<<RETURN TO FAQ PAGE
Why is software the key to Ultra-Fidelity?
An article by John Hughes
We at VRS were about the first in the pro world and audiophile world to realize that different playback software packages sounded different. In our early investigations we were also shocked to find that the developers of these packages all believed that all software sounded the same as any other software. Even their own packages that they had spent untold hours building, sounded the same to them as their competitors. This led to a four year quest to understand why software sounded different and how we could improve it. We do understand most of the reasons today and we have been in the process of developing our own software in conjunction with some of the best minds in the field. Our system should be out by early 2009, and we will have a preview available at RMAF 2008.
The following article is an attempt to explain why software sounds different, and what is needed to create good sounding software. I have tried to abstract some of this to layman's terms so that most people can understand it. Some of this abstraction might sound too vague to some pundits however, so I apologize for that ahead of time.
A bit of background on how software is developed.
98% of all software developed uses a few basic development packages, and are built from components, very much the way your computer has a power supply made by company X, a CPU by Intel, a motherboard by company Y. Most software is the same way. Part of the software (the user interface for instance) is built from a module that comes with the development package the programmer uses. All of the consumer playback packages that I know (iTunes, Windows Media Player, WinAmp, Foobar, JRiver etc.) and some of the pro audio suites as well, buy a license (or maybe own in house in the case of WMP or iTunes) for the audio handling modules that are used for audio playback. These multiple modules handle most of what happens to the audio in a playback software. An example of this software is Fmod (http://fmod.org/) which is widely used in audio players and might be the basis of Foobar and Winamp.
The programmer will take these modules and hook them into their program by plugging in software inputs and outputs, while using certain ‘filters’ or transforms available in the package. The most common being support for DirectX (the dreaded Kmixer) and volume control. All of these packages are based on sound architectures that are at least 10 years old (more like 15 years in most cases) and were never developed to be audiophile. In addition the hooking up of the module to the music file, transforming it for the player, and sending it to the DAC, all take mathematics. The math the modules use as a whole is not that great (or accurate) , and can change the sound. The basic math all these modules use was based on accuracy for 12 to 14 bit audio, and while the modules have grown to plug into and support more bits, like 24/96 or 24/192, the math inside the modules them is usually only accurate to 14 around bits.
Also, the math the programmer uses to hook it up is usually not very good from an accuracy point of view. In fact if a programmer spent the time to create the best math for the module they are using, it would probably add significant time and cost to the project. And bear in mind that virtually all programmers don’t believe there is a difference you can hear, so why would they bother? Also the math required to do this is very advanced, and not something most programmers know. When you do an equation that gives you a 99.9% accurate answer, then you multiply that by ten different transformations, you begin to have a significantly large gap in accuracy. That gap has a direct correlation with distortion of the audio signal. It is almost impossible to measure this accuracy with today's software measurement tools.
Another factor is that consumer players are built to work and mix into whatever other audio is running in a computer, the system sounds, another music source etc. Adding this 'mixability' into the software severely compromises the accuracy of the player. The best known aspect of this is the dreaded Windows KMixer, but Apple's Core audio has similar problems too.
If Apple were to decide to take the sound engine that is in iTunes and rework it to the highest pro audio standards it would probably cost them 5 million dollars (this is not just for development, but also for compatibility and beta testing etc.). It’s unlikely that Apple would ever decide this was worth the cost and aggravation.
There are a few pro audio software companies that have taken a different approach. Some of these companies have been making their software for 20 years! When they started making pro software, there was nothing available that could do the kind of instant transformations, handle more than two channels, and have all the channels arrive to the DAC or the recorder exactly synced after processing. So they set about building it themselves. Since they were working in an industry that valued accuracy over everything, most of them spent years creating an audio engine that was highly accurate. This meant they could not use most of the shortcuts and third-party modules available and had to create everything from scratch. This took years of work and great expense. But the resulting software has much more accurate math, and requires many fewer transformations to get the sound out to the Dac (in some cases reducing transformations by a factor of ten).
One of the pro audio audio modules that was developed this way is ASIO. As many know, using ASIO drivers with playback software can improve the sound quality, and deliver the music to the Dac through a clearer path. ASIO can't improve the erroneous math that happens in the player itself however.
In order to make a layman's comparison of the quality difference between software I would compare all the consumer playback software to varying CD players in the $100 to $200 range. Unfortunately, this is all most of us have been exposed too. Foobar is more like the $200 CD player. But serious pro audio software sounds more like a $3,000 to $10,000 CD player.
And to continue in the audiophile simile, the source is the most important piece of a system. Hooking a $200 CD player to a $5,000 amp is futile. Similarly, using iTunes to play music on a $5,000 Dac is also futile. This is really the basis of VRS’s dislike of iTunes, Foobar, etc. This is High-End audio, not Best Buy, and audiophile customers deserve better.
Since almost all audiophiles, and most high-end manufacturers have only heard the "$200 CD players" of software, they do not have a frame of reference in which to discriminate. Since we have not been able to demo what this type of software-based system can do, except in a limited basis at CES this year, and in a few demos on our systems, there is healthy skepticism out there about our claims.
The good news is that these differences are profoundly self-evident when you put anyone who likes music (audiophile or not) in front of a demo of really good mastering-quality software. It is this quality of highly accurate playback that we call Ultrafidelity.
