I’ve been looking into the Bongiovi DPS as a means to restore overly compressed recordings. It kinda sorta works, but that’s not really what it was intended for. But how does it manage to re-inflate some of the damage caused by excessive compression and limiting?
Well, to find out, we build one for ourselves and then look at the effects of each component. The DPS system is really composed of 3 parts, but only 2 of them apply for headphone listening. The third part, for room speakers, is a method of widening the stereo image. We won’t address that part here.
The first two parts are (A) a method for leveling out the sound track among the “frequencies that matter to the ear”, and (B) an equalization to overcome the limitations of your headphones & speakers.
That first part, about the frequencies that matter to the ear, is a bit of a broad stretch. But in the words of a highly experienced recording engineer, there must be some merit to the statement.
DPS leveling is accomplished by flattening the spectrum, going through a broadband compressor, then un-flattening the spectrum by the opposite operation to the pre-flattening. You can do this with a bass or treble shelving filter centered at 1 kHz, and producing a 24 dB differential between the depressed bass and elevated treble frequencies. Un-flattening uses the opposite shelving filter, centered in the same place, and providing the same 24 dB of differential gain.
The compressor is a bit of a wildcard, meaning that the choice of compression ratio needs to be decided. Not wanting to do too much harm to the musical content, I decided to start with a ratio of 1.5:1. And I decided to use RMS compression rather than Peak compression.
The compressor is really just a dynamic range compression, so it should be centered around the nominal 0 dBVU level, which in my case is at -23 LUFS. Hence the threshold needs to be set low, e.g., -50 dBFS, and the makeup gain should be adjusted to keep 0 dBVU at -23 LUFS. You can tell if you have the right amount of makeup gain by alternately bypassing the compressor and placing it in service during typical 0 dBVU passages. The two volume levels should be roughly equal.
Why would you want to do this? About the only reason, in consumer audio applications, would be to help overcome noise in your environment. What this does is level out the average of the music, within limits.
The spectral flattening ahead of the compressor keeps the bass drum and bass guitar from pumping the compressor on you. The compressor sees a nearly uniform spectral content, and holds that part level. The un-flattening adds back in the bass that was temporarily removed.
If you run a gentle compression ratio like 1.5:1, then it will squeeze your music into 2/3 of its original dynamic range, centered around the 0 dBVU level. That means the soft passages will be a little louder against the surrounding noise, and the louder parts will be not quite as loud.
Where something like this makes a lot of sense is in broadcasting where there are limits to the usable dynamic range in the transmitters and receivers. And in that case, you’d pre-compress before transmission, and uncompress and un-flatten the spectrum at the receiver.
Would I ever want to mix with this contraption in line? Probably not. But if the room is noisy, it helps when listening to recorded albums.
So far, nothing here expands overly compressed recordings. And you can prove this to yourself, apart from the common sense that we are compressing again, by watching Ian Shepherd’s PSR and PLR meter readings when this is inserted or bypassed. The PSR and PLR do not expand. If anything, they shrink a bit from the unprocessed sound readings.
So where’s the magic re-inflation of over-compressed recordings? As it happens, the 2nd stage of the Bongiovi DPS is a 10-band Graphic Equalizer. How that really ought to be set up depends on your personal tastes and whatever deficiencies in your headphones / speakers you want to overcome. And when there is treble boost, as there almost always is, then this is what causes our over-compressed recordings to show some signs of life. The PSR and PLR readings expand.
So, if we don’t need any dynamic range compression, then we can completely skip part (A) above, and cut right to the chase with nothing more than a simple 10-band or 30-band Graphic Equalizer. That’s the real secret to making music sound “better” with the DPS processing. For some people a little bit of compression from part (A) might also help, if they have a noisy environment or pathetically weak amplifiers.
That’s the secret! (and there’s my bit of treble boost!)
Not much to it after all. Perhaps the hard-won knowledge of a master recording engineer would be useful as a guide to how you should be setting your Graphic Equalizer for best results. But most people already have a pretty good handle on that for themselves.
Oh well… not quite what I had in mind for recovering damaged recordings. But there you have it. I find it remarkable how so little can do as much as it does.
But I have remarked before that the ears are relatively insensitive to short term crest factors. Except for perhaps the bass region where you want the pounding of live drums and bass to help feel the effects of loud.
But to prove my point about crest factors, consider the following. We all know that a simple square wave has an odd harmonic spectrum, where each partial declines in proportion to its harmonic number. And we all know what a square wave looks like, eh?
But now look at this one…
This is also a square wave… or at least it has an identical spectrum. The difference is that each partial has a 90 deg phase shift relative to the partials used in the first plot above.
The ear can’t tell the difference between these two waveforms. And as long as your DAC and amplifier can handle the crest factor, it sounds identical to the usual square waveform.
Eh? So really, what good is our treble boost really doing, apart from just making things a little brighter in our headphones? We aren’t really expanding the over-compression after all. We’re just boosting some high frequency components in the hyper-compressed recording. We can’t get back the bass boom that was squashed away, or the sharp impact of the snare drum strikes.