The Truth About Recording Levels

So Now What?

The problem with “casual” tests is they’re not scientific. As scientists like to say, the plural of anecdote is not data. Nor am I convinced by the claims of “dozens and dozens” of people who clearly never did a proper comparison either.

Audio forums are filled with anecdotal reports of “tests” that prove all sorts of nonsense. Other than measuring, there’s only one way to know for sure if preamps or converters or anything else sound different at different levels, and that’s by performing a controlled blind test.

That’s exactly what I did to create the mixdown files linked above. In this case I used the preamps in my Focusrite Scarlett 8i6 set 20 dB apart to put the record levels 20 dB apart. The “controlled” part was recording the same source simultaneously through both signal paths. And when you download and play the Wave files without knowing which is which, you’re listening blind which further legitimizes the results.

But the real proof is that the null residual is 50 dB below the music. A null residual that soft ensures that the tracks sound the same, even though the preamp and converter were hit 20 dB harder when recording one set of tracks.

If I had a way to match the levels of each track pair more closely, the null residual might have been even softer. A volume control having 0.1 dB steps can create a null that’s at least 38.8 dB down, but you need 0.01 dB resolution to match levels closely enough to obtain a null 58.8 dB below the main signals. Similar math proves that a null residual 50 dB down ensures the files are within 0.025 dB of each other at all frequencies.

So why do people swear they hear a change in sound quality when no such change is likely or even possible? According to the cited article, preamps and other audio gear respond with different coloration at different signal levels:

Analog gear (preamp, compressors, outboard signal processors, etc.) is … more like “Perfect, a bit unfocused, a little noisy, ‘tight’ sounding, spectrally distorted, CLIP.”

This is a classic example of how badly flawed our hearing perception really is. It’s not real. It’s Fletcher-Munson at work.

It’s the masking effect at work. It’s expectation bias and the placebo effect. It’s the short-term nature of auditory memory. It’s moving your head three inches to the side, which causes a surprisingly large change in the response reaching both ears. It’s comparing two different performances rather than one performance recorded two different ways at the same time. It’s failing to listen blind, so of course you can always tell which is A and which is B.

Indeed, it’s everything but what’s actually happening inside the audio circuits!