Advanced Compression
Intro
It took me a minute to get this blog together as it actual represents a real chapter in my yet to be published book “Painting The Picture Of Music”. Most of you, the readers, will find some information which has already been published earlier. For example M/S Compression and Dynamic-EQ. However this blog will go a bit more deeper into the magical, but above all, beautiful world of side-chaining. I guarantee you that you can’t live without it by the time you have finished reading this blog.
Over Compression
You all have read about this earlier in one of my blog post. But today I’m talking about the sense of setting a lower threshold than you need to get the job done. I though to start of with this before I even dig into advanced compression/ side-chaining.
As you know over compression will always make the sound worse, with the sole exception of percussive sounds (!!!) where it some times be a useful effect. Imagine for example that you have a sound source that plays occasionally but with silences in between. This is where usually over compression is best audible and most likely to happen. When you set the threshold, most likely you know where you’re aiming at of how much gain reduction you want to apply; hear and both say on the meter. This is controlled as we know by the ratio and threshold controls. Lets say we’ve set it in a way that we get 10dB of gain reduction. When the sound source passes through the compressor ask yourself this: “Does it ever go all the way back down to zero?”. If this is not the case and it only goes down to 3dB then you haven’t applied the set amount of gain reduction, in this case 10dB. What you did achieve is 7dB of compression reduction. The other 3dB could have been achieved by simply lowering the fader of your console. The problem starts when the sound source starts to play again. What happens next is that your compressor has to go from unity gain (0) all the way down to the full 10dB. The 3dB we just talked about, that the compressor also has to cover, will result in audible distortion from the compressor and distorts the initial transient. This brings us to rule number one of compression: “At some point in the track, while the audio source is still playing, the compressors’ gain reduction meter must indicate zero, otherwise the minimum reading obtained shows wasted gain reduction and thus over compression leading to the distortion of transients that follow the silences.
High Level Compression
This is most likely the best known usage for a compressor; to increase the apparent loudness of an audio signal. It is called high level compression as it works by reducing the high signal levels (peaks), bringing them down and closer to the low level passages. After this is achieved you apply make up gain to equal it’s input.
The danger however in this technique is that it is very audible and there should be great care in setting up the compressor and compromise between getting enough compression going and not to spoil the overall sound.
Low Level Compression
Ray Dolby is a great example to start the subject of low level compression off with. Roy Dolby told us back in the early A-Type noise reduction system that he left the high levels completely alone and modified the gain only of the signals that fall below -40dB. Every ordinary high level compressor is capable of doing low level compression. A low level compressor will bring up (expand) the audio signal that falls below the set threshold and leaves the high level content completely alone.
Parallel Compression
Another way of doing this is making dupes of the specific audio track. This technique is most commonly known as parallel compression. What you do is mix the uncompressed signal with the compressed signal. At the levels below the compressor’s threshold the two signals will combine to produce a 6dB increase in level. Above the threshold the compressed signal will be progressively reduced and add hardly any additional level to the mix. The outcome is a form of compression where you can get more dynamic range reduction as a result with fewer audible side effects, thus it will sound more human.
Please note that by applying this technique in the digital world you have to be careful of delay issues involved in digital processing, and compensate for the delay else you’ll end up with a mess of comb filtering.
Compression vs. Clipping
We have just talked about over compression and its distortion over it’s long term processes of working at the very least over periods of tens of milliseconds. If however you try to achieve very fast compression by using very short attack and release times, it will result in audible distortion of the low frequencies. What happens is that the compressor changes the actual shape of the waveform and of course you can push the compressor to its very limits up to the point where it has given all it has got to give.
Clipping however does work on a very short time scale. Think about transistorized circuitry which operations are based on the microseconds to any level that is too great for the power supply to cope with and cuts it short, resulting in harsh harmonics, which in addition results in apparent loudness and is called “soft clipping”. What actually happens is that the soft clipping of valve and valve emulated designs rounds rather than clips when dealing with the peaks. When used alone it results in the problem that it will only be affective on high level signals however clip-worthy peaks occur but in high and low level signals. To make full use of soft clipping I highly advice to use a compressor pre and the soft clipper post (use them in series). The reason for this is that compressor evens out general level of the signal, because the compressor works over a comparatively long period time, the peaks re not clipped but simply reduced to a more uniform level. Therefore the soft clipper has more material to work on. To go a step further add this signal to the unprocessed version and apply equalization to the processed version and select the frequency range that will be affected to add just the right bit of distortion without going into the extreme (mid-range).
