Fabfilter Pro Q3
I thought I should update the earlier Pro Q3 post with more information and a few extra equalisation techniques.
PHASE MANAGEMENT MODES
Pro Q3 offers three processing modes: Zero Latency, which is essentially a minimum phase filter; Natural Phase, a proprietary algorithm; and Linear Phase, with various resolution options.
- Zero Latency - matches the magnitude response of analog EQ’ing as closely as possible, without introducing any latency.
- Natural Phase - perfectly matches the magnitude response of analog EQs, it delivers the most accurate frequency response.
- Linear Phase – Linear phase filters only change the magnitude of the audio, while leaving the phase untouched.
MINIMUM PHASE OR LINEAR PHASE?
In order to choose the best processing mode, it is a good idea to understand a little more about the theory of EQ and Pro Q3’s processing modes.
We tend to think about the term “distortion” as meaning things that get fuzzy, but that is only one type of distortion. Phase is time, so to change the natural time relationship is to distort the phase. If the phase distortion is equal at all frequencies, it is linear; if it is frequency-dependent, it is non-linear.
In practice, a wide bandwidth filter will have relatively gentle phase-distortion, as you narrow the bandwidth there is less distortion at the extremes, and more within a couple of octaves of the centre frequency. With a very narrow filter, phase distortion crowds close to the centre frequency.
The example below is a 1 KHz Bell filter with a 6 dB boost and a Quality Factor (Q) of 1.0 and 4.0 respectively.
The below example is a 100 Hz High Pass filter at 12, 18 and 24 dB/oct respectively.
In Pro Q3 you are able to change the slope of a Bell filter up to a very steep 96 dB/oct. This is a very useful feature in mixing and mastering applications.
< 12 dB/oct
< 24 dB/oct
I was always concerned that Bell filter slopes above 12 dB/oct would cause undesirable phase distortion.
Apparently, I was wrong! There is not a huge difference between the 12, 24 and 36 dB/oct slopes.
PluginDoctor measures in radians not degrees, so here is a rounded conversion:
- 12 dB/oct 0.30 radian = 17.2 degrees
- 24 dB/oct 0.45 radian = 25.8 degrees
- 36 dB/oct 0.55 radian = 31.5 degrees
Phase Differences between Zero Phase and Natural Phase
At low frequencies there is little difference in phase response of these modes. However, Zero Latency does present a small deviation above 18 kHz.
Frequency Response - 10 KHz Bell filter, 6 dB boost
Phase Response - 10 KHz Bell filter, 6 dB boost
It is useful to think of the Impulse Response (IR) as representing inertia in the filter. Like stretching a spring and letting it go. The IR represents how the filter reacts to energy; it provides a visualisation in the time domain of how long it takes that energy to dissipate. This dissipating energy, in the form of oscillations, is what we call ringing.
Minimum phase filters are non-linear, they do not have left-right symmetry; the IR starts at its highest value with progressively damped oscillations decaying in time.
Linear phase filters, have left-right symmetry; therefore, they effectively half the post-ringing and instead, provide an equal and opposite amount of pre-ringing. Pre-ringing is a backwards echo, which is most audible on transients.
Both minimum and linear phase filters suffer from post-ringing. However, post-ringing is less audible when compared to pre-ringing, as it is “masked” by the source
In practice, filter ringing, for both minimum and linear phase filters, are less audible with “gentler” EQ curves (lower Q, gain and order) and higher corner/centre frequencies.
Lets take a look at the filter ringing behaviour of Pro Q3’s three processing modes at both low and high frequencies.
100 Hz Bell filter, 6dB boost, Q 1.0 - Zero Latency
100 Hz Bell filter, 6dB boost, Q 1.0 - Natural Phase
100 Hz Bell filter, 6dB boost, Q 1.0 - Linear Phase (Medium)
Zero Latency and Natural Phase both introduce similar post-ringing with Natural Phase having greater latency. Not an issue in delay-compensated DAWs. Linear Phase has significant pre-ringing which in some situations will be quite audible. Linear Phase filters are generally not recommended for low frequency processing.
10 KHz Bell filter, 6dB boost, Q 1.0 - Minimum Phase
10 KHz Bell filter, 6dB boost, Q 1.0 - Natural Phase
10 KHz Bell filter, 6dB boost, Q 1.0 - Linear Phase (Medium)
Natural Phase’s distortion looks worse than it is (image zoom), this is only over a few milliseconds and will likely be inaudible in most situations.
