Dave
very active
Posts: 480
|
Post by Dave on Sept 15, 2013 19:12:47 GMT
Hi Guys, Just a simple question this time but possibly the answers may not be . Given the relatively small range of bitrates and frequencies and combinations thereof that music can be converted to (just thinking 'FLAC' here), is there universal agreement on what constitutes a HiRes file? Is it any file that has either bitrate and/or frequency above Red Book figures of 16 and 44.1 respectively or is it a little more complicated than that? TIA, Dave.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,882
|
Post by solderdude on Sept 15, 2013 20:06:08 GMT
HiRes simply means high resolution and they mean considerable higher than redbook (CD)
You can increase resolution in 2 ways, more bits and a higher sample rate (cut-off frequency) 24 bit can thus 'resolve' an analog signal in smaller 'steps' and thus 44.1/24 is already higher resolution.
It should be noted that although DAC's have improved considerably once standard 'ladder type' DAC's have been replaced by bit streaming (DSD or 1 bit DAC's) and the resolution of the electrical output signal went up.
Most DAC's (even those that claim 32 bit) will thus NOT be able to reach an effective 24 bit resolution in real life situations. Thereby questioning the high-res part in the actual output voltage of that DAC.
24 bit is actually really needed during the recording and production (mixing) process in order to get at least 16 bit resolution as an end result. Its why it is used in studios. That extra 8 bits are needed for headroom during recording and rounding off samples more accurately in the production process. In the end result the 8 bits extra are such small steps they can only describe noise in the recording more accurately. a 24 bit resolution DAC is also essential if you are using volume control from the PC/source itself.
Sample rate simply says something about how many samples (of 16, 24 or 32 bit depth) are used per second. With more samples you can describe higher frequencies more accurately. To do this you have to make really BIG steps. 44.1 to 48 doesn't make any difference in describing a 10kHz signal for instance. You need to at least double the frequency (so 88.2 or 96kb) Such a signal can describe a 10kHz with double the amount of points so only increases resolution mildly. (176)192 or (352)384 doubles that resolution again.
So 16/176 isn't a real increase in resolution in reality as it is just a factor 4 and only for higher frequencies while 16 to 24 bits is an increase of 256x in resolution.
24/192 is thus about a 1000x more accurate as 16/48 and is thus hi-res.
In the end it is NOT the format that defines the sound quality but the recording itself. A well made recording (even when published on 16/44) can be considered hires music as you can hear all the lovely nuances. Badly recorded music on 24/192 can sound MUCH worse than a well made recording published on 16/44, even when available on 320kbs/MP3. There are plenty of well recorded CD's around that sound amazing. There are also quite a few DSD's and 'hires' uploads available in 24/192 that are nothing but upsampled/converted 44.1/16 masters.
How MUCH of an increase you get in electrical performance (DAC output signal) depends on the ability of the DAC to create very small voltages accurately. This is known as ENOB (Effective Number Of Bits) which for ALL DAC's is (considerably) smaller than the 'format' it can handle.
|
|
Dave
very active
Posts: 480
|
Post by Dave on Sept 15, 2013 20:26:09 GMT
Thanks Frans, Now that is what I call a really comprehensive answer to my question and, what's more, one that I can fully understand, so may thanks for that - I'm learning little bits (no pun intended ) slowly. I'm about to post another question that you may be able to advise on, not just for me but others may be able to benefit from it, so watch out over the next 15 mins. or so. Cheers, Dave.
|
|
Javier
Administrator
Digital bytes
Posts: 987
|
Post by Javier on Sept 16, 2013 8:51:18 GMT
Don't forget DSD which is a completely different beast.
