Upsampling Audio – Importance and Methods in Audio Mixing/Mastering

Upsampling is increasing the resolution & sample rate of the original digital audio. Some audio professionals call this “over sampling” although I prefer to call it up-sampling. For example, if the original audio is 16 bit/44.1Khz. If you like to convert it from 16 bit/44.1Khz to a higher resolution such as 32-bit float/96Khz; the process is called “up-sampling”. Upsampling is important in audio editing process particularly in audio mixing and mastering. The primary reason is that digital effects (used by the software plug-in or built-in with your recording software) introduced some kind of unwanted artifacts or distortion during audio processing. This audio processing can be compression and other non-linear editing. This will produce some aliasing effects on high frequency components of the original signal.

Demonstration of how artifacts are generated in a 16-bit/44.1Khz digital audio when applied with effects:

1.) Launch any recording software, (for example I use Adobe Audition).
2.) Go to Edit view, and go to File – New, Create new waveform. Assign sample rate and bit depth of 16-bit and 44.1 KHz. Select mono.
3.) Generate a 10 KHz sine wave tone. Go to Generate – Tone. Under “Flavor”, select Sine.

4.) Click OK and zoom, you can then see the sine wave generated at audible frequency of 10 KHz. See screenshot:

This is also the original signal frequency spectrum before any effects are applied:

5.) Now let’s apply compression to this 10 KHz sine wave at 16-bit/44.Khz resolution. For this example, I will be using the Waves C1 compressor. Select the entire wave and apply “Classic compressor” preset. Press OK.

6.) Select the entire wave and do Fast Fourier Transform (FFT) analysis. This will let you view the entire signal frequency spectrum to assess whether there are artifacts or side-effects generated by the compression. In Adobe Audition 1.5, this can be done by going to Analyze – Show Frequency Analysis. Select the maximum FFT size (65536) and use Blackmann-Harris. Finally click “Scan”. This is the result when compressing directly a 16-bit/44.1Khz sine wave (10 KHz tone).

Enclosed in yellow boxes are the artifacts generated by the compression. These are unwanted results or it’s also called “quantization distortion”. In real music, this will have an effect of degrading the audio quality particularly if you are over-processing a 16-bit/44.1Khz digital signal with a lot of effects or compression.

The Solution: Upsampling == > “Effects processing” == > Down sampling == > Dithering

Now you have seen the artifacts generated if you are applying effects directly over a 16-bit/44.1Khz. Let’s illustrate how you can avoid this. These are the steps:

1.) Generate the 10 KHz sine wave tone at the original sample rate and bit depth of 16-bit/44.1Khz, mono. Refer to the previous section for the procedure.
2.) Before you will apply any effects, “up-sample” first the digital audio. In Adobe Audition 1.5 (which I am using), you can simply go to “Edit – Convert Sample Type – Select “96000” under Sample rate – Select “32” bit under resolution. Select “mono” for channel. Make sure pre/post filter is checked and set to 999 for high quality. Finally click OK. This is the screenshot of the convert sample type settings. If you are using the more recent versions of Adobe Audition, you can refer to the manual and the process should still be similar.

The resulting wave still sounds and looks the same, except that it is now in high resolution 32-bit/96Khz.

3.) After up-sampling, apply the same “classic compressor” preset from Waves C1, and then click OK. Save the processed audio file as testing.wav. The saved audio wave file should retain its sample rate and bit-depth at 32-bit/96Khz. Close the file in Adobe Audition.

4.) Using Voxengo sample rate converter; down sample first the audio from 96 KHz to 44.1 KHz, just leave the bit depth at 32-bit float (do not changed it). This is my Voxengo R8brain settings:

5.) Open the converted file from Voxengo in Adobe Audition (e.g. testing_r8b.wav). Now dither it from 32-bit float to 16-bit. I use Waves IDR as the dithering plug-in. After applying dithering, you have successfully converted back the audio to 16-bit/44.1Khz.

6.) Now let’s see if there are any artifacts generated. To confirm; re-perform the FFT analysis; this is the result:

Thus you have noticed that there are no more artifacts/significant distortion on the compressed wave unlike when you apply effects directly on a 16-bit/44.1Khz wave. It looks smooth and clean. This is because the original audio of 16-bit/44.Khz has been up-sampled first to a higher bit-depth/sample rate before applying effects. Then it is down sampled to its original bit-depth/sample rate. The result is zero distortion, digital artifacts or aliasing. You can implement this technique whether you are working with digital audio in mixing and mastering.

Content last updated on August 15, 2012