Sonible smart:EQ 2
When I was asked to check out the recent update to the Sonible Smart EQ, a product I thought of as “spectral matching EQ“, I realized I had a few “automated” EQ matching tools already. When I started looking at my plug-in stash I found more “match-EQ” plug-ins than I recalled having, and realized I hadn’t really been using them for spectral matching since I hadn’t found such use very profitable. However, smart:EQ 2 claims to be more than a normal match-EQ tool, and I’m always interested in new possibilities, so I gave it a go.
First, the not-so-good news: getting it authorized was a minor pain for me. Sonible uses the iLok PACE system, which I already had installed (an old version, it turns out). When I tried to authorize I got an warning message about my PACE version, so I installed the newer PACE code provided by Sonible, and tried again. It worked . . . but only after I uninstalled and reinstalled it all. Not really too time-consuming, but not what I like happening. Hopefully few people will see such an issue.
NOTE: you need the PACE module, but you do not need an iLok dongle or even an iLok account. Also, the manual suggests downloading the PACE code from iLok, even though Sonible provided a PACE installation program with the smart:EQ 2 package. Might be best to follow the manual suggestion to use the latest PACE version, but the one I downloaded with the smart:EQ 2 code worked for me.
On the other hand, the iLok connection is good news for many - you can now authorize smart:EQ 2 with an iLok dongle, but note that you need to contact Sonible to get the dongle license set up.
You can run a demo mode to evaluate smart:EQ 2 before you buy it, but I’m not sure how long it operates that way. The manual states, “you can use smart:EQ 2 a couple of days without any limitations. (Please refer to our website to find out more about the current demo period of smart:EQ 2)”.
The Really Good News
Smart:EQ 2 has both an “intelligent” self-learning EQ balancing system using its smart:filter to “create a natural sounding tonal balance” and a set of very flexible parametric filters that include bell shaped (with Q from 0.01 to 20), shelf, low and high pass, and a “tilt” filter (raises or lowers levels below the frequency setting and does the opposite above the frequency). The smart:filter generates very high resolution curves, whereas the general EQ section has only seven bands. But the correction of these seven EQ bands can be “added” to the smart:filter curve to adjust the overall EQ.
I’d not found generic “professional” EQ curves to be useful in general - I mean, how could you expect an EQ someone came up with for one rock/pop/jazz song to help on another? But Sonible have approached this a bit differently. Smart:EQ 2 is not strictly a spectral matching EQ like others I have, and does not force a fixed EQ curve on your music. It analyzes the music and evolves an EQ curve based on the spectral content of the music and on a “profile” that Sonible has developed for different styles/tracks.
For full mixes, and useful in general, is the Standard profile. It does not impress any frequency-level bias other than at the lowest couple of octaves. If you use one of the other profiles, such as piano, kick drum, distorted guitar, etc., the curve developed will favor the specified source sound. These instrument profiles are not intended for a full mix (unless that mix is solo piano or guitar, for example), but for use on tracks or buses in a mix project.
Does It Work?
In short, I was surprised at most of the results I obtained using smart:filter. One instructive test was using the Standard profile on several very well-balanced mixes, both commercial CDs and my own finished projects. In such cases smart:EQ 2 suggested almost no correction, rarely more than a dB or so from flat. In such cases the tonal difference from corrected to original was so subtle as to be almost a non-issue and strictly a matter of taste. This is, in fact, a great result!
Professional Full Mix analysed with the smart:filter
I found with a full mix that had some elements slightly muffled or harsh, smart:EQ 2 usually improved the sound notably. On really “bad” mixes, the results were sometimes a little better, but the music still needed to be fixed in the mix - if an instrument or vocal is too low in a mix, there is no way that EQ can correct the problem!
One surprise was using the smart:filter on a single instrument: I’ve been recording a number of Bach and Mozart piano compositions using some physically modeled pianos, and had a reasonable tone using several different piano models, but when I let the smart:filter try an analysis using its piano profile, I was impressed with how it “opened up” the sound. Given some time I would have made similar adjustments, but it took only a couple seconds for smart:EQ 2 to reach its conclusion.