Instead of using "long" words (ie number of bits) like PCM encoding uses, DSD uses a single bit but with extremely high speeds producing an analogue like wave form. Basic DSD (aka DSD64) like the one we can find in SACD disks samples at 64 times the speed of regular CD (16/44.1KHz), that is 2,822,400 Hz Vs 44,100 Hz. Some material is becoming available in DSD128 which doubles the sampling rate to 5,6MHz (5,644,800Hz). Some players like JRiver, HQPlayer or Foobar can do real time conversion from PCM to DSD up to DSD512 (JRiver tops at DSD128) which means an astonishing sampling rate of 24.576Mhz!!!! HQplayer is the only software currently available capable of doing DSD to DSD upsampling staying within the DSD domain. In the DSD ultrasonic noise is the main problem. When using the lowest sampling rate (DSD64), ultrasonic noise starts increasing rapidly after the top of audio band (+20KHz) and needs a very steep filter at around 50KHz. Using higher speeds reduces the problem by displacing the noise higher up, doubling the speeds doubles the point where the noise starts building up, but has the inconvenient of doubling the size of the resultant file. An uncompressed DSD64 song of 4 minutes of length will produce a file of approximately 170MB, the same in DSD128 will almost double to 340MB, further increasing to DSD256 will again almost double the previous to about 630MB and DSD512 will produce a huge 1.2GB file. There are currently two DSD file formats, DSF and DFF, the former allows file tagging like MP3 or FLAC but no compression is possible. The later is the opposite, offering no possible tagging but allowing for compression through the DST algorithm cutting the file size to nearly half the size of the original.
DSD was originally used as an archiving format. It was a direct unprocessed dump from the output of Delta Sigma (DS or SD) analogue to digital converters prior to PCM conversion. It works much like the RAW format of digital cameras before conversion to JPEG. As Delta Sigma has become the mainstream technology for quite a few years, it is possible to say that with a very, very few exceptions all recordings made in the last 10 to 15 years were born as DS.
As conversion from DSD to PCM, a process called decimation, is lossy, one could argue that DSD is high resolution at least in the sense that it is more faithful to the original. Of course that would be true as long as there is no subsequent processing of the signal that implies any DSD to PCM conversion. The problem is that tools that allow editing of pure DSD are scarce verging on non-existent and terribly expensive so the chances of pure DSD recordings are extremely limited unless they are direct transfers from analogue tape masters to DSD.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,882
|
Post by solderdude on Sept 16, 2013 9:54:24 GMT
DSD64 has 'about' the same resolution as a 20/96 PCM file albeit with different noise levels (higher for DSD) and different 'problem areas' and properties. This means that DSD can also be considered HiRes as it can resolve smaller signals more accurately than 16/44 can.
A DSD-'signal' is a very old format already which predates the digital audio area (introduction of CD around 1983) as it was invented in 1954 already (yes, 60 years ago) The first practical (and very crude) PCM signal transmission was also around 1950 but the basis for it was 'invented' already 100 years earlier and was a way to 'multiplex' multiple telegraph signals via one wire.
Also one should realise that all Delta Sigma DAC's (a lot of them are such these days) basically put out a DSD stream and the DAC chip itself performs the PCM to DSD conversion. It stands to reason, if a recording is made in DSD format) to skip the DSD->PCM->DSD conversion. a DSD stream (after channels have been pulled apart (de-multiplexed)) is closer to an anlog signal and has different impulse reproduction. Now we see more and more bitstream DAC's but they were already made in 1988 by Philips and Sony had quietly (they didn't know how to sell 1 bit in the growing amount of bit DAC's) applied those types of DAC's and nobody noticed the Sony's would be souning different in those days. Technics had their 4 bit 'Mash' converters with noise shaping technology and made lots of advertisements to 'sell' it as better sounding, yet Sony had already quietly introduced 1 bit converters.
Both DSD to PCM and PCM to DSD conversion are 'lossy' because of the completely different 'mechanisms' involved.
a DSD64 stream works with a frequency of 2.8224 MHz (1.4112 MHz per channel = similar to 2.8224Ms/s) where a PCM stream 16/44 has 1.4112 Mb/s sample rate (not counting tags and error correction overhead and for stereo) which comes down to a frequency of 0.7056 MHz which is 4x slower than DSD64.
|
|