Anyone who is familiar with the large body of work that Steven Wilson has created, either on his own or with collaborators, will be aware of his exceptional ear and attention to detail. Not only a talented musician and songwriter but an equally capable producer, mixing and mastering engineer. A contemporary music jack of all trades with an entirely clear vision of what he is trying to achieve and the knowledge and expertise to achieve it. If you have never heard of him that is entirely possible as he largely inhabits the music genre known as “Prog Rock” which, in general, receives little in the way of mainstream airplay. I for one am only a fairly recent admirer of his work. The heyday of the genre was the 70’s but it seems to be seeing a bit of a resurgence of late, not that you’d know about it from popular media.
Hand Cannot Erase is his fourth studio album as a solo artist, formerly being the creative brains behind the now disbanded Prog Rock band The Porcupine Tree. It was recorded principally at AIR studios in London during 2014 and is inspired by Joyce Carol Vincent, a resident of London who’s death went unnoticed for three years at which time they found her dead body in her London bedsit. In Steven Wilson’s own words,
The basic story, or concept of the record – it’s about a woman growing up, who goes to live in the city, very isolated, and she disappears one day and no one notices. There’s more to it than that. Now, what’s really interesting about this story is that your initial reaction when you hear a story like that is, ‘Ah, little old bag lady that no one notices, no one cares about.’ Vincent wasn’t like that. She was young, she was popular, she was attractive, she had many friends, she had family, but for whatever reason, nobody missed her for three years.
Artistically speaking, I find it a provocative and moving album and as far as the music is concerned, one of my favourites of his solo album work. That said, the quality of the recording doesn’t live up to his high standards and sends my ears into fatigue quite quickly. This is in contrast with other solo work such as 4 1/2, Grace for Drowning and The Raven that Refused to Sing, the Porcupine Tree album Deadwing or the self titled Blackfield album, all of which, in my opinion, surpass the sonic quality of this recording.
So What is Wrong with the Recording?
At the outset, lets be clear. Making quality music recordings is hard, even harder when the structure and dynamics of the music is complicated. Steven Wilson’s music certainly fits that description aptly so in a sense, I find it remarkable that he manages to produce the number of good recordings he does, given the challenges he must have faced.
The problem in this case is essentially one of dynamics and the shifting tonal balance that has resulted from those dynamics. By dynamics here I am not talking of transients in the music but rather the dynamic markings of the music itself. How it might go from piano (quiet) to forte (loud) and so on, and how the balance of the sound changes in those changes of loudness. The tonal balance shifts too much between loud and soft sections.
What is Wrong with a Strong Shift in Tonal Balance?
The fundamental problem is one of hearing fatigue brought about by frequency specific temporary threshold shift and of masking.
In the case of temporary threshold shift, our hearing perception is driven by the nerve impulses generated by the hair cells in the cochlear of the inner ear. Like all nerve responses, they are prone to fatigue if over stimulated: the stronger the stimulation the less sensitive they become. For example think of sticking your hands in a hot tub of water (washing dishes if you still do it by hand like I do) or looking at a bright light and then looking away. In the case of the hot water it feels really hot when you first stick your hands in but after a while you begin to notice it less and less. Looking at a bright light and then away and you see a “shadow” of that light in you vision, superimposed on whatever you are now looking at. Similar sort of things happen with your ears when subject to loud sound.
That may or may not be a problem depending on the spectrum of the sound causing the threshold shift. Why? Because it isn’t just one nerve ending that you are affecting but hundreds and they are tuned to different parts of the spectrum, so if you stimulate them differentially (ie. one set more than another) then the threshold shift will be differential too. To explain visually, consider this contrived example.
In this example we subject our hearing to loud sounds that are confined to the mid frequencies of the spectrum. Before exposure our ear frequency response was flat but after exposure to the loud mid range the ears suffer temporary threshold shift in the mid band the the sensitivity drops. As a consequence of the over exposure we no longer hear the detail in this band very well and it will be hours before the effect wears off. The diagram is an exaggeration of the truth and if you monitor at quiet levels you won’t suffer threshold shift but at levels that make the recording music sound realistic (nominally 75-80dBA) it is more than possible to suffer the ill effects of differential temporary threshold shift.
In the case of masking, because our hearing has finite frequency resolution and because of non-linearity in the mechanics of our hearing, loud sounds from one band of frequency can swamp neighbouring bands, making it difficult to hear the sound in those bands. Diagrammatically this can be represented with the following.
This represents the effect of masking of one tone by another of arbitrary frequency and level. The y-axis of the graph represents how much the threshold of hearing is reduced by the presence of the masking tone at any given frequency. The different plots represent the threshold shift for different levels for the masking tone (higher the graph the higher the level). Masking is most at the frequency of the pure tone and either side of the masking tone the threshold of hearing is raised in level proportionately. Because of non-linearity a pure tone masks frequencies above the tone frequency better than below. The non-linearity creates harmonics of the pure tone and those harmonics act to mask the higher frequencies. It is worth keeping this point in mind and it clearly explains the power of excessive bass to give rise to a muddy indistinct sounding mix.
