This plot shows a unity‑gain reference trace (red, in the middle), along with the transfer functions for conventional parallel compression (brown, top), and polarity‑inverted parallel compression (green, bottom).
I’ve played with ‘out of phase’ parallel compression (where the compressed signal path is polarity inverted) a few times in the past. It was recommended to me by a mixing teacher 16 years ago, and it sounded good — but, honestly, I didn’t really understand what I was doing! More recently, I heard it being discussed in Dan Worrall’s ‘Parallel Filters, God Tier’ video, in which he said this is actually expansion. It seems to be a convenient way of doing expansion (I usually don’t find using an expander very convenient), but I have a few nerdy questions.
First, parallel compression is known to be less damaging to transients and this is what I love about it. I wonder, would expansion using out‑of‑phase compression be similarly gentle towards transients?
Second, what advice do you have for using such techniques? Would you use similar settings as for parallel compression (eg. high ratios) or gentler ones for expansion? I’d love Hugh Robjohns to write a Part 2 to his 2013 parallel compression article [www.soundonsound.com/techniques/parallel-compression], discussing details of such techniques with similar transfer plots!
SOS Forum post
Hugh Robjohns, SOS Technical Editor replies: In parallel compression, the loudest elements (the transients) are carried mostly in the direct path and are thus largely unaffected. In a conventional expander, signals above the threshold (loud transients) are unaffected anyway; as far as transients are concerned, there’s no net benefit in using a parallel expander setup compared with a standard expander.
In a parallel compression setup, because low‑level signals are heavily compressed and added to the (low‑level) direct path, there’s an increased signal level at low levels, hence the upwards compression effect.If you invert the polarity of either path (usually the compressed one), low‑level signals are subtracted from the direct path, resulting in a form of downwards expansion. But while quiet signals get even quieter, the behaviour is not the same as it is with conventional downward expanansion.
You requested a transfer plot, so I’ve obliged! I sent an Audio Precision (AP) test set signal into my DAW (SADiE) on two separate channels. One channel was routed at unity gain direct to the output, which was fed back to the AP test set. I ran a transfer plot which is the central (red) line, completely linear from ‑55dBFS to 0dBFS, as you’d expect. The second channel had a stock compressor inserted, with the threshold set at ‑42dBFS, a ratio of 50:1, and no make‑up gain. With that second channel feeding the output as well, at unity gain I measured the upper (brown) transfer curve, showing a conventional 6dB uplift below the threshold, reducing gently above the threshold. In other words, the classic upwards‑compression curve you’d expect from parallel compression.
There is an obvious difference between a conventional downward expander and this parallel configuration...
The lower (green) trace is the result of flipping the polarity of the compression channel. This shows a gentle and progressive downwards expansion curve, in which the attenuation builds towards the threshold at ‑42dBFS. Below this threshold, pretty much total cancellation occurs, because the opposite‑polarity compressed and direct channels have identical signal levels. There’s an obvious difference between a conventional downward expander and this parallel configuration: a standard downward expander maintains consistent gain reduction for signals falling below the threshold, but does nothing for those above; whereas this parallel inverted‑compressor arrangement results in a progressively increasing attenuation all the way down, and total gating for signals below the threshold. Quite a different effect!