Even more surprising was using some of the profiles on instrument buses. As you likely know, it’s not usually profitable to adjust track or bus EQ in isolation since the combination of sounds in the final mix may result in masking, or in combinations of sounds you can’t evaluate without the full mix, so I was skeptical about using the various profiles addressing individual tracks/instruments. This said, I found using smart:EQ 2 with the appropriate profiles on instrument buses (bass, drums, percussion, rhythm guitar, lead guitar, etc.) helped in many cases to make them “stand out” more clearly in the mix. In fact, one project from a few years ago, a badly recorded (not by me!) college band, which I had set aside after obtaining a reasonable basic mix (all the music director wanted at the time), ended up actually sounding good. I used 8 instances of smart:EQ 2, seven on buses and one on the master, setting up each with appropriate Sonible profiles. Took only a couple minutes. I admit it would have possibly taken me an hour or more to get to that point. This result was very impressive.
This piqued my interest, so I tried about a dozen different mixes, choosing a very early mix state for most of them, adding a smart:EQ 2 to most buses (bass, drums, guitar, etc.), then running the smart:filter analysis. On most projects I also I popped a copy of EQ2 on the master bus using the Standard profile. I then rendered the result and played my former final mix and the smart:EQ 2 update, switching between them after adjusting for loudness bias. In every case I was impressed - at the least, smart:EQ 2 provided results similar to my mix, with any difference being a matter of taste, but on a number of them, smart:EQ 2 really “opened up” the mix better than my original efforts. Embarrassing, but true!
So How Does It Do It?
I decided to analyze what smart:EQ 2 does when you use the smart:filter. It does not simply impress an EQ curve on the result. The correction it suggests, even using the Standard profile, is based on the frequency distribution of the source music, so, for example, a track with a lot of bass (like a bass guitar track) results in a different curve than say a piano or trumpet. I decided to see what smart:EQ 2 would do using noise as the source track, white noise, pink noise, brown noise, and some other colors (yes, there are at least six noise colors!). White noise is easy to generate if you have a white noise generator (!), and some noise generators include pink noise, which I use a lot for testing. White noise has equal power in any given bandwidth (say every 20 Hz or every 100 Hz), while pink noise has equal power in equal ratios of frequency, for example, 40 to 60 Hz has the same power as does 80 to 120 Hz, or 4000 to 6000 Hz.
White noise is very “hissy” (and annoying - but does approximate some percussion instruments like hi-hats) while pink noise (spectrum slopes down 3 dB per octave) has more bass and less treble, and roughly approximates the musical spectra of many full pop and rock mixes. Analyzing a range of both “classic Rock” and more recent music, the overall spectrum pretty much falls between Pink and “muddy Pink” (from -3 to about -5 dB/octave).
Below is the average spectral density made using Voxengo SPAN for a few verses from Bryan Adams “Summer of 69” one of my favorite rock songs - the slope from about 80 Hz to 10 kHz is close to -3 dB/octave, like pink noise, with a few bumps and dips at various frequencies which are characteristics of the instrument mix . Also notice the significant roll-off below 60 Hz, which indicates appropriate high-pass filtering was used to remove subsonic rumble. When I applied the smart:filter to this mix, it suggested no significant changes since it’s already an excellent Bob Clearmountain mix!
Summer of 69 average spectral plot
Using white noise and the Standard profile, smart:filter yielded a curve that is mostly pink noise - from about 100 Hz to 15 kHz it slopes down 3 dB per octave like pink noise, but below 100 Hz it also slopes down to roll off the low end, much like the Summer of 69 spectrum.
However, if you start with pink noise as the source, the result has a slope of about 3.9 dB/octave from 100 Hz to 15 kHz, rolling off below 60 Hz and leveling above 15 kHz. And if you start with a steeper spectrum roll-off, like heavy dance music (rolling off at 5-6 dB/octave), the result has a slope of about -4.5 to 5.5 dB/octave above 60 Hz and rolls off below 60 Hz.
These results are of course only for a static noise input using the Standard profile - when it analyzes actual dynamic music and instrument parts the results are more complex, and using smart:EQ 2 profiles other than the Standard one biases the resulting EQ to favor certain sounds. As I found, many of these EQ adjustments work surprisingly well.
A kick drum EQ curve developed using a kick drum buss (3 mics), kick drum profile and smart:filter
A smart:filter result for a bass guitar track using the bass guitar profile
Electric guitar profile applied to an electric guitar track
And It Gets Better!
As mentioned, smart:EQ 2 has a set of parametric EQs (called Interactive EQ) that can add to (or subtract from) the smart:filter curves, and can be used on their own without any “smart” curve if you like. They are extremely flexible, and each of these filters can be applied not only to the stereo audio of the resident track/bus, but also applied to the Mid signal, or to the Sides signal of the track. This opens up many possibilities for sound stage shaping and is a feature I’ve found on few other EQ’s.