Returning to “Hand Cannot Erase”
Listening to the album from start to finish, the first track that begins to fatigue my hearing is 3 Years Older. Looking at the average spectrum of the entire track there is a hint of a reason but for the most part the track looks well balanced. The view presented by Har-Bal is shown below. The green and yellow plots are the re-mastered track average and peak spectrum plots and the light and dark grey ones are the peak and average spectrum of the original recording. There is a hint of the mid-range being somewhat weak by about 3-4dB with the remainder being well balanced. Interestingly enough, my first hearing of this recording was with the previous incarnation of my monitors crossover design which included a design error that raised the mid-range sensitivity of the speakers by 3dB above flat, making this recording sound substantially better than it does on my new crossover design with a verified flat response in situ in my studio. This is essentially one of the main benefits of Har-Bal because it brings a level of objectivity to the study of track balance that is simply not repeatably possible with monitoring and listening. This is because essentially all monitoring setups will not be perfect and any anomaly in monitoring will end up biasing the track balance. If in adjusting for best sound balance, you find that the resulting spectrum looks biased it is a clear indication there is a problem in your monitoring.
But as hinted earlier, the problem with this recording is more than just a biased overall balance. If the track is split in time into its structural parts and each part analysed, a different picture emerges. To build my re-mastering filter for this track I ended up breaking it into 16 segments. Looking at the spectrums for segment 9 the track looks very well balanced with little need for EQ modification. Any modification of the track via static equalisation to fill in the 3dB apparent weakness in the mid range would unecessarily colour the track in segment 9.
Moving on to segment 14 a more stark imbalance is revealed. Here we see the bass drum and bass guitar stand 10-12dB above the mid-range content leading to a significant degree of masking of the mid-range. The distorted guitar solo sits 6-8dB above the mid-range from 2-4kHz. It should be noted that this region is the region of highest sensitivity in human hearing so by having such a high level here can result in significant fatigue from over stimulation and corresponding temporary threshold shift. It certainly plays a major part of the fatigue I experience in the original recording. Remembering that masking is most prevalent for frequencies above the loudest sound, it is likely that this guitar part significantly masks other parts around this region, most likely percussive (cymbals mainly) sounds.
This illustration of how tonal balance can shift markedly through a track is a major issue in attempting to obtain a stable and satisfying balance. Static equalisation simply can’t cut it because adjusting the EQ for best sound in section 14 compromises section 9 and so forth. The problem partly arises from instrument parts changing relative balance between different passages (ie. from a soft to a loud part). In a mixing context you’ll be coming up with a static set of fader settings on tracks to get the best balance but because the balance changes through the track a static group of fader settings just wont cut it. Commonly, in situations like this the instrument parts will be compressed to glue the track together because the act of compression will compress the change in balance in parts to a smaller more tolerable one. However, compression comes at the cost of reducing the impact of the recording through loss of dynamics. It is technically feasible to make the fader settings automated so they follow the change in balance through the track, but in practice this is a very complicated task and I’m not aware of such a strategy being used in any widespread sense. Alternatively, moving mix balance issues appear to be commonly handled through the use of multi-band compression but again, this results in a loss of impact and possibly artifacts.
Looking at the track level histogram and the average RMS level figures for the track it seems clear that Steven Wilson was aiming to preserve as much of the dynamic content as possible. As such, I think it unlikely that he used a lot of compression during mixing and similarly, if he did use multi-band compression during mastering it seems like it was modest.
Now contrast this with what is achievable with Har-Bal. In splitting the track along structural lines we change a dynamic and moving track balance issue into a static one. Each segment is treated with static EQ. No compression is required so the dynamics within each segment is entirely preserved. All that is modified is the relativity between segments. As the EQ is entirely static within a segment there is no possibility of processing artifact. Artifact is only possible at the transition from segment to segment but if the EQ is carefully designed this is entirely avoidable, even when EQ shifts substantially from segment to segment. This can easily be understood by the fact that if segments are split along lines of track structure (ie. when a part begins and ends) we naturally expect to hear a change in sound colour. Only when the segmenting is chosen incorrectly does it become difficult to achieve artifact free processing.
Har-Bal provides the means to create a uniform track balance that maximises intelligibility and minimises fatigue without the relative guesswork of relying entirely upon your ears. Simply put, it is very difficult to be 100% reliable in your decision making simply through listening, if only because your judgement in listening will be biased by any flaws in your monitoring, which can even include the state of your hearing at the time you are mixing or mastering. In my case, the quality of my hearing can and does vary substantially from day to day due to respiratory allergies and possible gastric reflux (raw onions aren’t good for my hearing). With the information that Har-Bal presents me it becomes plainly obvious when my ears are not behaving themselves and biasing my perception. Also, unlike processing through multi-band compression, processing in Har-Bal is entirely explicit. Whatever is being done to the track is presented to you exactly, not a vague feeling about what is happening.
