DSEQ 3 Dynamic Equalizer from TBProAudio
DSEQ 3 is the third version of TBProAudio’s very comprehensive dynamic spectral equalizer. I have several dynamic EQs, but most have a limited number of filters and operate in the time domain which limits their frequency selectivity due to filter limitations. DSEQ 3 operates in the frequency domain which enables very tight filters when needed as well as having very many of them (over a thousand!). I was actually running comparisons of vocal de-essers when DSEQ 3 caught my attention.


DSEQ 3 with Input Signal spectral plot (grey), Gain Reduction plot (blue) with the Spectrum Slope panel in upper left of
the plot area – overall pre-filter curve is the yellow line and settings for each pre-filter are shown in panel below the spectral plots.

A Dynamic Spectral Equalizer?
I have several dedicated de-essers, one twenty years old that I’ve found hard to beat in use. Simple, few controls, and it takes out the sibilance very effectively without changing the tonality of the signal. However, I had a vocal to clean up that wasn’t just plagued by “hissy” sibilance, but had considerable “whistle” tones and some odd microphone noises. It was beyond fixing with any normal de-esser that provides a band-limited frequency dip in the sibilant frequency range since the whistling tones and other odd mic noise came through even when the sibilance was cut.

Since I had been very impressed with TBProAudio’s ABLM 2 loudness matching tool (which I use every day, both in mixing and in testing gear), I decided to see how well their dynamic spectral EQ might fare with this vocal. The short answer is, extremely well. I could quickly adjust the gain reduction applied to the hissy sibilance and the whistling tones were pretty much automatically dipped to extinction just by adjusting the threshold.

And trying more tracks and mixes with various issues really dazzled me! DSEQ 3 (DSEQ from now on!) is excellent for de-essing vocals, taming resonances, smoothing harsh frequency build-up, reducing ringing, balancing mixes, even clearing up masking issues like the classic kick versus bass guitar, very transparently. This is because it doesn’t just duck the full signal or even a small frequency range, but can address extremely narrow frequency bands as needed. Or it can easily be adjusted to dip a broad range of frequencies. Oddly, DSEQ 3 (or 2 or 1) has never been reviewed in Gearspace. Version 3 actually has been out for over six months as I write this and has had a number of updates and improvements, so I found it very capable and perfectly stable in my tests.

The Main Event
Like most plug-ins today, DSEQ has a setup menu, a preset system, A/B comparison and a previous/next preset control along the very top of the GUI. The toolbar below the top edge includes fifteen controls that may seem daunting, but basically are some plug-in settings including over-sampling, quality modes and some processing features that I’ll describe later. On the far right is the now familiar wet/dry control for parallel processing if you wish.

But the controls you will be using the most are in a strip down the left side of the GUI. These are Slope, Selectivity, Channel Link, Attack, Release, and Threshold. The last three should be familiar since most compressors and dynamic EQs have them, but the first three are less common. Slope adjusts the spectral balance of the input signal that is sent to DSEQ’s gain control. It is basically a tilt-EQ, centered around 1kHz, that boosts highs and drops lows for positive values and boosts lows and drops highs for negative values, but this boost and cut are applied only to the signal sent to the gain control, not the main audio signal. This may seem a bit odd if you don’t realize that Slope controls the gain-reduction – boosting highs will tend to create more gain reduction in highs rather than boosting levels there while boosting the low frequencies will cause greater gain reduction in the low end (depending on threshold and other settings of course). So when you first play with Slope and it seems to be doing the opposite of what you expected, this is the reason. If you are familiar with spectral plots you know that pink noise decays at 3 dB per octave, and music generally has an overall spectrum falling off from about 3 dB per octave to as much as 5 dB per octave. The Slope control, for one thing, enables you to optimize how DSEQ handles harshness is the music you are treating.



You can view the input spectral display as it’s affected by Slope and to obtain equal treatment across the full audio spectrum, adjust slope until the input spectrum is approximately flat, or bias the control as you wish. You can turn on a Show Spectrum Slope mode in the Analyzer popup menu (see below) to get an estimate of the pre-Slope control input spectrum slope. Handy to see how the music is distributed, frequency-wise. And the Slope control has two automated modes; first a semi-automated method using the Calculate Slope button – this sets an optimal slope value based on main input signal spectrum. Second, there is a fully automatic Adaptive Slope button that enables DSEQ to continuously calculate the optimal slope (the same value shown in the Spectrum Slope window mentioned above). Note that changing the setting in the manual slope parameter window will increase or decrease the calculated adaptive slope additively, so set it to zero unless you really want to modify the slope determined by the Calculate Slope mode or Adaptive Slope function.

