Why record audio in high resolution

I always ask my clients to record or export their tracks in better than 44.1 kHz 16 bit resolution, preferably in 48 kHz 24 bits. A very common reaction to my suggestion is “why should I bother with this if the final master will be in CD audio format anyway?”. Here are the reasons why you should record audio in high resolution.

The quality of a digital audio recording is technically determined by the sampling rate and bit depth used during recording (we’re not speaking about mics, preamps, interfaces etc. now). If you want to reproduce an analogue sound source digitally whose maximum frequency ranges up to 20 kHz (the upper limit or human hearing), you’ll have to use a sampling rate of 40 kHz at least (see the Nyquist–Shannon sampling theorem for more info). Since no A/D converters are perfect 44.1 kHz was chosen as the standard in many aplications, including CD audio. This means that your A/D converter will capture 44,100 discrete samples from the continuous analogue signal per seconds.

Now let’s see bit depth. Bit depth can be referred to as the resolution of a discrete sample. The CD audio format can contain up to 16 bits of information per sample. This determines the maximum dynamic range and signal-to-noise ratio. In 16 bit audio this is theoretically 96 dB, in 24 bit audio it’s 144 dB. So when for example compression or equalization is applied during the mixing and mastering session you further limit the dynamic range and lose bits, i.e. you degrade audio quality.

A good way to demonstrate the negative side effects of this is to use an example of digital image processing.

I took this high resolution image:

original high res image

First I applied some color and tone corrections on it (just like equalization in audio) then I applied a “cut out” filter which reduced the details with a nice character (similar to compression). Here’s the result:

01_processed_before_resample

What would’ve happened if the image were resized before applying the corrections and filters on it? Take a look at this picture:

02_processed_after_resample

Comparing it to the previous one you’ll find it blurred and less detailed. And that’s the point.

If you record audio in high resolution much more detail and transparency can be preserved during mixing and mastering. So when your song is converted to the lower resolution target format in the very last step you’ll lose much less from the sound quality than if it were recorded in CD audio format right in the beginning.

Why record audio in high resolution

Sample rate conversion explained with SoX examples

Sample rate conversion is a big dilemma that comes up quite often. You know you need it sometimes but not sure when and why. This entry will help you out with real world examples using one of the best – and actually free – sample rate converters called SoX.

The sampling rate of an audio file is very similar to the frames per second of a film. For example in CD audio you have a standard sample rate of 44,100 Hz, which means that each second of the audio file is “built up” using 44,100 samples. These samples are captured during the analogue-to-digital conversion when a continuous analogue signal is converted to discrete samples.

What samlple rate should you choose for recording your tracks?

The upper limit of human hearing is 20,000 Hz. According to the Nyquist-Shannon sampling theorem you can restore the original analogue signal without information loss if you choose a sampling rate which is at least the duplicate of the highest frequency present in your analogue signal. Otherwise some ugly distortion will occur (called aliasing). So if your analogue source goes up to 20,000 Hz, you will need a sampling rate of at least 40,000 Hz; the standard has been set to 44,100 Hz.

Why to choose higher than 44,100 Hz sampling rates?

When your tracks are processed with digital processors (including software plugins), these tools will usually work more precisely (i.e. with less artifacts) at higher samlple rates thus the results will sound better.

I usually suggest recording audio files at a sampling rate of 48,000 Hz and keep this rate until the end of the mastering stage, then convert in the very last step.

The final sample rate of your songs will be determined by the standard requirements of the medium it will be published on. For example if you’re planning to release your music on DVD, you should use a sample rate of 96,000 Hz from the very beginning. However, most often a CD audio format will be needed too, so conversion will most likely be necessary. And here can things go pretty ugly…

The built in sample rate converter in most DAWs are not that top-notch at all. If you simply load your files in your DAW and export them at a lower sample rate setting you will probably lose sound quality. Fortunately there are many great tools to overcome this issue including my go-to high-end sample rate converter called SoX, which I think is the best free sample rate converter. But using it can be tricky as it has no graphical user interface. But no worries, here comes the solution:

  • Download a copy of the latest SoX version for Windows here: http://sourceforge.net/projects/sox/files/sox-win/
  • Extract the files or install the program (I prefer zip packages but it’s up to you)
  • Unzip the archive below and copy the files to the directory where you unzipped/installed SoX:
    sox-examples.zip
  • Create shortcuts to the .bat files on your desktop (right click->send to->desktop (create shortcut))
  • Now simply drag the wave files (yes, you can add more simultaneously!) to the desired shortcut icon
  • The converted files will be put to a directory named ‘converted’ within SoX’s folder. Each example is set to the best available very high quality conversion.

