MeldaProduction Redefines Multiband
MeldaProduction seems to be ceaselessly pushing back boundaries in computer audio manipulation – here’s yet another frontier being opened up.
by David Baer, Sept. 2017
Multiband audio processing plug-ins have been around for quite some time, most notably multiband compressors. In software, there’s a certain economy of production when developing a multiband effect. Once a developer has the code for one compressor, why not bundle several in a single plug-in and charge more?
Typical multiband effects offer a mechanism to split the frequency spectrum into two or more bands. In use, the audio for each band is sent to a processor programmed in a way different from its neighboring bands. Multiband compressors are particularly useful given that compression of bass frequencies may need to be done very differently than higher frequency content.
Of course, there are many more kinds of multiband effects than compressors. Probably no vendor has more multiband capability on offer than MeldaProduction. Their line of multiband effects (most having names ending with “MB”) include processing ranging from distortion to tremolo to convolution to anything else that makes sense as multiband. And then there’s the mighty MXXX which can put a different processing chain of FX on each band. Those who know nothing about MXXX may wish to learn more in an in-depth review I wrote when it was first released, which may be found here.
For all the types of multiband effects available, nobody has brought out a multiband filter effect, however. The reason is pretty obvious. Multiband segregates the audio content based upon frequency. A filter removes frequencies based upon how it is programmed. Mixing these two processes makes no sense, right? Well, hold that thought.
Multiband the Old-Fashioned Way
We are going to look at some fresh ways MeldaProduction has redefined what “multiband” means. But first, let’s start with conventional frequency-spectrum multiband which has been supported from the beginning.
The image just above shows the band selection window available in any MeldaProduction MB effect. In this case there are four bands. The X-axis shows frequency. When sound is present, a customizable analyzer display also is presented in this window, normally a typical real-time FFT spectrum display.
The M and S indicators on the right of each band may be clicked to enable mute or solo for the band. The little icon in the lower left corner is for bypassing the processing for that band (but not muting that signal). Per-band gain may be adjusted up to plus/minus 24 dB by moving the center horizontal line up or down (which has been done in bands 1, 2 and 4). Panning on a per-band basis can also be specified.
Positioning the bands is done in one of two ways. The white dot in the center of all but the two outer bands allows the band to be moved in its entirety. Otherwise band boundaries can be positioned individually. It’s all very straightforward, intuitive and very easy to use.
One thing is not obvious, although it should rarely, if ever, cause problems. The gain and pan controls are applied before the band signal goes into the processing. Mute and solo happen after the processing.
The background characteristics are supplied in a window invoked by right-clicking on the grid display. In doing so, we see this window:
Upon first invocation, we specify the number of bands (up to six). Thereafter we can insert a new band to the left or right of current band (the one selected in the grid). We can delete the current band. The Expand Band button temporarily solos the current band and expands the band to full range.
We can copy and paste band settings between instances of the MB plug-in. The Reset buttons should be self-explanatory. The Auto-set Limits by Analyzer function is beyond the scope of this article but details can be found in the PDF manual of any MB plug-in.
The real action happens in the column to the right labelled Crossover. In the window image above, we have the Analog option selected. There are four frequency spectrum choices:
Analog – signal segregation via good old-fashioned analog crossovers that impart no latency but exhibit phase shifts
Analog LP – the LP stands for linear phase; no phase shifts here but latency is present
Linear Phase – LP but in this case digital crossovers, not analog
Hybrid – see explanation just below
For the two analog options, we see the Slope parameter below enabled for modification. For the two analog crossover types we can select the crossover slope from the list displayed. The crossover slope dictates how gradually/abruptly the crossfade between neighboring bands happens.
The two non-analog options do not offer the Slope parameter. With LP the slope is high but frequency dependent. This is the option recommended for mastering applications. The Hybrid crossover uses slopes less abrupt than the Linear Phase option.
All the other parameters in the lower part of the Crossover panel for these four options are disabled. You may only alter controls here that are relevant to the crossover type selected above (a welcome recent enhancement that avoids a lot of confusion).
So, there you have the classics. We will spend no more time on frequency spectrum multiband. The new capabilities are what we’re really here to explore.
Panorama Segments as Bands
The first big departure from the conventional in multiband signal segregation uses the stereo soundstage as the field to be subdivided into bands. Now, we need to be clear right up front that this probably doesn’t work the way you would assume.
Let’s say your track is of an ensemble with a trumpet player just right of center. Using the layout in the grid shown just above, you might think then that the trumpet will be sent to the processing for band 3. But how is that actually going to work? The software does not know there’s a trumpet present, let alone where it’s located in the LR soundstage. Even if the software did know that, there’s no such thing as a “trumpet filter” to pull out that part of the audio. [However, if anyone knows of a Trump filter, please contact us at your earliest convenience! – Ed]
The stereo signal is passed to all band’s processing, but the amount of L and R in the signal is adjusted based upon the band configuration. Except in the case where the leftmost or rightmost bands is zero width (more in a moment on this case), some amount of both L and R will go into a band’s processing stage.
Let me try to explain how this band segregation works with a simple example. Imagine the input signal consists of a constant-level tone in the left channel and silence in the right. Imagine also that each band’s processing stage is neutral – it just passes the audio straight through.
The first thing to observe is that no matter where we place the band boundaries or how many bands are present, the output signal level remains constant. This is exactly what we would expect.