Selectivity determines how much the action of a singular dynamic equalizer affects its neighbours. Higher values limit the effect to a more singular dynamic equalizer, but may create some distortion. Note that there can be over a thousand dynamic equalizers operating at once and the very highest selectivity setting will have them essentially operating independently which can create odd artefacts. I found between about 50 and 85 was often best, but up to “Sharp” (100) was useful at times depending on the needs of the processing (i.e., de-essing works fine with low values while removing sharp resonances is more effective with higher values). Very low values (under 20 or so) can lead to pumping, in case that’s what you want!

Channel Link controls how strongly left/mid and right/side channel processing is linked. Low values provide more separate processing. For a stereo master a value of 75% is useful. A setting of 0% provides dual-mono processing.

Threshold has a couple extra features not found in most compressors and dynamic EQs – in addition to manually setting the thresold there is a control to automatically set the optimal threshold value based on the main input signal. As if that weren’t enough, there is also an Adaptive Threshold mode to make DSEQ continuously calculate and update the optimal threshold level. And there is also a Gain Reduction strength control that boosts or lowers the gain reduction relative to the normal level, from 0.01x to 10x with the default factor at 1x. This works in a similar manner to the ratio parameter of a compressor, and is very effective using high gain in surgical cases such as removing resonances. One additional control for threshold is a Max GR that limits the maximum gain reduction to a selected value in case you want limited reduction levels.

Filters Anyone?
You may wonder why an EQ (which is a kind of filter itself) has its own filters? There are a dozen parametric filters controlled in a panel below the large spectral plot area – these are pre-filters that act on the signal sent to the dynamic EQ gain control (not on the input or output signal of DSEQ) and can be very effective in helping focus the dynamic EQ action on frequencies of concern. There are 16 filter types plus a gain control – there are slope and Q control as appropriate, and as with all good filter designs now, you can use the mouse to drag gain and frequency, as well as the mouse wheel to change Q or slope. And there are more “tweaks” available including a Channel setting that controls which part of the signal is affected by the pre-filter (all/left/right/mid/side). The user manual describes all this in good detail.

And there is More!
There are more features to describe. The central portion of the GUI is a spectral display, but it is not just a normal spectral display. In addition to showing the usual spectral density of the input signal (post-Slope control as mentioned earlier), it can display a gain-reduction view which is most handy – I found I used this the most. It can also show a display of the pre-filter response, turn on the Show Spectrum Slope panel described earlier, even turn on a spectrogram display (level versus frequency plotted on a timeline).

And below the pre-filter area there is yet more. On the left is a Custom Threshold control panel and towards the right, the Smart AI panel. DSEQ can work with either fixed threshold values or, using the Custom Threshold, work using custom curves calculated from any audio file or from the recorded input signal itself. This provides a high range of flexibility, though I found a fixed threshold modified using the pre-filters to emphasize or de-emphasize frequency ranges usually covered what I needed.

DSEQ also makes it easy to tame digital harshness using its smart AI function. Presets include a number of AI assisted functions, and selecting a desired AI learn mode, you start playing the file you wish to process, press the learn button and DSEQ calculates the optimal slope and threshold parameters in seconds.

And we have not yet taken a look at the top toolbar mentioned earlier. This has controls for processing functions (oversampling, quality and processing type: L/R or M/S), display functions, and several very useful operational features.


At the left is oversampling control and you can see there are two settings (1x/1:1) – this is because you can select different (or the same) oversampling for online and offline processing (offline rendering includes bounce, mix-down, and audio export modes). The Quality setting also can be separately set for online and offline processing with 7 different quality modes: eco-eco, eco, normal, high and ultra, ultra 2 and ultra 3. I found normal and high to be fine from most of my tests, but the lower the quality level, the higher the aliasing that can occur. The Mode control selects left/right or mid/side processing while the Analyzer window selects what is displayed on the spectrum view (All = the summed left/right spectrum, left, right, mid or side).