Another feature I almost forgot to mention is the control you have over the smart:filter results - you can adjust a computed smart:response using its weighting curve control that enables you to change the gain, center frequency and range of the effects (the green line in the images above and below). Very useful as can be seen below.
Here is the electric guitar profile modified with a low shelf to cut below 200 Hz (to help the bass guitar and kick) and a 16.5 dB boost centered at 3 kHz, but applied only to the Side signal, as can be seen in the detail box at the bottom.
This last example took less than a minute to modify from the "stock" electric guitar profile and expanded the stereo delay already present in the track to a wonderful expansive sound. I could accomplish the same effect using a stereo-to-Mid/Sides convertor, a separate EQ, and another M/S to stereo convertor with appropropriate routing, but that would take far more effort and time than using smart:EQ 2.
Spectrums and Display
As you can see in the screenshots above, there is a built-in spectrum view, with pre, post and combined pre and post views. This is always handy if you need to find a particular frequency sticking out of the mix to adjust with the Interactive EQ. And what you can't see above is that the smart:EQ 2 window can be resized as you wish.
One feature that I found of limited use is the “States” section. The idea is good, but it could be more useful. In short, it enables saving up to 18 smart:filter curves so that you can apply different treatments to different sections of a song. However, it does not save any EQ adjustments you have made, which limits its usefulness, and you must separately switch these states using automation in your DAW. You could, of course, also use the DAW automation to change your EQ settings along with the State settings, but that gets a bit tedious! It does work, although it can create sudden changes in tone, which may be useful or not - that’s a trick used by many engineers to make verses, choruses and bridges stand out from each other, but needs to be applied carefully
Tested using a PC Audio Labs Rok Box PC, Windows 7 64 Bit, 4-Core Intel i7-4770K, 3.5 GHz, 16 GB RAM. A single instance of smart:EQ 2, measured in REAPER, used from 0.3% to 0.66% cpu resource depending on the type (32 bit, 64 bit, VST or VST3). So you wouldn’t likely use an instance on every track of a 100 track project, but as I noted, I found it very useful on instrument/vocal buses, and on the master bus. And if you are new at mastering, using it on a full mix could prove educational.
The software uses linear phase filters, which means there is latency (it just goes with the maths) and the delay is 4096 samples. Some people had trouble with their DAWs not accurately reporting this latency with the first release, but that appears to have been corrected. I found REAPER and two versions of Studio One (3.5 and 4.1) reported the latency precisely, both at 96 kHz sample rate and at 48 kHz. I also used a digital impulse function and square waves to observe pre-ringing, and found it behaves exactly the same as any linear phase EQ I tested at the same settings, and frankly I’ve found the concern over pre-ringing is much ado about little. Compared to a minimum phase filter, the audible changes range from totally inaudible to interesting! And when audible, it can sometimes be just what you want! See Our Audio Blog - Crave DSP for an excellent explanation of linear phase and minimum phase filters.
I look forward to more profiles, and possibly a way to create your own, since as a user you can now only modify the provided profiles using the parametric EQ curves which severely limits the complexity and resolution of your correction curve.
The smart:filter mode works well for many types of tracks, better than I would have guessed, the parametric EQ section provides some extreme (and subtle) curves, and the Mid-Sides mode is a very useful function. Of course, the smart:filter is not totally magic (!), and you may still add gain riding and other adjustments to finalize a mix, but it helps get you closer to a final mix, quickly. I’ll certainly be using smart:EQ 2 frequently in the future.
The smart:filter function works well for many styles of music and the profiles help individual tracks and buses to stand out clearly in a mix.
The interactive EQ section can be used to modify smart:filter results to adjust sounds to taste.
The interactive EQ section is flexible with good selection of filer types, and a wide selection range of gain and “Q” settings.
The Interactive EQ section filters can be applied to stereo, or the Mids or Sides of a stereo signal, which is very useful.
The weighting curve control enables adjusting the strength and focus of the smart:response curves
The States function enables saving up to 18 smart:filter curves within smart:EQ 2 itself.
The built-in spectrum view with pre, post and combined pre and post views is handy when adjusting the Interactive EQ.
Resizable window is very handy.
The States function enables saving only smart:filter curves, and not Interactive EQ curves.
User Profiles can only save smart:EQ 2 profiles modified by the Interactive EQ section, not fully original curves.
smart:EQ 2 | The Intelligent Equalizer Plug-in by sonible