standard pro audio circuits, that would be +4dBu.

Consumer level audio circuits are typically -10dbV nominal (with respect to 1V).

However, not all of the audio sources that we bring into Q-SYS are going to be +4 nominal, right?

Microphones, for example, typically come in at around -40 nominal.

This is why Q-SYS and almost every other DSP makes use of an analog preamp.

This allows you to increase or decrease the gain of that device

before it flows into an A-to-D converter,

at which point we jump from the analog domain to the digital domain.

An analog preamp typically provides the most control with the least noise.

A-to-D conversion takes small samples of the analog audio waveform,

where the signal enters the ‘digital’ domain.

These samples represent the magnitude of the waveform at that point.

A signal is typically sampled at a fixed rate.

According to the Nyquist Theorem,

you must sample at twice the highest frequency that you plan to reproduce.

As long as we're sampling at least double that 20k upper limit,

we shouldn't experience aliasing or have incorrect sample wave form.

A Q-SYS sampling rate for example is 48k.

Different A-to-D converters will have different supply voltages, so the input signal is scaled down from that value.

That introduces the idea of dBfs, or full scale.

Many integrators get confused because certain DSP manufacturers

represent their input meters in dBu and train on gain structure accordingly,

recommending 0dBu nominal input.

Other DSP platforms such as Q-SYS use input meters in units of dBFS, not dBu.

So if you apply that other knowledge in Q-SYS,

every time a microphone has a peak, it would force it into clipping.

Units really matter in this situation:

if you look at +4 dBu nominal as compared with +24dBu clipping,

if we leave +20 dB of headroom in our digital system (0dBFS clipping),

our nominal level is going to be -20 dBfs.

If you're a Biamp user and you set that to zero then you've essentially eaten all that headroom

because your nominal level is essentially 0 dBFS.

In Q-SYS meters, our nominal becomes -20 dBfs.

Of course, after we take samples, we can do anything we want, right?

Q SYS is a floating point DSP, which means that I can apply an infinite amount of gain and attenuation.

The alternative would be a fixed point DSP,

which means that introducing gain beyond a certain amount of dynamic range will start to introduce error.

In Q-SYS, you can only clip samples at the conversion point to and from the analog domain,

however you can apply infinite attenuation or gain.

Alright, let's take a break right there and we'll come back for the next section.