When to use a high pass filter in the sidechain of your compressor

When to use a high pass filter in the sidechain of your compressor

If you’re using a compressor to make a track in your project sound loud and upfront, then choosing whether or not to use a high pass filter in your sidechain could be important.

Why? Because how your compressor perceives volume may not be the same as how the human ear perceives volume. In this post, I explain why it’s sometimes a good idea to put a highpass filter in the sidechain of your compressor.

Let’s look at the Fletcher-Munson curves:

Lindos4

(I’ve taken this graph from https://en.wikipedia.org/wiki/File:Lindos4.svg used in the article https://en.wikipedia.org/wiki/Fletcher-Munson_curves – the work is listed as Public Domain).

The “equal loudness contours” tell us that if we’re listening to a 1 kHz sine wave at 60 dBSPL, then a 100 Hz sine wave will have to be about 10 dB louder to be perceived as being the same volume! Or, looking at it the other way round, a 1 kHz sine wave would have to be about 10 dB quieter to be perceived as being the same volume as a 100 Hz sine wave.

Of course, these curves depend on who is listening, and how loud the sound is.

And we rarely listen to individual sine waves on their own, as most sounds consist of lots and lots of harmonic partials. Sound in the human ear is detected by outer hair cells in the cochlea, which mechanically encode an extremely narrow frequency range. So although I’m definitely not an expert on this, it certainly does seem like you can think of the human ear as containing a large number of sine wave detectors: http://auditoryneuroscience.com/?q=ear/dancing_hair_cell

These equal loudness contours also explain why the mid range is so critical to a mix: the balance of instruments in the mid range of frequencies (very roughly 500 Hz to 5 kHz) defines the mix.

Your compressor

Let’s look at a massively simplified model of how a compressor works:

Compressor1

Audio goes in. The audio signal gets split in two: one branch goes towards the sidechain detector, and the other branch goes towards the output via the big red circle. The red circle is just a volume control which may or may not be reducing the volume at any moment in time.

The sidechain detector decides how loud the signal is right now. (This may even be fed by an outside source instead of the incoming audio signal, the classic house example being sidechaining a kick drum to duck a whole mix). Then the gain reduction element decides how much the volume of the signal should be reduced at any moment in time.

A standard sidechain detector, without filtering of any kind, will treat sounds of any frequency in exactly the same way. For example, a 1 kHz sine wave which peaks at -12 dBFS, and a 100 Hz sine wave which also peaks at -12 dBFS, when fed through the same compressor one after the other, will both induce the same amount of gain reduction. Not only that, but both of these signals both have exactly the same RMS power of -15 dBFS. But the 100 Hz sine wave will sound much quieter to human ears.

So if our sidechain detector fails to take equal volume contours into account, the compressor may compress low pitched notes or sounds more than it should.

For example, say you want to compress the lead vocals in your project. You want the vocals to stand out at the front of the mix, loud and clear. You load up your favourite compressor plugin (or patch in a hardware compressor if you’re rich) with no sidechain filtering. When the singer sings lower pitched notes, you lose perceived volume.

Another example: you strap a compressor over the master bus of your mix. There’s an eight bar intro with vocals and guitar. But when the bass comes in, the mix goes quiet.

The solution

The solution is to modify our sidechain detector to include a high pass (low cut) filter. Many compression plugins let you do this very easily. We usually want a moderate high pass filter with a steepness of 12 dB / octave. We don’t want to remove the low frequencies – just lower their volume. The corner frequency of the filter will depend on the sound we’re compressing and what we’re trying to achieve.

Remember that we’re not filtering the sound you hear coming out of the compressor, only the sound that controls the compressor’s behaviour:

Compressor2

A note on high pass filters: the only frequency that a high pass filter truly removes is 0 Hz (DC offset). A 12 dB / octave high pass filter means that frequencies below the corner frequency will still be audible for several octaves below the corner frequency.

An extremely contrived example

I’ve programmed a very simple synth part which plays triangle waves that are two octaves apart.

I’ve compressed this synth part using Waves RChannel with a ratio of 4, attack time of 0.00 ms, release time of 30 ms, and I’ve set the threshold so that the compressor is doing approximately 3 dB of gain reduction. (Note that RChannel uses lookahead in its compressor, so you can use extremely fast attack and release times without distortion). Here’s what it sounds like:

 

Note that the lower pitched note is much quieter than the higher note. The two notes have almost identical peak level when they come out of the compressor, and are being compressed by the same amount of gain reduction (approx 3 dB).

And now I keep all the settings identical, but I add a 500 Hz high pass filter into the sidechain of the compressor. The high note is still compressed by the same amount, but the low note ends up being compressed less:

 

This is much better – the two notes sound like they have the same volume, even though in reality the higher pitched note has a much lower peak level. The higher note gets compressed by approximately 3 dB, and the lower note gets through with only a fraction of a dB of gain reduction.

So when should I use a high pass filter in the sidechain?

  • In this post I’ve mainly discussed using sidechain filtering for pitched instruments, in particular when the fundamental pitch of any of the notes played moves into the low mids or below.
  • Pitched bass instruments (bass guitar, synth bass) can sometimes benefit from a high pass filter in the sidechain. Your initial reaction to this might be “How can the compressor compress at all if you remove the fundamental frequencies from the sidechain?”. Well, to start off with, the high pass filter doesn’t remove low frequencies, it just lowers their volume. If you can get a filter shape which matches the inverse of the Fletcher-Munson curve, you can have very even-sounding bass. However, this depends on the tone of your bass, and sometimes may not be a good idea.
  • For compression on buses / masters, a high pass filter in the sidechain can also help greatly.
  • For non-pitched instruments (ie, drums) the sidechain filtering can affect the resulting amplitude envelope, but this is beyond the scope of this post.
  • Sidechain filtering in general is a useful technique for dealing with problem frequency ranges (eg, deessing), but this is also beyond the scope of this post.
  • Multiband compression / dynamic EQ is also an incredibly powerful (or gentle) tool that can work very well, especially on basses, but again is beyond the scope of this post.

Ultimately, you have to use your ears.