Our Audio Formats
At ProMates our ultimate aim is to provide the best HD music quality and, if possible, deliver at the sample rate at which it was originally recorded or mastered. We offer a growing selection of music and as an alternative you can purchase many of the albums in a lower sample rate, as not everybody needs the highest possible definition.
If you have purchased a music download with a sample rate between DXD and 44.1 kHz 24 bit, you get an MP3 download free of charge. You can find the available download directly from your account page and you will be able to start to download each music file 2 times.
DXD or Digital eXtreme Definition is a PCM audio signal with 24 bit resolution that is sampled at 352.8 kHz. The format was initially developed for Merging’s Pyramix workstation as a recording and editing format for SACD’s. Since the technology was introduced in 2004, many of the Classical albums rewarded with “Best Engineering” or “Best Recording” album awards, have been recorded in DXD, then converted to DSD and then published as an SACD. Please also visit: this article in Stereophile.
Some modern music stores have a built in sample rate conversion utility. One of the reasons for this is that it is very convenient for the store, since they only have to upload an album at the best sample rate, and then select which formats they want the album to be published.
In the professional audio world however, it is widely recognized that there is a huge difference in the quality you receive from different sample rate converters. The very few existing, top quality, sample rate converters are all either stand-alone applications or built into professional digital audio workstations.
At ProMates, we didn’t feel that was good enough. We have therefore chosen to generate all our down samples with Pyramix DAW, and only convert samples between DXD, DSD, 176.8 kHz, 88.2 kHz, and 44.1 kHz or between 384 kHz, 192 kHz, 96 kHz, and 48 kHz in order to minimize the impact on the quality of the sound. Since we are generating all our downloadable files in-house, we don’t sell single tracks.
DXD is 352.8 kHz/24 bit and compressed with lossless FLAC.
DSD or DSD64 is the one bit format used for SACD’s, sampled at 2.8224 MHz and compressed with lossless DSF or DSDIFF
PCM 44.1 kHz to 192 kHz/ 24bit is compressed with lossless FLAC and AIFF-C.
MP3 is in most cases compressed to 320 Kbps.
When you have made a purchase from ProMates, you will normally be able to download a digital booklet in pdf format.
At ProMates Music Store, you are also able to purchase an entire sound hardware/software solution capable of playing back all sampling rates, including 384 kHz/24 bit, DXD, DSD64, DSD128 and DSD256.
Digital HD Audio Formats
It has been suggested that since most humans can’t hear much above 15 or 20 kHz, there is no need for HD audio with higher sampling rate than 96 kHz. In response to this, we simply ask that you use your ears; It is more persuasive to listen and observe the clearly audible differences between the HD formats, and then try to find an explanation for these differences, than simply to dismiss them as unnecessary based on nothing more concrete than an uninformed opinion.
What would be helpful to ascertain the validity of HD sound technology would be more scientifically controlled double blind listening tests. The tests that have been conducted thus far have had somewhat mixed results: As evidenced by this test conducted at McGill University in 2007, and a similar investigation undertaken by the Boston Audio Society in the same year.
On the other hand, we know that many audio professionals record in very high definition on a daily basis, and we seriously doubt that all these audio professionals, experts in the medium of sound and acoustics, can be completely wrong.
So, in an attempt to understand the issue more clearly, let’s try to look at some essential characteristics of the different HD formats:
The first area of interest is known as Aliasing Intermodulation Distortion (AID). AID is actually a combination of two issues. If an AD converter is present for a signal above half the sample frequency (Nyquist), these signals will be mirrored around the half sample frequency (sf/2) and create new frequencies without harmonic relationship to the original content (Aliasing)
As a result, AD converters have added a steep low pass filter that, with very few exceptions, commences suppression at 45% of the sample rate and achieves full attenuation at 55% of the sample rate. Due to the filter characteristics, this type of Brick Wall filter does not attenuate much at sf/2, and a significant part of the content, with frequencies just above sf/2, are therefore mirrored around sf/2.
Usually these mirrored frequencies are inaudible. Despite this, a problem occurs when they are reproduced in a loudspeaker, and to a lesser extent, when they pass through a non-linear electronic circuit like a power amplifier. Here all frequencies close to each other intermodulate and create different frequencies (f1 and f2 = f1-f2 and f1+f2).
If these new frequencies have a harmonic relationship, then this issue doesn’t represent a problem, however if some of the content has been mirrored around sf/2 and has therefore lost its harmonic relationship with the rest of the content, this intermodulation will create new audible unharmonious frequencies. Typically in the area between 1 and 5 kHz, precisely where they dramatically affect the quality of the sound..
Figure 1 shows what happens when a sample with a frequency of 4.5 kHz, and the natural harmonics of 9, 13.5, 18 and 22.5 kHz are AD converted with a 0.45/0.55 filter at a 44.1 kHz sampling rate. As can be seen, the frequency of 22.5 kHz will be dampened somewhat by the filter, and then mirror around sf/2, and this creates a new frequency of 21.6 kHz. 21.6 kHz will then intermodulate with 18 kHz from the original content, and create a clearly audible frequency of 3.6 kHz.