M(id)/S(ide) Compression
This is one of the subjects I covered earlier in one of my blog posts. For those who missed it here is it again:
We are all familiar with what M/S does and where it stands for (Mid/ Side). M/S compression will give you another angle to how you will tread your stereo buss (mix buss). The M channel is the whole sum of the song, and the S channel represents the difference between left and right.
So M/S compression lets you compress and control the center, and sides of your stereo mix separately. This, all the sudden, allows you to bring up your vocals without affecting your instruments, bringing them back. Taming kick and snare on your overheads, or even emphasizing wide early reflections in the stereo field.
Example: A lot of times you’ve finished a mix, however sometimes the vocal gets slightly buried when the backing track (instruments) get loud. If you go on and try to compress the overall mix, or by the use of narrow band compression of the vocal’s frequency range, you will notice that the compression will ruin the great sounding backing track (instruments). M/S compression will come in handy during this stage. M/S compression can help to isolate the center image (M channel). How? By only compressing the M channel, bring up the center when the signals get loud, or another option is to compress the M channel and expand the S channel. This way you help to control the vocalist and open up the specific band. By compressing the S channel only, anything panned center is unaffected and the compression only affects signals panned left and/ or right that are out of phase. Loud signals in these modes will cause a momentary reduction in level of the S channel and therefore resulting in a narrowing of image width. Another option is multi band M/S compression but I’m not going to touch that option today as it offers more possibilities. If you do work in analogue I strongly recommend to use a stereo compressor (which side-chains are linked, the reason for this is that you don’t want any phase-shifts happening or an imaging change. In L/R compression it’s not guaranteed absolutely zero, as analogue compressors will not 100% handle both channels absolutely equally and therefore some degree of image shift and phase shift might occur. However in M/S compression, any disparity between the channels will not result in any degree of image shift, but in a variation of the width of the stereo image which is less obtrusive than the degree of phase and image shift occurring in L/R compression. I personally would say it works way better than L/R compression and you will find that out for yourself once you’ve played around with it a few times.
To achieve M/S compression simply pass the M signal through one channel of a compressor and the S signal through another. Plugin wise you have to this in two stages.
Advanced Side-Chaining
How To
Let me first start of for those who are fairly new to this technique, how it is achieved. To perform dynamic-equalization (frequency conscious) you’ll need to patch an equalizer (doesn’t matter what quality) into the side-chain of a compressor (parallel a signal so it enters the normal input of the compressor, and at the same is connected to the side chain input via the equalizer). The audio signal you want to process gets patched through the I/O. What you do next is make the compressor more sensitive to the frequencies you want it to dip. On the equalizer you select the frequency or frequencies or bandwidth you’d like the compressor to dip, instead of cutting you give them a boost. The reason for boosting is simply: when the compressor hears an exaggerated response in frequency, frequencies or the band, courtesy of the EQ boost, it will exceed the compressors threshold and make the compressor reduce the level of the audio signal’s specific frequency, frequencies or band. You’re compressor has just become an incredibly flexible and creative EQ. Makes sense right?
The release time of the compressor has to be set right in order to let the compressor attenuate the boosted frequencies, courtesy of the EQ, by setting up the release time so that it releases and returns the track back to unity gain (0) immediately after the frequency has ceased. The attack time has to be set around 50 musec (.05 ms) and an release time of 50 to 60 ms will get the compressor in and out fast enough to attenuate the frequency portion, but leaving the rest of the signal untouched.
Soft Knee & Hard Knee
Lets go beyond side chaining as known to human kind and introduce something very interesting: ‘Distortion Triggering’.
