About Digital Calibration

Why calibrate? | Sample Offset | Latency | Test Setup | Conclusion | ^ Studio Page

Why calibrate?

In the age of analog tape, of which I only caught a last short glimpse, every master tape started with a 1 kHz test tone to align the equipment and make sure the material would play back as intended. When digital recording arrived, this was no longer deemed necessary, and in fact it is much easier now to play back a file with consistent quality. But to assume that "digital" automatically means "correct" is tempting fate.

This little article was written to show you two easy ways to calibrate your digital equipment, enabling you to work independently and trust your own ears and eyes, regardless of what a manufacturer prints on their spec sheet.

We need only two tools here, pink/white noise and time-align clicks, which can be used to get to a near-perfect timing situation when you mix the tracks in your favourite digital audio workstation (DAW).

Pink or white noise are random forms of sound, distributed over the whole frequency spectrum. The power is equal for all frequencies for white noise, and decreases by 3 dB per octave for pink noise. Ideally, you pick the noise type that results in a flat line on the spectrum analyser (see graphics below). You can use this for various calibration tasks. The basic idea is that if the evenly spread noise comes out unevenly, something has gone wrong in the signal chain. This can show up in various ways, for instance as a low-pass filter, a comb filter, or a notch filter of various orders, altering the noise spectrum in a way that you can both see and hear.

A time-align click is a very short impulse that goes through all channels, allowing you to get your channels in sync. If you have a digital interface and you hook up other gear to it, say, an ADAT and an AES/EBU device, particularly if you don't use a common clock source, you are likely to end up with phase and timing problems.

In principle, this is nothing new. The good old 1kHz test tone and the slate in the film workflow do essentially the same. The difference is that in the analog domain, latency is not that much of an issue. Pink/white noise and time-align-clicks are not a standard technique in digital recording, mainly because many people think that digital is so great it doesn't need calibration, or are not yet aware of the latency and sample delay issues that we constantly have to deal with.

Sample Offset

If a stereo signal is offset by n samples, you run into problems if the signal runs through a mono output. The phase of the waves, especially in the top end, gets confused and will be cancelled out. This varies with the frequency of the audio, as well as the sample rate. This has little or no effect as long as we stay in the stereo domain, but is very audible with a mono output or when summing/mixing multiple channels.

The following example pictures demonstrate the effect of various sample delays at 44.1 kHz sample rate. The mirror point for the audible effects is the Nyquist frequency, i. e. 22.05 kHz for a 44.1 kHz sample rate, 48 kHz for a 96 kHz sample rate, and so on. Consequently, for a 96 kHz sample rate, you won't get an audible low-pass filter for a one sample delay, but the comb filtering will eventually reach the audible frequencies as the offset increases to 2 or more samples.

I've created a short example file for demonstrating the low pass filter created by a 1 sample delay at 44.1 kHz SR. The integrity of the noise is less than perfect, but it should be good enough for picking up the effect. The first 5 seconds show left and right channel out of sync by one sample, the last 5 seconds show L and R in sync.

In general, this is the most common problem in digital signal chains right after latency and pops & clicks, but much easier to fix. All you have to do is delay one of the channels, until you see a ruler flat test noise curve, where the test noise sounds (and looks) the same as the test noise that came out of the generator. This allows for optimum calibration of the digital signal path along multiple channels.

Latency

Latency is the same as delay, only that we use this term for a different situation. Latency generally refers to a processing delay between digital devices as the samples travel from one interface to another or as any kind of calculations are performed on the audio data. This is easier to resolve than the previous problem—all it takes is an impulse which you can use to align the tracks, whether they be mono or stereo or whatever.
The technique is very simple: At the beginning of a recording session, you pass a short impulse through all input channels. This way, you can acoustically and visually align the timing of the channels later in your DAW.

Test setup

For the noise part, simple connect a test tone generator to all of your channels and play the noise into all of them, using a splitting cable or mixing console. It is important that the same noise reaches all inputs at the same time and, ideally, with the same volume. The test tone generator can be a dedicated device or even your DAW, if it can output a test signal and record it on different channels simultaneously.

In the noise part, we are only calibrating the electrical side of the signal flow. To go into the details of compensating microphone positions in the acoustic domain would imply a rather complicated setup, which exceeds the scope of this article.

The time-align click goes through the same signal chain, only that you use your test tone generator to produce a short click—any short impulse will do. Think of it as the electrical equivalent of a slate, as used in film work. Be careful not do overload the inputs. With the time syncing, it is much easier to repeat the same for the acoustic domain, simply use a real slate at the very same position of the sound source(s).

Conclusion

While this may all sound trivial, it is the source of much confusion around digital recording. I suggest that you start a session with a test signal that consists of a mono pink noise source, sampled by each channel, as well as a time-align click.
This is most useful when you are using a new connection or new equipment, or a signal chain that is not known to work perfectly. The only disadvantage is the addional 2 minutes that you need to set up the test generator. The advantages are true phase coherence across all channels, as well as optimum latency control. For a setup that you are using regularly, this has to be done only once.

By the way, test generators are easy to find in both software and hardware versions. Most DAWs feature a test generator as one of the built-in plugins, or otherwise you can easily find them as a standalone software. Alternatively, you can achieve the same by obtaining a hardware test generator that includes pink/white noise.

Disclaimer

This is mostly a lab test scenario and has little or nothing to do with what you hear in a real-world recording situation. In the test examples, L and R channels contained the exact same material and were summed to mono. In a real-world scenario, phase differences and cancellations may contribute to a certain sound style as much as the music itself, so not everything that can be measured as being out-of-phase is necessarily bad. We should continue to trust our feelings: Whatever sounds good, is good!

This article is not targeted to cast any negative light on any products on the market. It only demonstrates a easy method to calibrate your digital equipment in a way that makes you completely independent of what any given manufacturer is inclined to tell or not to tell you.

I'll be happy to answer any questions your might have to the very best of my knowledge, just drop me an e-mail.