And in referring to a uniform track balance I don’t mean destroying the balance so everything is the same. To illustrate my point, if you take a look through the spectrums for all segments in the track it should be clear that my re-mastered version is respectful of the original balance of Steven Wilson’s mix. All I have done is apply EQ to minimise masking and fatigue.
Consulting the segment EQ frequency responses it is clear that the EQ does indeed change throughout the track and quite substantially. One thing you will note from studying those frequency responses is that the responses involve a lot of high Q peaking and cutting. Though you might think that such a response would bring about obvious colouration it doesn’t. The responses are critically tuned to the content of the track using the Empathetic EQ technique that I encourage. Again, this is something that is only possible using software like Har-Bal, perhaps only Har-Bal, because it provides the means to build a response tuned to the music. This is simply not possible through listening and a conventional EQ.
The Re-Mastered Spectrums for all the Tracks
To complete the picture the re-mastered spectrums for all the tracks on Hand Cannot Erase are shown below. Again, in studying them you will find that my re-mastering is respectful of the original mix but aims to reduce masking and fatigue. In fact, just looking at the overall average spectrum you could be fooled into thinking not much has changed but in many instances that is not the case. As a result of the changes made with Har-Bal there is a substantial, easily recognisable and artifact free change in clarity. The fact that it doesn’t look like a substantial change is just indicative of how good Steven Wilson’s mixes were in obtaining a uniform average track spectrum. Most of the substantial changes are contained within the passages most sonically problematic. The best presented tracks in the original were First Regret, Perfect Life, Transience and Ascendant Here On… with the remaining tracks suffering from significant shift in tonal balance during the respective track.
Obtaining a perfect mix balance in a final master is invariably hard to a achieve when using conventional methods, even if you are blessed with a great musical ear. Hand Cannot Erase serves as a perfect example. I have chosen to illustrate the deficiencies of contemporary mixing and mastering with this recording because of the admirable abilities of Steven Wilson, not out of any twisted desire to shame. It illustrates that even if you have exceptional talent, it can still be difficult to fix problem recordings with just your ears and conventional gear.
The two most common and major issues in recorded music is loss of definition due to masking and fatigue associated with possible frequency dependent temporary threshold shift caused by excessively strong content in a narrow band of the spectrum. This is further compounded when the balance of a mix shifts markedly through the song / track structure. When a track structure is static and balance doesn’t change much throughout, the track is generally easy to mix and master. This is because the balance issue is a static one. In music with a complex structure and much use of musical dynamics the likelihood of shifting mix balance is high and if it does occur it is difficult to compensate for using conventional technique. This is very much the case in Hand Cannot Erase.
Most commonly, for these types of recordings the go to solution is firstly tracking compression to help glue the mix together and multi-band compression to even things out in the mastering stage. Both techniques result in loss of dynamics and leads to a final master with less impact than is otherwise possible.
In contrast, Har-Bal provides the means to address shifting mix balance in a track with relative ease(using segmented static equalisation that preserves the dynamics in the recording). Har-Bal’s unique approach of giving summary statistics of the track in the appropriate context (Track and segment average and peak spectrum and level histogram) makes it easy to identify and correct masking and fatigue related issues.
Through the use of Har-Bal I was able to transform an album which would begin to fatigue my perception within minutes into one that I can listen to from start to finish without fatigue and with a lot of enjoyment. You don’t need to take my word for it. You can test it yourself, but being a copyright protected recording, I cannot make the recording available to you. You will need to purchase it. Furthermore, if you have a licensed install of Har-Bal 3.7 then you can study the filters I created in detail.
Instructions to Demo the Re-master
Assuming you have access to the Hand Cannot Erase CD, rip the tracks to .wav file on Windows computers or .aif on Mac OSX computers.
Download and then install the player app on your computer (download links are provided below). In the case of Windows machines run the installer and in the case of Mac OSX machines extract the application bundle from the .tar.gz and place it where you like on your file system (desktop is a good place for it). When you run Har-Bal Player on Mac OSX for the first time, right click on the Har-Bal Player icon and select the Open menu item. You will see a message saying Har-Bal Player is from an unidentified developer and do you want to run it. Select the Open button to run the software. The next time you wish to run Har-Bal Player simply double click on the Har-Bal icon.
Download the impulse response files for the tracks (Hand Cannot Erase .hbfr files in link below).
Run Har-Bal Player, configure the output playback device and then click on the file open icon and enter the .wav (or .aif) file and corresponding .hbfr file in the dialog as illustrated below. Once opened and configured for playback, press play and listen. You can toggle the EQ in and out using the toolbar buttons.
If you don’t have this CD and don’t want to purchase it and you still doubt the ability of Har-Bal to cope with the problem of shifting mix balance in a track then you can hear and see how this can be handled in a demonstration video by following the link below. The video runs to around 1h 15mins so if you just want to hear the results scroll playback to the last 10-15 minutes or so.
Demonstration of Track Splitting
Har-Bal Player Download Links
Har-Bal-3.7-Player for 32 bit OSX
Har-Bal-3.7-Player for 64 bit OSX
Har-Bal-3.7-Player for 32 bit Windows
Har-Bal-3.7-Player for 64 bit Windows