Every one is familiar with compressors soft knee and hard knee settings which in basic words go from immediately (hard knee) from uncompressed to compressed at the exact set threshold rather than the gentle bending of the soft knee types. The knee type is an essential factor in its sound, however very few compressors allow you to modify the knee curve in any way and even if they do it’s a pre-fixed setting for either hard or soft not even a mixture of both which in cases such as parallel compression are very desirable.
So is this it you may wonder? No, there is a trick that enables you to change and control the knee curve of ANY compressor which has a dedicated side chain input which most compressors do have. Instead of hooking up your equalizer hook up a distortion box (e.g. guitar stomp boxes). You won’t hear any of the distortion sound as it’s not in the actual sound path but it functions and is hooked up as a trigger. What happens in the circuitry is that the distortion box sends soft and/ -or hard clipping into the compressor as a control value. This clipping will bend the shape of the knee curve of the compressor itself depending on it’s operation mode and type: peak/ RMS detection. It will result in a way more musical and different sounding compressor that you’ve ever heard. Cool trick eh?
Pre-Delay Compression
Another option for the side chain is to insert an advanced version of the signal to control the level of the signal itself. Hardware compressor can never anticipate or prepare for what is going to happen, they base their reaction on what information is coming in. With this trick it allows you to prepare the compressor for what is coming.
What you do is simply make a dupe of the track you want to compress and shift it in time with respect to the other track. Connect this advanced dupe to the compressors side chain input (delay of around 50 ms should do the trick) and the delayed version to the normal input of the compressor. What happens next is that the compressor reacts more smoother and musical to the transients of the sound being processed, more realistic. You can also do this the other way round which will be very useful for percussive instruments which depend of their transients and therefore require a slow attack time so that the initial transient come through unaltered before its body gets compressed by the compressor.
Pre-Attack Compression
This trick works exactly the same as the Pre-Delay trick above but now we focus on the initial transient attack information and delay the signal only by around 10 ms resulting in additional punch to percussive instruments such as kick drum, snare, clap, and so on.
Vocal Sibilance Example
For example lets create a so called “De-Esser”. We patch in an equalizer to the compressors side-chain. The quality of the equalizer doesn’t matter as it won’t be in the audible stereo path. Usually I start by boosting the area above 5,000 cycles (5kHz). You can do some broad strokes as the energy above the sibilance band is usually not enough to unintentionally trigger the compressor. This is also another way so I don’t have to fuss with the high-frequency cutoff, and it insures me that all the sibilance’s will be attenuated. On top of that I’ll low cut everything below 5,000 cycles to increase the compressors sensibility above 5,000 cycles and decrease them below 5,000 cycles.
To get to the compressors ratio’s I usually start of with u:1 (unity to one) so I can really hear the effect and then back off the compressor to where it sounds right.
High Pass Filtered Side-Chain
High pass filtered side-chain is most commonly used to reduce the influence of low frequency instruments on the gain reduction circuit. For example kick drums, percussion, and so on. To achieve this you set the high pass filter of the equalizer up to 500 cycles. You’ll notice that your drum kit will sound more open.
Subject Of Thought: Compressor Ratios & Release Times
What ratio should I use? That’s a common question. However I’d like to forget about thinking in ratio’s, why? Cause it’s music, it’s something spontaneous. It’s creative. For example when we talk about ratio we talk about the ratio of which the compressor compresses the signal beyond it’s knee curve, e.g.: a compression ratio of 2:1 and a 10 dB increase of input level, will result in a 5 dB increase in level at the compressors output (simple math: 10 / 2 = 5). But is this the right approach? Assume that the threshold is subject to the knee curve which leads to logarithmic compression. HOWEVER beyond this point the compression lessens and the curve reverts back to a straight line, leading to no compression. Transients usually cause problems with compressors, resulting in a compressor going completely out of line and uncontrollable for a short period of time. The answer: Why not letting the transient pass through so we can focus on controlling the steady-state of our audio signal? We (peak) limit after compression anyways to control the initial transient. Add on to this musical release times in whole and dotted note values? Fortunate enough I’m already accommodating these issues with a major plugin developer.
Embrace your creativity!
The Music Surgeon
Mirror Mixing: The Official Blog 