Zero Phase, Natural Phase or Linear Phase?
For the majority of mixing and mastering applications Natural Phase will deliver the best results. Nevertheless, Zero Latency is very comparable in quality, and would be best suited for latency-critical applications like performance and production. However, phase issues arise with these modes when you mix a filtered signal with another similar unfiltered signal, or one that has been filtered in a different way. It is very likely that the different phase components of both signals will not align, and cancel each other to some extent.
Linear phase filters provide an answer to this problem, as they only change the magnitude of the audio, leaving the phase linear. Therefore, linear phase filters are recommended for phase-coherent applications such as: individual equalisation of the top and bottom snare mics; processing on parallel channels, like a drum parallel. It is worth noting this is not just an EQ paradigm, it is applicable to any processor that contains filters (in the audio path, not side-chain). For example, Tape emulations, saturators with filters, anything multi-band, as these implement crossover filters.
In addition, the phase distortion inherent in minimum phase filters introduces colouration. This colouration is associated with “dimension” and “depth”; this is how we hear sound in the physical world (i.e. phase is distorted as sound propagates through a medium such as air). If additional dimension and depth is undesired, a linear phase filter will provide the required transparency.
Unfortunately, linear phase filters are not “perfect” and can create some very undesirable side effects, including increased latency, and pre-ringing, which is highly undesirable in audio. However, the adverse effects of linear phase filters can be somewhat negated. Latency is not an issue with delay-compensated DAWs; the amount of pre-ringing can be made less audible with “gentler” EQ curves (lower Q, gain and order) and higher corner/centre frequencies. The choice of resolution is also important.
The latter point requires further explanation:
In Linear Phase mode, the frequency spectrum is divided into “bins” of equal width in Hz. For example, with Medium resolution at 44.1 kHz, there are 4096 spectrum points, so each “bin” is about 11 Hz wide (44100/4096). As you can see in the example below, if you require a sharp peak or dip around 50 Hz, a resolution of 11 Hz does not provide enough points to properly specify the curve. The frequency curve will be an approximate, it does not map to the specified curve depicted in the display. It will be less steep and have ripples. Another way to put it: at each point (0, 11, 22, 33, 44 Hz etc) the curve will be exactly what it should be, but it will ripple in between if the gain change between the points is quite large. You need more points to get a better approximation. In Low resolution mode, Pro Q3 uses half the resolution, so you end up with a 21.5 Hz bin size at 44.1 kHz.
50 Hz Bell filter, 6 dB boost, Q 1.0
50 Hz Bell filter, 6 dB boost, Q 4.0
At a Q of 1.0 Linear Phase Low and Medium stand up particularly well to Natural Phase. However, issues arise when the Q is increased to 4.0, especially with a Low resolution setting.
In practice, I have always found Linear Phase Medium to sound the best out of all the resolutions. However, if your DAW is struggling with delay-compensation, Low resolution is very good, especially with lower Q’s and higher centre frequencies.
Dynamic EQ changes the gain of an EQ band dynamically, depending on the level of the input signal. This makes it possible to perform subtle and surgical edits similar to a multi-band compressor, but with greater control over the band.
The dynamic behaviour of Pro-Q 3 is highly program dependent: attack, release and knee all depend on the processed audio, the frequency range of the EQ band and the current dynamic range. This results in very natural and smooth sounding compression and expansion.
The Range parameter sets the “limit” of dynamic compression/expansion for a band, ranging from -30 to 30. dB.
In the above example, I passed a 1 KHz sine wave at 0 dBFS through Pro Q3. I set up a Dynamic band centred at 1 KHz with a threshold of -3 dB and a Range of -9 dB. As expected Pro Q3 compressed the sine wave by 3 dB.
With the Range set at -9 dB, if the sine wave is boosted with an extra 3 dB, there would now be 6 dB of compression, as 6 dB of compression fits with in the 9 dB container set by the Range.
Lets take a look at the Range as a “container” for dynamics below.
Same scenario as above, a 1 KHz sine wave at 0 dBFS, a 1 KHz Dynamic band centred at 1 KHZ with a threshold of -3 dB. However, this time the Range is set to -2 dB. We are unable to compress the signal by 3 dB as this falls outside the 2 dB Range container. Therefore, we are limited to a maximum of 2 dB of compression.