In my examples there’s a file called ‘44.1kHz_16bit_dither_noise_shaping_VHQ.bat‘, use this only for final masters: no audio processing should be done to the files afterwards!

If you wish to convert to a different sample rate just open the bat files with a text editor and change the number 44100 to whatever you’d like to.

For advanced users: I added the -G option to every conversion to prevent clipping. If you want to leave the gain unchanged simply delete the -G switch in the bat files using a text editor.

If you’d like to see how SoX actually performs in sample rate conversion compared to your DAW or even high-end hardware boxes then take a look here: http://src.infinitewave.ca/

Happy converting!

Sample rate conversion explained with SoX examples

Stereo image fix after applying analogue processing

Running stereo buses or full mixes through analogue devices like compressors, equalizers, summing boxes etc. is fun. They can add that magical touch to your track – a bit of stardust, the so-called 3D effect, punch, fullness and all sorts of adorable favours. But many times you face a problem even with high-end devices: the stereo image shifts…

I was mastering a track recently which has been run through a very expensive channel strip unit to add some shine to the mix. Sounded gorgeous except for that the mix – most noticeably the bass and the vocals – slightly shifted to the left, which was kind of annoying.

I asked my client to send me the mix without the channel strip processor applied on it. Then I measured the RMS levels for the left and right channels on both versions of the song (you can use Voxengo SPAN for such tasks). It looked like this:

Original (in the box) mix: Left: -18.2 dB, Right: -18.3 dB
Mix run through the channel strip: Left: -19.4 dB, Right: -20.4

You can see that the right channel has become 0.9 dB quieter compared to the relative levels of the original mix. To compensate for this I used the free Stereo Tool VST plugin from Flux where you can adjust the levels of the left and right channels independently (besides many other options). I increased the level of the right channel by 0.9 dB: now the vocals and the bass came back to the center. Problem solved!

Just a final note. Always check the mix with your ears too because RMS values are sometimes misleading. So don’t compensate for the stereo image shift in the mix blindly.

Stereo image fix after applying analogue processing

29 mixing tips to improve sound quality

Here are some tips for mixing your songs which will help you to achieve more clarity and definition. None of this is craved into stone but can help you out in many situations.

  1. When using equalizers it’s usually better to cut than boost
  2. You can achieve more clarity if you high-pass every track around 80 Hz, except bass instruments
  3. Use a wide Q setting for equalizers unless you need surgical adjustments
  4. If you need more punch for your bassdrums boost high frequencies instead of lows which will “sharpen” the attack
  5. Use small amounts of reverb – it’s better to feel it instead of hearing obviously
  6. Shorter reverb settings contribute to more clarity in the mix
  7. Add both reverb and delay to pads to make them sound bigger
  8. Add a little ambience reverb to dry sounding drums and bass tracks
  9. Filter the delay effects with high- and low-pass filters to avoid your mix sounding busy
  10. It’s generally better to use a low compression ratio for buses
  11. Applying parallel compression on drums and vocals will result in a more powerful sound
  12. Watch drum transients – make them sound snappy instead of smeared
  13. Organize instruments into groups and apply bus compression instead of compressing every single track
  14. Don’t mix too hot, apply proper gain staging by setting your channels to 0 VU using an analog style VU meter (prior to mixing)
  15. Use an autopanner on some of the tracks to make them sound more alive
  16. Pan backing vocals to sides
  17. Avoid hard panning of single instruments
  18. Spread instruments in the stereo space
  19. Place lead vocals, kick, snare and bass instruments in the center
  20. If an instrument sounds too narrow apply a 10-20 ms channel delay between the left and right channels
  21. Avoid the “big mono” effect when every sound comes from two sides
  22. Don’t invert the phase for only the left or right channel of a track – this will lead to mono compatibility issues
  23. Fix smaller noise issues with a noise gate instead of a denoiser plugin
  24. Keep your hihats lower in level, too loud is ear fatiguing
  25. A little chorus on the bassline can make it sound fuller
  26. Check your mixes on multiple systems: e.g. on headphones, in the car, on laptop speakers etc. to make sure they translate well
  27. If you don’t have a decent monitoring system use a spectrum analyzer to spot errors in the frequency balance
  28. Build up the mix: start with the drums, then the bass, the vocals, lead instruments and finally the background elements such as pads, strings. This will result in better balancing.
  29. Always compare your finished mix to commercial releases
29 mixing tips to improve sound quality

saving-a-mix-containing-weak-mp3-samples

I recently worked with a DJ who remixed an early 80′s song. There were two transitions from the remix to the original song which was a low quality, mono mp3 file. So when the song played the original parts, the previously wide and dynamic sound collapsed, it sounded thin and grainy.

