I will admit upfront: I am not an analog guy. But in the world of audio, analog is king. The digital world, however, is bourne out of practicality. In this post, I’ll examine some formats and pros/cons of them.
It is practical to store digital recordings. A digital recording does not degrade like an analog recording does. An analog representation of a bit may degrade, but either the bit is correct or it is not, either it is zero or it is one. But because it is represented in an analog medium, it still has some degradation that can happen, but in a more reliable way where you can verify if degradation has taken place or not – via checksums or more formally hash function values; hash functions are mathematical functions that map one set of numbers to another set of smaller numbers.
These functions are used to check integrity of stored data. If data is rehashed, then the new hash number must match the old hash number, or the data has changed. So instead of keeping two identical files and comparing them bit for bit if they are identical, one can just generate a hash value from the source file and use that to later check against new hash values from the same file to see if it has changed in any way.
The functions are also used in exchange protocols and networks such as Peer to Peer (P2P) networks to ensure integrity when files are searched for and multiple fragments fetched at the same time from multiple peers.
So the hash values are just numbers that are associated to the data via a hash function. There are several functions, algorithms and lengths. Since the target set is smaller than the source set (the hash values have a fixed length), there is always the possibility of collisions, meaning two different source numbers will be transformed into the same target numbers. It may seem strange for you to think of digitized music as numbers, but remember digital files are just series of bits, and a series of bits is just a binary number, and there is a one to one mapping from any number in any number system, to another number in any other number system. So much for theory related to data integrity.
So, the bottom line is that digital storage and processing is practical and valuable.
The problem with digital audio is two- or three-fold:
- faithful digitization (analog to digital conversion (AD))
- faithful “analogization” (digital to analog conversion (DA))
- faithful mixing and processing of digital audio data
Since the advent of the Compact Disc (CD) format (and the Red Book standard), which specifies digital audio storage of 16 bit 44,100 Hz stereo resolution, requiring a bandwidth of 2 * 16 * 44,100 bits per second (b/s), i.e. 1,411,200 b/s, a debate has gone on about the virtues of digital vs analog audio.
Opponents of digital audio will invariably say that digital audio will never sound analog, that digital audio is an analog signal hacked to pieces, and that the analog source curve can never be reconstructed. This is in fact true. It never can be. The question is, how much precision is necessary before further precision becomes irrelevant in the face of human hearing limits and reproduction system capabilities.
Then there is the proces of actually converting the digital signal back into the analog domain (or vice versa, from the analog domain to the digital domain). Some will say that this can never be done with the precision and faithfulness needed. I don’t believe this to be true and while I do not know AD/DA system is the best on the planet so far, I have no doubt that the very impressive technology we have to day will mature even more in the future and will finally put this analog vs digital debate to rest. I believe we will have a future without compromise, or, a future without compromise of relevance. That is to say, or put another way: if a tree falls in the forest, and you’re not there, do you really care about it. It does happen, but if it is not physically possible for you to sense or percieve it, on any level, then it will have no relevance to you.
So what is digital audio anyway.
Well, digital audio, as you may have guessed, is not just digital audio. 😉
There are two fundamentally different ways to deal with digital audio. One is based on a single-bit representation and and the other is based on a multi-bit representation. The single-bit, or delta, idea is simple. Each digital sample is represented by only one bit, but the sampling frequency will be in the megahertz range to increase accuracy. The other idea is to use multiple bits for each sample and then use a less extreme sampling frequency. This is called Pulse Code Modulation (PCM).
It is possible to convert between these two codings. One may think it possible to convert from one to the other, losslessly, without distorting the information. But is in fact not possible to convert audio from a single-bit (s) representation to multi-bit representation (t) and back (s’) and have the s and s’ be bit for bit identical.
Each format has its virtues but the main problem with delta signals is that they cannot be easily mixed. They have to be converted to multi-bit form first and then mixed. So delta form is best used when the digital representation will not be mixed and manipulated afterwards, otherwise it doesn’t make that much sense.
Some AD converters work in a two-stage proces where the source signal first goes through a delta converter which is then followed by a multi-bit upsampling converter.
So what formats belong to what paradigm?
- CD (multi)
- SACD (delta)
- DVDA (multi)
- DXD (multi)
CD is the Compact Disc format we all know. It’s 16 bit 44,100 Hz stereo, unencrypted, unwatermarked (to my knowledge). There is copy-protected bastard offspring from the Compact Disc format, but it is not compatible with Red Book and cannot be called a Red Book Compact Disc.
SACD is coded in Direct Stream Digital (DSD) and then successively coded in Direct Stream Transfer (DST). The first format is a delta representation, the second a compression format for the DSD delta form. SACD has a maximum resolution of 1 bit 2,822,400 Hz ?-channels (the frequency equals 44,100 Hz * 64). It also includes Pit Signal Processing (PSP), a copy-protection mechanism used to watermark digital audio.
There is a lof of theory around converting between delta and multi-bit, but I’m not very familiar with it, so I’ll not delve into that here, interesting as it may be.
DVDA is well known and has a maximum resolution of 24 bits 192,000 Hz stereo. DVDA is encrypted.
So what is this DXD?
Digital eXtreme Definition (DXD) is another, newer, format which takes high-resolution to the next level. It specifies an audio signal as 24 bit 352,800 Hz data. It is a high-frequency format, 8 times as frequent as CD audio, and uses 8 more bits per sample than CD audio.
I came across DXD when looking at DA hardware. I found the DAD AX24, that is, the Digital Audio Denmark AX24 converter. It can be configured as both an AD and a DA. Denmarks Radio (DR) apparently uses this converter to record classical concerts in DXD. So I recon’ it must be quite good.
“The main benefit of DXD is the much better impulse response which are capable of capture the ambience around the instruments”
Since it has both AD and DA it can actually be used to record live events. I mean, where else do you get DXD data?
That is a good question! But actually Linn Records has recently unveiled online downloads of Studio Master quality digital recordings in the FLAC format. It’s not quite DXD, but it’s better than CD!
I don’t know how good the AX24 actually is, but to me it looks like a very interesting product and with online downloads booming, I hope to see hires formats boom as well.
We’ll see in due time what the future holds for us…
- Digital Audio Denmark (DAD)
- Digital eXtreme Definition (DXD)
- DAD AX24
- DXD Recordings
- DXD at Danmarks Radio
- DXD Discussed
- Direct Stream Digital (DSD)
- Pulse Code Modulation (PCM)
- Linn Records
- Hash Function
- Cryptographic Hash Function