Range and Soft Knee
Fabfilter in their genius have implemented a “variable” soft knee: the wider the range, the softer the knee. This feature ensures a smooth sounding dynamic response.
In the above example, we have the same 1 KHz sine wave at 0 dBFS, a Dynamic band centred at 1 KHz with a threshold of -1 dB, and a Range of -9 dB. If Pro Q3 had a hard knee we would see 1 dB of compression, not 1.5 dB.
Settings are the same as above, but with a Range of -24 dB. As you can see, we now have 4.1 dB of compression - confirming “the wider the range, the softer the knee”.
This is very useful on a practical level, lets say you are controlling an overly dynamic vocal; the threshold is set at the ideal level but the compression either sounds too “hard” or you are not getting “quite enough” compression. Increasing the Range will provide more compression, as the onset of compression is at a lower level due to the softer knee. This slower onset of compression is “gentler”, as the ratio is gradually scaled up to the threshold setting. Conversely, lowering the threshold would provide more compression but this will be “harder” compression and has the potential to noticeably react to other parts of the vocal.
Images below are examples of knee settings:
< 3 dB Knee Setting
< 24 dB Knee Setting
PRO Q3 OR PRO MB?
Like Pro MB, Pro Q3 can dynamically control a range of frequencies. However, Pro Q3 has some advantages and disadvantages when compared to Pro MB.
Pro MB is a fully featured multiband compressor/expander with controls for time constants, ratio, knee, and lookahead, as well as the ability to side-chain, upward compress and downward expand. Therefore, Pro MB is ideal for multiband compression/expansion duties, where control over dynamics are most important but over a broader range of frequencies. For example, if you want to compress the upper mid range of a stereo master. In Pro MB you can modify the shape of the compression with the attack and release. Set the amount of compression with the ratio, adjust the knee behaviour. Hard knee immediately compresses at the set ratio when the signal passes the threshold. Soft knee is when the onset of compression is gradual; starting say 6 dB below the threshold at a low ratio and increasing to the set ratio when the signal reaches the threshold
In addition, you cannot use Pro Q3 as a downward expander to attenuate spill/noise or as an upward compressor to bring up the sustain of an 808. You are unable to modify the degree of stereo unlinking of bands. Although Pro Q3 now allows for external side-chain compression, you are unable to set a frequency range for this trigger, like you can in Pro MB. For example, if you wanted to side-chain compress your sub bass triggered from an external kick, but you wanted to define a band around the ‘click’ of the kick.
Pro Q3 is a parametric EQ with the ability to downward compress or upward expand bell, shelf, and tilt filters. Pro Q3 is more surgical when it comes to frequency bandwidth and contour, and really shines when you want to downward compress or upward expand a very specific band of frequencies. For example, a bell filter to emphasise the click of a kick drum, or compress resonances with a narrow Q.
These specific applications are impossible in Pro MB, as you are working within the limits of crossovers and the narrowest band is around half an octave. Crossover slopes are limited to 48dB/oct in Pro MB, Pro Q3 slopes go up to 96db/oct.
In addition, Pro Q3 also has a wide range of shelf shapes ranging from analogue, Baxandall, resonant and anything in-between.
Pro Q3 is also great for applications that require both equalisation and dynamic control. For example, your vocal is a little “dull”, so you boost the high frequencies with a high shelf, the timbre is perfect but there is now some sibilance. Instead of inserting another plug-in in your chain in the form of a de-esser, you can easily add a dynamic band within the shelf to control the sibilance, just set the threshold to compress only when the sibilance is present.
Static EQ - Attenuation
Static EQ bands are desired when you want to boost or attenuate a range of frequencies throughout an entire track or mix.
- Room Tones - you want to attenuate as much of these frequencies as possible without effecting the instruments timbre. A dynamic EQ would not make sense in this situation because room tones are static throughout the recording.
- Extraneous Frequencies – if there are frequencies within the track or mix that are not desired. For example, the kick drum is pumping nicely but a little too “boomy”, attenuating this frequency range with a static EQ will eliminate the boomyness whilst retaining the punch. Additionally, if there are excessive sub or lower mid-range build-up throughout the entire track but the dynamic range is balanced, a static EQ dip will be the most appropriate course of action. Conversely, utilising a dynamic EQ within these situations would disrupt the dynamic balance.