First I tried to revitalize dynamics and match the frequency curve of the different parts. Then I split the original parts to 3 frequency bands and applied a short 10-12 ms left-right channel delay on the highs. And so it sounded awful… Although this technique often works fine with individual tracks this time it did not. The original parts did not have enough clean air to work with and even worse, the instruments and the vocals just did not come apart.

I thought it once again and found a very simple yet wonderfully working solution. I routed the original parts to two different busses. On the first buss I made the required dynamics and EQ adjustments so the samples sounded bigger and cleaner. On the second buss I set up a tempo synced stereo delay 100% wet, and processed the echos to sound smooth and clean with a weak low end. I turned the volume fader all the way down and then started to slide it upwards very slowly. At a very low setting the mix started to sound quite fine. The original samples had the same punch as the newly produced tracks and a nice stereo width without hearing any echos distinctively. Job done!

saving-a-mix-containing-weak-mp3-samples

Free VST plugins for professional use

Well, not a mixing technique this time but a very useful entry for many of you I’m sure!

I’ve always been very critical about selecting the best tools for my studio which ensure a seamless mixing and mastering workflow. Through the years I’ve came across several freeware VST plugins and other free audio software too. Here’s a list of the free software I use regularly and which are on par with the most expensive ones:

A very musical sounding HP/LP filter, great for mastering as well:
http://www.brainworx-music.de/en/plugins/bx_cleansweep_v2

One of the best software audio analyzers:
http://www.voxengo.com/product/span/

A general purpose mid-side plugin:
http://www.voxengo.com/product/msed/

My go-to stereo tool and analyzer:
http://www.fluxhome.com/products/freewares/stereotool-v3

A great tool for widening the stereo image:
http://www.voxengo.com/product/stereotouch/

A utility for delaying left/right or mid/side channels:
http://www.voxengo.com/product/sounddelay/

A neat autopanner:
http://www.meldaproduction.com/freevstplugins/mautopan.php

Delay effect with a very good sound:
http://www.e-phonic.com/plugins/retrodelay.php

And finally one of the best sample rate converters available (not a plugin and runs from command line; a short tutorial will follow this entry soon):
http://sox.sourceforge.net/

Free VST plugins for professional use

Mixing with headphones

Although mixing with headphones is something I wouldn’t recommend to do there are situations when you’ve got to go for it. So let’s see what’s wrong with it and how you can compensate for the disadvantages.

When you listen to music through headphones you experience a quite unnatural ‘in your head’ sound. First, room acoustics is totally excluded which results in a dull and fully dry sound. Second, the sound doesn’t come from the front but goes directly into your ears which causes total stereo separation. Also, your ears respond to sound pressure coming from speakers differently then when coming from headphones. That’s quite bad so far… But here comes the good news!

To compensate for the undesired side effects of listening to music through headphones you can use virtual speaker simulation tools. Most of them operate with head-related transfer functions (HRTF) under the hood to simulate acoustical transfer from loudspeakers to ears and reflections from the human body, and binaural room transfer functions (BRTF) to simulate room reflections. For me the most convincing results were produced by the now discontinued Focusrite VRM box. My second choice would be TB Isone by Toneboosters which is a binaural room simulator in VST plugin format. Other options cover (not exclusively) 112 dB’s Redline Monitor, G-Sonique’s Monitor MSX5 or Beyerdynamic’s Virtual Studio. For best results you should use high quality headphones with a full band flat frequency response especially designed for pro audio applications.

Still, I wouldn’t solely rely on such a tool during a mixing session. However, I would recommend using it to double check your mixes to hear how they would sound in various listening environments. Also, if you don’t have a decent loudspeaker based monitoring environment you’ll most probably produce better-translating mixes by using a speaker simulator when monitoring through headphones.

Mixing with headphones