Finally there are seven ‘buttons’ that activate different features: the LP/NP switch (here showing LP) selects between the linear or natural phase dynamic filter mode, the next button is a Bypass switch which bypasses all DSEQ 3 processing, and the button marked with brackets [ ] selects Pre-filter monitoring to listen to the incoming control signal through the pre-filter settings. The Δ button is a delta switch that lets you hear the difference signal: input minus output. This enables hearing the impact of pre-filter and dynamic equalizers more clearly. The button with little headphones turns on the filter band listen mode to hear the output signal in a specific pre-filter band with all DSEQ processing active.

The last two buttons are the external side-chain toggle and a spectrum analyser freeze control if you want to study the display with it not bouncing all about. Well, not really the last – to the right side of the snowflake freeze button is an area with the letters AB-LM Lite which controls a light version of the ABLM loudness matching tool and works very well to eliminate loudness bias when using the Bypass control. Very nice to have this right in DSEQ.

Operation
Presets can save almost all settings including over-sampling level, quality mode, processing mode, and even the analyser mode, as well as settings for slope, selectivity, channel link, attack, release, and threshold. And if any pre-filters have been set up, they will also be saved in a preset.


DSEQ 3 Preset manager window

While there is a lot in DSEQ (and I have not covered everything in this review!), I found it very easy to use in practice. While I’ve described many features and functions, starting with a preset such as “Drums – all”, “EGuitar – soften highs”, “General – remove ringing”, or “Master – deess” will set up all the necessary parameters to get started, and then only a little adjustment of Threshold, maybe Attack and Release, and pulling the peaks of the pre-filters in an appropriate frequency range up for more gain reduction, or down for less, can result is incredible control.


DSEQ 3 “zapping” a resonant frequency near 1 kHz in a fiddle track – no need to find the frequency manually.

Sharp resonances are often tamed, even eliminated, just using the init(ial) preset (everything flat) and adjusting the Threshold and gain strength control as needed. Using the input spectral view enables seeing what is causing a problem, and dragging a pre-filter to the problem area and adjusting its gain and Q can often tame the problem quickly. While it is possible to create some rather odd artefacts with extreme settings, I found I could obtain excellent, transparent results using a fairly wide range of settings.

DSEQ 3 is definitely an extremely versatile dynamic EQ that can help solve many kinds of problems on tracks, buses and full mixes.

Technical
Requires Windows 7, or later, OpenGL 2 GFX card/Mac OS X 10.11 or later, Metal GFX card. Plugins provided for Win: 32/64 Bit VST, 32/64 Bit VST3, 32/64 Bit AAX or OS X: 64 Bit VST, 64 Bit VST3, 64 Bit AU, 64 Bit AAX .

I tested DSEQ 3 using a PC Audio Labs Rok Box with Intel Core i7-4770K CPU @ 3.5 GHz, 16 MB RAM running Windows 7 64 bit. CPU usage varies with oversampling: from a low 0.5% with no oversampling and High Quality setting, up to about 2% at 4X with Ultra 3 quality. Latency varies with the Quality setting from 384 samples with eco-eco to 24,576 at Ultra 3 at any sample rate. At the very usable High Quality level and no oversampling, the latency is 3,072 samples. So not a processor for real-time use in tracking.

Pro:
An amazingly flexible dynamic EQ with some extremely useful functions at a very fair price
Over a thousand dynamic filters at once if you need them
Pre-filter monitoring, Delta mode and Filter Band listening mode are especially useful for fine tuning correction
Selectable linear or natural phase dynamic filters
ABLM Lite eliminates loudness bias when using bypass
Several automated parameter modes to help streamline use
Full DAW automation support
Wide range of processing quality levels
Useful EQ cheat sheets (OK, I didn’t describe this among other extra features!)
Can use main input signal or external side-chain signal for control (I barely mentioned that)
Demo version available
Free upgrade for owners of previous DSEQ version
Did I say, a very fair price for such a powerful tool! Excellent value for money

Cons:
Nothing that bothers me. If you are an audio engineering neophyte you might find DSEQ 3 a bit of a challenge, but a few hours playing with it will likely make you an admirer.

https://www.tbproaudio.de/products/dseq