Not too long ago I posted about a new plug-in called Dynamic Spectrum Mapper from the brand new company Pro Audio DSP featuring ex-employees of Sony Oxford. I got in touch with one of them, Paul Frindle, to ask him some questions about the new company and the new plug-in.
Tell us a little about your background.
I came from a musical family and since I was a child I have always been fascinated by music, technology and reproduced sound and was happily messing with Hifi and guitar amps from my early teens (much to my parent’s horror). By my late teens I was playing in bands and spending all my spare time on valve amp designs, trying to get hair-raising and spine tingling sounds from my home made guitar rig.
My first brush in professional audio came around 1970 when I worked as designer for a small company, which amongst other things was constructing very early 8 track recorders for celebrity musician clients, using modified versions of commercial real-to-real machines. Following this I had a period of working in various consumer repair shops and playing in bands by night and my first job in a real studio work came in the mid 1970s when I eventually landed a job in a studio in Paris. I was thrown in the deep end there and did everything from maintenance, recording, mixing, studio design and in my spare moments fiddled with new designs for compressors, noise reduction and such like. It was a fantastic experience.
After a few years I came back to England and spent some eventful months working in London at Trident studios, but having failed to find any reliable accommodation after 6 months I eventually ended up working in Oxfordshire at the Virgin Manor studios as a maintenance and mobile recording engineer. It was there that I was later introduced to Colin Sanders and his nascent SSL Company and eventually became a designer for them, working on the E-Series console, early assignable console designs and the G-series follow up console. Later I worked on their early digital audio initiative, where I was landed with the daunting task of making early converters ‘sound good’.
Realising analogue consoles had reached their practical limit in complexity, and large scale digital systems would be needed to satisfy market needs, five of us left SSL to set up Oxford Digital Ltd, where under contract to Sony we embarked on designing the OXF-R3 console system. Later, in around 1994 Sony took over the Oxford Digital outfit in order to put much needed resources into the burgeoning project so, along with other key members of the design team, I became a Sony employee.
Can you please tell us about your breakup from Sony Oxford (Sonnox)?
Well the break with Sony occurred when they offered early retirement to senior engineers over the age of 50 years old. At that time I was suffering stress related health symptoms and wasting more and more time away with associated illnesses, so I thought it best that I accepted the offer. Several other key staff also accepted early retirement and shortly after Sony entered into a protracted period of selling the department on to third parties. Two management buy-out teams eventually bought out their respective parts of the outfit, one of which comprised the highly successful plugins project, which went on to form Sonnox.
It was during this time, when myself and another ex-employee were pretty much idle, that we hatched the notion of making our own plugins and designing some of the many product ideas we still had not been able to realise. By the time the Sony buy out had finally completed I had been away from Sony more than a year and we were both pretty much buried into what we were already doing.
New company, new plug-in. Dynamic Spectrum Mapper is really something else. How on earth did you come up with that idea?
This product idea was initially born from a nagging dissatisfaction with existing multi-band compression techniques. The processes that use conventional band filtering have phase issues between bands that cause the whole thing to change timbre constantly in a weird and unnatural fashion. Other more recent designs that use phase corrected filters have impulse response issues that soften instrument percussive events and ruin definition and impact by turning ‘bumps’ into ‘whumps’ and ‘bangs’ into ‘wangs’ – none of which sounds too great to me! I had sounds in my head that might be achieved if different processing methods with fundamentally less annoying artefacts could allow much more flexible and creative control, without giving me a headache.
Added to this was the long standing and annoying problem that conventional compression momentarily brings down the whole track often with single events in one part of the frequency spectrum. Multi-band compression may avoid this problem to some extent, but only by imposing some simplistic crossover of frequencies you provide manually – which do not in the least match the nuances of the program. The missing dimension from multi-band compression is a true underlying representation of the actual program spectrum, presented with enough bands to make a believable attempt at matching the original sound. It struck me that if this could be achieved, quite apart from making a damn good job of compression, all sorts of other exciting stuff could be done using this level of complexity and accuracy in the dynamic control of the frequency spectrum.
How did you realise this in the DSM?
Well, for any of the above to make sense at all we had to fix the bad sounding filtering problems I mentioned first off. Quite perversely (for the purist), the first slightest hint I had of this ‘sound in my head’ came from using MP3 encoders in the 1990s. . These things split the program into bands using FFT methods and dynamically throw away data in the frequencies each side of the most prominent frequencies, on the assumption that we can’t hear what’s lost, because of ‘masking effects’. But as we all know you can hear it just fine – and being no stranger to the various sounds of data truncation after so many years of messing with converters and digital stuff, I could extrapolate in my head and imagine what it might sound like if the levels of those bands were being changed, rather than just being truncated to less bits. I realised back then that these methods were something that could potentially produce the required result. From then on I wanted to try this idea as part of a compression process – it had been logged in my head along with the heaps of other stuff built up over the decades that I’d like to try – if I ever got the chance.
I have been fortunate enough to work with people who have the required strength of knowledge, fluidity in maths, innovation and imagination to make this idea a reality. Of course early trials sounded pretty awful, but I could hear through the nastiness to perceive the nuances of what I wanted buried within the unwanted artefacts. After than it was a long drawn out process of innovation, improvement and endless listening trials until we got something that was spot on. At that point it was worth adding the necessary and interesting functions and endlessly tweaking the thing into what we have today. The process drove us all to distraction on several occasions, but this is always the case with design at this level. If you havenâ€™t been driven close to the point of breakdown, the chances are you havenâ€™t tried hard enough and there is more still to do!
What would you say the typical application for DSM would be?
