Crave DSP Crave EQ 2 by Sound-Guy
Crave DSP: Crave EQ
In a recent review of the Sonible Smart EQ 2, I added a reference to Crave DSP since there is a fine explanation of minimum phase and linear phase filters on the site, and found Crave makes an EQ themselves. I also found it was getting good buzz on the Gearslutz forum ( Crave DSP releases Crave EQ v1.3 ). Testing it, I found it has some unique features as well as excellent audio quality.
Crave EQ looks like a parametric EQ!
What is It?
The Crave EQ is a 16 band EQ with four different filter implementations available. There is an analog model and a digital model (both minimum phase designs), a linear phase design and a “transparent phase” design. Crave EQ is a “straight” EQ, not a dynamic or “smart” EQ, but it is a “musically friendly” EQ modeled on the Butterworth filter design. The Butterworth design produces a very flat frequency response in the passband region and has more linear phase shift characteristics than other filter designs, which is desirable for audio work. Butterworth filter - Wikipedia
Hopefully you already know about filters and phase issues, but if not, check out Our Audio Blog - Crave DSP for some information about different designs, as well as tips on using linear phase filters. Basically, real hardware filters always introduce phase shift, the magnitude depending on the filter design, the frequency, how “steep” the filter is, and how much gain is being applied. In addition, due to the resonant circuitry of a hardware filter there is ringing after the input signal stops. Some people may think this is a “bad” thing, but in fact, it’s what we’ve heard in recordings for many decades, all the great music of the 60’s, 70’s, 80’s, and so on - and is still the design of many sought-after EQs.
When digital signal processing came along, many designs attempted to emulate minimum phase filters, but digital processing enabled something I recall from signal processing courses at university - the ability to make a filter with no phase shift. But there was a catch - such a filter requires negative time! The use of latency, delaying the output long enough to account for negative time, made such a strange concept viable!
Latency is familiar to anyone using a DAW today since many plug-ins require some delay for heavy processing. A linear phase filter also requires latency, which makes it unsuitable for some applications like live sound or tracking, but for mixing and mastering, latency is not a problem. However, a linear phase filter presents one additional issue in the negative time region, called pre-ringing. Just as the maths showed back at school, there is a “reflection” of the signal that shows up before the signal actually occurs! Back then we laughed about this as being impossible, but using latency, it is just a software engineering trick!
This effect can be heard using very short “clicks” (like single cycle or short burst of square waves) if a steep filter with high positive gain is used. The effect is subtle even in this case, and just as with phase shift and analog ringing of a real-world analog filter, it is not in itself “bad”, just slightly different. In a mix with many instrument tracks, pre-ringing is pretty much a “don’t care”, and considering that Grammy winning mix and mastering engineers often prefer linear phase EQ to minimum phase, you shouldn’t lose sleep over it! However, it’s great to have a choice of filter types.
The Crave EQ Digital model has zero latency while the Analog model has 48 samples of latency (1 msec at a 48 kHz sample rate); both of these filters are minimum phase designs. The difference between them, other than the 48 sample latency of the analog model, is the digital model produces slightly different phase characteristics at high frequencies. I was able to measure this using Room EQ Wizard, but could not detect any real audible differences.
The Linear Phase model has a selection of latency time/samples (six settings from 75 msec/3,600 samples to 1,500 msec/72,000 samples - times based on 48 kHz sample rate). The more samples you use, the more accurate the EQ shape is at low frequencies, but phase shift is always zero, and any small difference from an ideal filter decreases above a few hundred Hertz.
The unique Transparent Phase model provides analog modelling at low frequencies and shifts to a linear phase-type model at higher frequencies, with the “crossover” frequency adjustable from 80 Hz to 3.2 kHz in six steps. This enables optimizing the results based on the filter frequency settings and the signal involved. Pre-ringing can be minimized while keeping latency and phase shift relatively low. Pre-ringing with this filter model is limited to less than the latency time, and my tests found it was actually about half of the latency time, unless you count signal levels down more than 80 dB! You can actually use this filter model for tracking or live use at its higher frequency crossover settings (3.2 kHz has only a 1.25 msec latency delay, and there are 2.5 msec and 5 msec latency settings which may also work for some live/recording situations).
Transparent Model has several "crossover" frequencies
One problem with some software filter implementations is aliasing - I have some plug-ins that create aliased frequencies almost as strong as the fundamental frequency. Aliasing is bad, much worse than pre-ringing or phase shifting in my experience, because the spurious frequencies are not harmonically related to the signal being aliased. And the aliased frequencies are lower than the signal that creates them, so a 5 kHz tone may have aliased sounds as low as 100 Hz, as well as multiple tones on up to 5 kHz. This is not good, and many designs require oversampling (which can significantly increase cpu load) to reduce the aliased levels to insignificance. Crave EQ has no need for for oversampling - it creates no aliased tones that I could measure down to -180 dB FS. Very clean indeed.
Rather poor and great aliasing behavior - note the level scale is -180 dB FS at the bottom.
Like other user-friendly EQ’s, Crave EQ has some nice touches. The GUI, which while not unique (pretty much all the parametric EQ’s I have, follow basically the same format), is organized nicely, and can be continuously resized from 300x450 pixels up to your entire screen size. The frequency, filter and meter axes can also be continuously scaled, and the GUI is macOS Retina and Windows high-DPI compatible. Of course it has controls for frequency, Q and gain, as well as filter type. New bands can be created by double-clicking, and right-clicking the node of any band opens a menu to set a number of parameters. These include filter type, filter slope (which is in addition to the width-slope of the Q control), channel selection (you can process only the Mid or the Sides, or the Left or Right channel of a stereo track, or the full stereo signal), you can solo the band, silence a band, or remove it.
Stereo L/R processing mode
Stereo M/S processing mode
The spectrum analyzer is very configurable, and can show input, output and both signals. There are a number of other short-cut actions to save time setting up various parameters, and the automatic loudness compensation mode is very handy if you suddenly pop in a large peak.
There are three level meters, one on the left shows the filter gain, and there are two scales on the right side that puzzled me briefly - the one closest to the spectrum analyzer window is the dB scale for the spectrum analyzer itself, but the one on the far right has a scale reading down from zero (to as low as -180 dB - continuously adjustable), and up from zero to +30 dB. It is a peak meter, and my tests found it is a reasonably accurate true-peak meter with the zero level being 0 dB FS (it measured within +/-0.2 dB of my best true peak meters). In other words, it can measure true peaks up to 30 dB above full scale! While this is a signal level you should never be using, most DAWs today can handle peaks over full scale without clipping, but it’s best for many plug-ins to keep the signal levels below 0 dB FS, and this meter helps monitor that.
True Peak reading is shown above right loudness scale
Crave EQ has a Mono mode intended for use with mono tracks - mono uses less cpu resource than stereo, 20% to 70% less from my tests - which is nice to have. However, one DAW I used for testing, REAPER, has a situation with mono that you should be aware of. REAPER has no true mono tracks. It uses a minimum of two audio channels, so a mono file actually plays as dual-mono. REAPER users have complained about this forever since it means that with many plug-ins you must process both audio channels! Currently, the mono mode of Crave EQ (and of other plug-ins I have) sees only the left channel and sends audio only to the left channel. Keith Wood, the creator of Crave EQ, knows about this issue and plans to fix it, but in the meantime you should use a stereo mode (Stereo L/R or M/S) with mono tracks in REAPER. This is not an issue with DAWs that have true mono tracks, and not a serious problem with REAPER since Crave EQ uses so little processing power, but it did throw me for a few minutes!
How Does It Sound?
The Crave EQ sounds excellent, adds no aliased signals, adds no distortion, and provides near perfect Butterworth filter performance as well as the very clean transparent phase mode. Each filter model has its own characteristics, and while used in a full mix you might notice very little difference, used on individual instrument tracks it allows you to take advantage of each filter’s strengths. Given it’s low cpu load, one could easily use Crave EQ for all EQ needs. I know I’ll be taking advantage of it, and especially the transparent phase mode, in future projects.
I used REAPER, Studio One 3.5 and Studio one 4.1 for testing in a PC Audio Labs Rok Box PC (Windows 7 64 Bit, 4-Core Intel i7-4770K, 3.5 GHz, and 16 GB RAM). To estimate cpu load I used REAPER’s Performance Meter: a single instance of Crave EQ with just a few (1-4) bands of the analog or digital models used about 0.05%. More bands increased this, with the analog and digital models reaching a peak of about 0.09% with 16 bands. Using the transparent phase model increased cpu usage to 0.07% (one band at lowest latency) to 0.12% (16 bands at highest latency).
As expected, the linear phase mode created the highest cpu load, but still not very high at 0.10% for 1-16 bands at 75 msec latency, up to 0.25% using 16 bands at 1,500 msec latency. However, I doubt anyone really needs that setting - I found 75 to 250 samples in the linear phase mode worked fine. By comparison I have a few other linear phase EQ’s that use from 0.11% (with 10 bands) to 0.42% (with only six bands), and one very fine mastering EQ that uses 0.90% of cpu resources. Of course when mastering you need only a single EQ, but being able to use dozens of linear phase EQ’s in a mix project without taxing the system is a good thing.
Crave EQ is an excellent audio tool at a very reasonable price (US$69) with excellent sound and features, including its unique transparent mode. And it was easy to install and authorize unlike a number of programs I’ve used the past year.
Very fine sounding ”clean” EQ, with an excellent emulation of the classic Butterworth filter response.
The linear phase mode has six different latency settings available for those who ‘crave’ a perfect linear filter response (at 72,000 samples of latency!) and allows more normal humans to run with only 3,600 latency samples.
Unique transparent phase mode works well, limiting both latency and pre-ringing of a linear phase design, while enabling lower phase shift than an analog or digital design.
No aliasing and no need to use oversampling.
Low cpu load.
The creator, Keith Wood, is very responsive to inputs from users and has added features and fixed issues very quickly.
Keith is planning to add a number of features soon based on user inputs, including a frequency to key/keyboard snap toggle feature.
I found no ‘bugs’, had no crashes. Appears very stable.
Not really a “con”, but Crave EQ is a clean, static EQ, with no dynamic EQ bands, no EQ-match feature, no “tube” sound emulation, no extra “bells and whistles" better accomplished with other dedicated tools. As one forum user recently commented, “I have got the impression that for many people a comprehensive list of features is the most important criterion when purchasing an EQ plugin.” But for me it’s sound quality and ease of use, and the Crave EQ delivers at a very reasonable price.
Crave EQ, professional equalizer plugin for mixing and mastering