Static EQ – Boosting
- Enhancing Frequencies – if you want to enhance frequency ranges throughout an entire track or mix without altering dynamics or groove, a static EQ is best suited,
Dynamic EQ – Downward Compression
- Overly-Dynamic Frequencies – these bands are great for attenuating unwanted frequencies that are not constant throughout a track or a mix. For example, a vocal that gets shrill during certain parts of a chorus, a dynamic band can be set to only compress these frequencies when they become excessive. Additionally, when mastering a stereo mix, a dynamic EQ can control a spikey bass or lead note. A static EQ would attenuate this frequency range regardless if the bass or lead note were playing, thus adversely effecting other elements of the mix.
Dynamic EQ – Upward Expansion
- Adding Dynamic Energy - upward expansion is great when you want to add dynamic energy to the source. For example, adding “click” and “punch” to a kick drum, boosting a static EQ would also brighten the tail of the kick, which sounds unnatural, a dynamic EQ will boost at the transient and then dissipate quickly, just like a drummer striking the drum with more velocity.
Dynamic EQ – Downward Compression with Upward or Downward Gain
- Downward Comp with Upward Gain – if you want to boost the low frequencies of your bass but an occasional note is too loud. Lets say a 3 dB gain boost works perfectly apart from that note, which is an additional 3 dB louder. Here, you would set the threshold to only be triggered by the “rogue” note and set the range so that the note is compressed by 3 dB, thus levelling the bass out.
Downward Comp with Downward Gain – lets say you have a synth that has excessive midrange and is a little too dynamic. Compressing the midrange by 6 dB sucks the life out of the synth, as it is over-compressed; however, it does sound balanced frequency wise. Compressing the midrange by 3 dB provides good dynamic control but does not sound balanced frequency wise. And 6 dB of static gain reduction does not control the dynamics. This is a great situation for combining a dynamic band with static gain reduction. For example, you could compress the midrange by 3 dB to control the dynamics with an additional 3 dB of static gain reduction to balance the frequencies.
Another example, a stereo-master has too much upper midrange throughout the entire track but in particular during the choruses. Instead of automating a static EQ to attenuate even more during the choruses, you could set a dynamic band with static gain reduction throughout the song but with additional compression during the choruses.
Dynamic EQ – Upward Expansion with Upward or Downward Gain
Upward Exp with Downward Gain – if you have a kick drum that is too “boomy” and lacking “punch”. 3 dB of static gain reduction rectifies the boomyness but not the lack of punch. Of course, downward compression would reduce the dynamics further. Therefore, combining static gain reduction with upward expansion is the most appropriate strategy for this application. In this example, I set a dynamic band with 6 dB of static gain reduction (3 dB more than desired) but with 3 dB of upward expansion, this provides the 3 dB of gain reduction required to control the boomyness, but with 3 dB of expansion to add more punch.
Another example, a stereo master is too bright but a static EQ or dynamic band in downward compression mode loses high frequency energy and sounds dull. You could therefore, use static gain reduction to balance the high frequencies and then add some dynamic life back with upward expansion.
- Upward Exp with Upward Gain – lets say you have a synth that is lacking in the midrange both in frequency and dynamics. Expanding the midrange by 6 dB sounds too dynamic; however, it does sound balanced frequency wise. Expanding the midrange by 3 dB provides good dynamic energy but does not sound balanced frequency wise. This is a great situation for combining a dynamic band with static gain. For example, you could balance the midrange frequencies with 3dB of static gain boost with an additional 3 dB of upward expansion to emphasise the dynamics.
Static and Dynamic Combinations
I believe one of Pro Q3’s strongest features is the ability to combine both static and dynamic bands.
For example, a high shelf boost with a narrow bell dynamic band. This would be great for the vocal example mentioned earlier, where the high shelf boost produces the desired tone but introduces sibilance. The dynamic band could be set to compress whenever there was sibilance. This is great as it reduces the need for a de-esser later in the chain.
You can also combine dynamic bands together. For example, a wide midrange bell, adding some upward expansion, coupled with a narrow bell, downward compressing a spikey synth note. Or a low frequency shelf, subtly upward expanding the kick and bass, coupled with a bell focused on the kick with more aggressive upward expansion.