There are loads of potential applications for this process, with examples included in the set-up list. They range from achieving conventional tasks dramatically much more quickly than with normal devices, and to a much higher subjective quality. These include things such as programme compression, limiting and maximising, advanced vocal processing, instrument timbre modification and such like. However beyond this it can also achieve things that are virtually impossible to do well with other single applications, such as compression of extremely difficult material, character mapping between whole mixes, tracks, vocal parts and instruments, continuity matching of diverse programme types, play out character for broadcast, correcting varying microphone proximity effects in video, film production and difficult outside broadcast situations. And of course there are a whole range of new creative possibilities in making novel sounds and dynamic modifications to programme of all types, which cannot be achieved easily by other methods. The list could go on and on. You can mess about with it to create all sorts of things. I find this all very exciting.
One of the problems with new applications like this is getting the message over, in a world where conventional categories of device are most immediately recognised. The DSM is a multi-talented device and a far cry from a single function device. Yes, it can tackle these conventional tasks extremely well, but it would be misleading to simply call this a limiter or a dynamic EQ or any other singular function. It is absolutely not just an EQ, compressor and spectrum analyser shoved mindlessly onto the same GUI!
Of course we could have used this technology to make a suite of single application plugins by selective restriction, and sold them all separately as more recognisable functions. But this would have been bad value for the users and would have sadly limited what can be achieved by having the whole function presented in its completeness. The combination of the whole thing is what makes it so powerful.
Do you have any tips & tricks for the DSM to share?
The major thing thatâ€™s worth getting across is just how easy and fast it is to use. Operationally you can look at it a bit like a compression device with the added dimension of an underlying frequency spectrum, which is acquired simply by hitting the â€˜captureâ€™ function.
Once you have got your preferred timing, ratio and threshold settings in place for the kind of work you are doing, most of what you want to do can be achieved by simply hitting the â€˜captureâ€™ button on the part of the programme that sounds best.
For instance, programme maximisation can be achieved by hitting the capture during the very loudest passages and leaving the DSM to drag everything else up to that level with a similar frequency spectrum content. Because the underlying spectral content is from the real track at itâ€™s loudest, the impact of that loudness is maintained. You can then hit the limiter function and increase the gain or threshold to get great volume enhancement and impact.
Vocal processing can be achieved by hitting the capture button during a part where the voice is sounding great (e.g. not â€˜essingâ€™ or â€˜screechingâ€™). The DSM can then be just left to apply this to the whole track. By increasing the threshold you can ensure that only the loud bits get processed and the rest is unchanged. And since the underlying spectrum it is using to process is actually from the real voice itself, the results are more natural and believable. Itâ€™s uncanny to experience.
You can map the characteristic sounds of one track onto another, just by capturing one and playing the other through it. So for instance you can quickly achieve continuity between whole tracks in mastering situations, or you can treat your vocals to the overall sound of another, even captured from a completely different take. You can even capture the spectrum of one kind of instrument and apply it to another. And if you just want the frequency character without the dynamic compression, you can just hit the â€˜freeze gainsâ€™ button and it will act like a complex EQ.
And of course you can manually tweak the response of the compression spectrum using the EQ style controls and selective timing sliders to make creative changes to the sound. For instance, preferentially increasing the HF recovery speed using the HF REL can produce some great HF sheen to the sound â€“ great for things like stringed instruments, orchestral parts, acoustic guitar and vocals. Slowing the LF attack preferentially using the LF ATT slider can increase the warmth and impact of heavy percussion and bass instruments â€“ great for drum track mixes, bass parts and heavy dance tracks.
You can also do interesting stuff by messing with attack and decay time settings â€“ for instance faster settings are what I have used to make the analogue tape softening and aural exciter style pre-sets. And since the capture function is fully adaptive it compensates for your settings, so however wacky your creative timing settings are (even with attack times slower then release), hitting the capture button will drag all the average levels back to match the original track.
I could go on, but you probably donâ€™t want to publish an alternative operation manual 🙂
Can you tell us about any future plans? New plug-ins, porting, etc, etc.
Well, generally I personally would like to pursue genuinely new applications and explore new artistic areas â€“ including the many nuances that have been lost in the conversion from analogue to digital processes over the last few decades.
We live in an era I could not have imagined 30 years ago, where unprecedented signal quality is at our disposal for amazingly small costs and there are a many great plug-in products out there producing high standard results, which are technically speaking well beyond anything analogue could ever manage. So I feel that making yet more variants of what is already out there would be a wasted opportunity and not really what the users crave most.
Essentially sound engineering is a creative art and my deep passion is to appeal directly to this specific aspect of what people do, by providing new artistic dimensions and freedoms people can draw upon to make great art that we all, musicians, engineer and listeners alike want to experience. Like everyone else involved in creative art, I crave that precious feeling when something puts shivers down your spine â€“ because itâ€™s so darned amazing! Perhaps we are all hedonists at heart?
As designers we are a bit like humble service providers to the art itself and the people at the sharp end tasked with creating it. After a lifetime of involvement in striving to make those â€˜magical artistic momentsâ€™ by whatever means I could lay my hands on, at whatever task I was engaged in, I have a head full of sounds, nuances and impressions I would like to realise in future products. And like anyone else who has engineered for a living, I also have whole catalogue of frustrating technical and operational â€˜blocks to successâ€™ I have suffered, which I would like to break through and sweep aside for future engineers.
At this moment I canâ€™t actually say whatâ€™s coming next, but if in any way it facilitates any of the above sentiments, I will be more than content.
Big thanks to Paul for taking the time to answer. Expect a review of Dynamic Spectrum Mapper this or the coming week.