AID is without doubt contributing to the “hard sound” of a CD, and the effect can, to some extent, be compared to an detuned piano. If all three strings which create the sound of a key are in perfect pitch, the sound will be warm and rounded. If only one string is slightly out of pitch, the sound will be harsh and shrill.
The effect of AID is minimized at higher sampling rates, due to less amplitude of the frequencies close to sf/2. It is worth noting that top microphones, such as the DPA 4007, have an almost flat frequency response up to 40 kHz and thereafter only suppresses at 6 dB pr. octave.
It is indeed possible to manufacture an AD converter with full attenuation at sf/2 and therefore with no AID. The problem for audiophile music lover is that most music is recorded with a 0.45/0.55 filter at all sampling rates.
The good news is that AID is not a problem for DSD and DXD, because the amplitude of the frequencies above sf/2 is well below the noise floor.
Another byproduct that is worth mentioning is the different audio formats ability to reproduce a pulse. Sound is composed of different frequencies, with different amplitudes and of pulses: A pulse, for example can be the sound heard when a string is struck.
As can be seen in Figure 2, DSD has an almost perfect pulse response of a 3 us pulse. The pulse response for DXD is also great, and then the amplitude is reduced at lower sample rates.
Before and following the digital representation of the pulse, one can also see pre and post temporal ripples. These ripples are caused by anti-aliasing filter, and the steeper the filter is, the more of the energy of the pulse will be transformed into pre- and post ripples.
Likewise, it is possible to minimize the pre-and post ripples, especially at high sampling rates, by using a softer anti-aliasing filter. Unfortunately, only the AD converters that can also be used to record in DXD or DSD, have the advanced technology required to alter the filter steepness. Other AD converter manufacturers build their machines with the 0.45/0.55 filter. This comes with the converter chip, also at 192 kHz sampling.
How much amplitude and timing of the pulses affects the quality of the sound and the perception of a stereo image, have not yet been fully investigated scientifically.
Figure 3 shows the noise spectrum for various formats after conversion in a typical high-end AD converter.
As can be seen in the graph for DSD, conversion increases the noise dramatically just after 20 kHz. The total amount of noise generated is so great that it represents a problem for amplifiers and speakers. Phillips, who owns the SACD and DSD format together with Sony, has therefore specified that SACD players must use a filter which attenuates noise above 50 kHz.
Unfortunately, using this type of filter means that the otherwise perfect pulse response of DSD is suppressed to a significantly lower level than the pulse response of DXD.
The quantization noise of DSD128 is similar to that of DSD64, but with proper noise shaping the noise starts to raise one octave higher at around 40 kHz. This is a huge improvement compared to DSD, but a low pass filter is still needed in the DAC.
Lately we have also seen album releases in the DSD256 format. Here the noise first starts to rise from around 80 kHz, and it is therefore only necessary to deploy a soft low pass filter in the DAC from 130-150 kHz, with minimal attenuation of the pulse response. As a result, the total pulse response of DSD264 exceeds the pulse response of DXD, and the net result of this is a very small but identifiable improvement in the audio quality.
All types of music usually requires some editing and mixing before it can be released. Since any kind of DSD is a 1-bit format, it cannot be edited without having to convert it to either DXD or the slightly older DSD-wide. This procedure tends to cause a problem in that further quantization noise is added to the signal when it is converted back to DSD. It does not take many conversions or too many multi tracks added to the mix before the cumulative effects of the noise reach problematic levels.
Furthermore, it is impossible to perform sample rate conversions back and forth between formats without any impact on the sound quality. If editing and mixing is needed it would be better to release albums in DXD instead of DSD256, since the file size of DSD256 is bigger than DXD files, and the impact from the sample rate converter is greater than the small differences in sound quality between the formats.
For the reasons outlined above, the majority of the recent SACD releases, especially those which have received awards, have tended to be original recordings, that are edited in DXD and then converted to DSD only once.
However, there are also many examples of 96kHz recordings that have been published both on CD and SACD, that have given SACD’s a bad reputation..
In the modern world, the reality is that most people download their music from the Internet, and the ready availability of bandwidth and increasing memory will likely make the common usage of the highly compressed mp3 format redundant. A DXD file is approximately three times larger than a DSD file and 12 times bigger than a CD of the same length. In the not so distant future due to our improving technology, these differences will likely be irrelevant.
I have been one of the driving forces behind the successful introduction of DXD to audio pro. I have spent several years traveling from one high-end studio to another, completed many listening tests and demonstrated the different types of audio formats. Today more than 400 of studios and concert halls have purchased DXD equipment.
For me, it is clear that there is an undeniable relationship between resolution and quality. Each format seems to have has its own expression and conversely, its own limitations. To reproduce classical music, for example, I have always preferred the quality of DSD.
Yet, it was not until I experienced DXD that I was no longer able to hear the signal as a digital reproduction. In contrast, DXD is calm and warm, with a deep, resonant and well defined stereo perspective. So much so that it evokes memories of the heady days of analog.