Now, assume we use the layout shown above where each band is precisely 25% of the panorama. When we solo the respective bands, we see the approximate following drops in level of the left channel signal:
- Band 1: -7.5 dB
- Band 2: -10.5 dB
- Band 3: -14 dB
- Band 4: -20 dB
If you do the math, you’ll find that those four attenuated signals will produce one attenuated by 0 dB when mixed – in other words, the original signal. The above numbers are for the left-channel signal (where the right channel is silence). Naturally, the attenuation right to left is symmetrical.
What algorithm is used to calculate the level sent to the processing of each band? That I cannot tell you, but it has nothing to do with the width of the band. You can make the first or last band zero width and it will still send significant output to that band’s processing. Furthermore, I found that in the test case (left: constant signal, right: silence) that no matter where the band boundaries were placed, every band will have higher output than the one to its right, irrespective of the location of the band boundaries.
So what if you wanted to have exclusive processing on the left and right channels? Easy – create three bands, bands 1 and 3 having zero width (the boundaries are right up against the grid edges) and mute band 2. You might need to play around with gain a little as well, but it gets you what you want. When band 1 has zero width, its processing stage gets zero amount of the R channel signal. And of course the same applies to band 3 being zero width.
Mid/Side crossover type is almost identical to the Panorama type. The L/R signal is just converted to M/S on the front end and back to L/R on the back end. The same kind of signal strength attenuation happens as is done in Panorama. If you understand Panorama (and understand what M/S is all about in the first place), there should be no mysteries here.
Mid/Side is probably the more useful of the two stereo crossovers in practice, but Panorama (Left/Right) is much easier to explain and demonstrate, which is why much more explanation was given here.
The next crossover type is Level. This one actually has a much wider range of applications than what is suggested in the documentation. Note that the grid X-axis units are dB (silence to 0 dB, left to right). The idea here is that you can create an envelope-follower modulator and make that control a decision process that decides in which band the signal currently belongs. Sounds a little difficult to set up, right?
Actually, it’s quite straightforward if you take a few steps back and look at basics.
Note the Level parameter in the configuration section. This actually sets the position within the band spectrum. The position will always be within one band and only one band at any one time. How do you control to what value this parameter is set? Answer: anything at your disposal. You could slave it to a multiparameter that is slaved to a MIDI CC. Or you could assign a modulator to it which could be an LFO or could be an envelope triggered by a sidechain signal. Or you could of course make the modulator an envelope follower and the multiband would behave as described in the documentation.
The Slope parameter governs how gradually the bands transition from one to the next.
This multiband crossover option has an enormous range of possibilities. Don’t think of it just as way to route the signal to a particular process based upon the signal strength.
The Disabled crossover type is very easy to describe. You may set up as many bands as you need separate parallel programs. In Disabled crossover mode, identical copies of the input signal are sent to the processing of all bands with no segregation of content. An example use case would be where you wished to implement parallel compression as an insert.
This is the one type where the output level will be higher than the input. If there are three bands and the processing is neutral (just signal pass-through), then the output will be three times as strong as the input. So use this option with care or at least enable the Limiter button found on all MeldaProduction plug-ins.
Parallel compression might be the most common application of Disabled multiband, but it is by no means the only creative use for this. Now, about the multiband filter idea …
Spectrum and Tonal/Transient
We’ll look at the last two crossover types together since they both first split the signal into many narrow frequency bands (thus are called the Spectral crossovers). The first, Spectrum, simply applies signal segregation based on the level of the individual narrow bands. I had a difficult time thinking of a practical use for this one. But a helpful fellow on the KVR MeldaProduction forum offered this explanation. The main purpose of this one is to separate the important tonal part of the signal from noise. The tonal part is likely to have significant amplitude and the noise component low amplitude.
Far more interesting is the Tonal/Transient crossover type. With this one we have discrimination between the transient part of the signal and the non-transient part, referred to the Tonal part. What’s perhaps different here than some other conventional transient detectors is the fact that it’s done individually for each of the spectral bands.
The parameter panel is shown to the right. The Level Slope, Smoothing, Tone and Spectral Resolution parameters are available for both types of spectral crossover types. Transient Release and Transient Resolution is enabled only for Tonal/Transient. I am not going to go into any detail about what the parameters do – that is well-covered in the documentation. However, if you have any experience with transient detector plug-ins, you should have a reasonably good feel for in what ways they may be used.
MXXX and Out
No MeldaProduction plug-in can better take advantage of extended multiband capabilities than can MXXX. While transient/tonal discrimination might make good use of different compression settings, for example, the ability to use one kind of FX processing chain for transients and an entirely different chain for non-transients is huge.
The possibilities opened up with the non-traditional multiband capabilities we’ve looked at in this article are more than a little bit intriguing. Not that MXXX needed it, but these capabilities particularly make MXXX a more desirable plug-in than it would be without the extended multiband feature. MXXX is not inexpensive, but it is looking more compelling all the time. Note: MeldaProduction has recently added a “rent to own” model subscription for its full catalog of software, which might make the acquisition of MXXX (and everything else) a possibility – details at the URL below.
For those whose budgets won’t accommodate acquisition of MXXX, however, maybe a small addition to the Melda line would be a consolation prize – so how about adding that MFilterMB to your catalog, MeldaProduction?
For more about the MeldaProduction software catalog go here: