|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
192.181.133.141
An audio buddy of mine recently sent me an email in which he quoted his friend talking about the introduction of digital into the telephone industry. His friend worked for AT&T and was instrumental in incorporating digital technology to replace the analog system in use at the time. Here's what he wrote:
After reading this, it occurred to me that this might be a good method to use for comparing the various digital formats being used in audio today. The method to evaluate and compare the different digital resolutions would be to multiply the sampling rate by the quantization bit rate for each digital format. That way the various digital formats could be compared on an equal basis. For example, the resolution of 16/44 Redbook would look like this:Redbook resolution = 44,100-Hz x 16-bits = 705,600-bits/s per channel
DSD64 (SACD) resolution = 2,822,400-Hz x 1-bit = 2,822,400-bits/s per channel
If we allow 16/44 Rebook to be the standard to which the other formats are compared, we can see how much higher resolution the other formats provide. For example, here's how SACD compares to Redbook:
2,822,400 / 705,600 = 4.000-times Higher Resolution than Redbook
Here's a table of how some of the other digital formats compare to 16/44 Redbook:
As can be seen from the table, DXD has a resolution that falls in-between the resolution of DSD128 and DSD256. It also shows that 24/176.4 and 24/192 PCM both have significantly higher resolution than standard SACD. I have a feeling that DSD512 might be overkill when you compare the additional memory required for storage. On the other hand, DXD might be the most efficient digital format from the standpoint of digital resolution relative to the amount of storage memory required.Anyway, I thought this type of comparison was rather interesting so I wanted to see what some of you other digital enthusiasts think about this concept.
Thanks!
John Elison
Edits: 09/02/20Follow Ups:
It IS an interesting analysis however you cannot equate bits per second to resolution. Imagine in a different domain that we have 4k uncompressed video (oh to dream!) at 30 Hz. So that's like 740MB/s . Now suppose we go back to SD 480 and it is just 28MB/s. BUT now let us send that out at 1000 Hz! Whee!!!! Now we are at 28,000MB/s and we have achieved "38 TIMES" as much resolution. Of course, not really, the picture will still not be as clear as the 4k. Because, the resolutions are in different domains.
Doesn't the same sort of thing apply in sound? There's a magnitude domain and a time domain, they are apples and oranges, you can't just multiply all the bits. DSD has far less magnitude information but much more often, it's a different animal with different problems. Bits aren't "resolution" they can just be used to achieve it (whatever "resolution" is which could be a whole other thread).
Yeah, the numbers are just grabbed off a calculation site and maybe not correct, but it's the principle of the thing. I own a number of SACDs I really like, however sorry Sony's original graphic of some kind of pure reconstruction is pretty much silly nonsense, as is the "256 times" resolution claim on the HDTT piece.
Plus it is all for nought if you're just scraping the same old ancient degraded tapes yet again, with no hope to improve anything. What, now we need DSD4096 to re-re-re-re-re-re-re-release "Kind Of Blue" because there was some nanogram of musicl information that had never been recovered off the tapes previously? [rolling eyes emoji]
I find 96/24 to offer the best overall compromise in true musical resolution. If a recording is mastered in 192/24 as are many classical titles, I choose that because downsampling involves compromises. I have no interest in upsampling even when that "softens" the results as you find with DSD content. I'll take the original.
I have many 176/24 recordings that were originally released in SACD. DXD was originally introduced for editing SACD content.
I think what you have is an analysis of raw file size, but not entirely correct because you can use lossless compression on those PCM files to make them smaller w/o any loss in audio quality. For example you can compress with FLAC or ALAC and upon playback there's zero loss in quality.But I don't think it's really a comparison of resolution, just file size and maybe bit-rate for streaming. Additionally, you reach a point of diminishing returns rather quickly in terms of actual audible differences.
IMHO the biggest factor is how well the original material was recorded / mastered. I'd take a well recorded and mastered CD 16/44.1 file over a "higher resolution" 24/192 or DSD512 file from a lesser quality recording any day.
And for folks who rip vinyl, I doubt that there's anything to be gained going beyond about 24/96 as there's no more useful information or dynamic range than that coming off the LP.... perhaps a bit more noise to pickup but that's about it.
Edits: 09/05/20 09/05/20
You're right! It's just a file size calculator. I just didn't think it through completely.
Thanks!
John Elison
It's still an interesting analysis.
If you take the bit-rates and calculate them out to Mb/s for 2-channel audio and compare that to the internet service that you have, or cellphone data plan, you'll see what your minimum bandwidth requirement must be for streaming.
In the case of cellphone data, you will also see that "hi-res" will quickly eat into your data plan - unless you have an unlimited plan. But even that gets tricky because some cell services severely cut off your bandwidth if you go beyond a limit - even for certain so-called unlimited plans.
Okay! Here's the bit-speed for the various formats. However, this doesn't include the reduction from FLAC compression.
Edits: 09/05/20
.
Gsquared
Without understanding the information value conveyed per bit, "Bit density" is a meaningless metric to use in comparing PCM to PDM ("DSD").
What matters is the level of information that is conveyed per unit time. "Resolution" relates the accuracy of the decoded information per unit to represent the original signal. Since a 1 bit converter only has two states compared to a given N-level multi-bit converter, the DSD format needs far more symbols to be transmitted per unit time to convey the same information as a multi-bit converter. DSD is therefore the least efficient way to convey information and why coding in DSD results in larger file sizes. The problem to solve is to recover the encoded signal with the least error.
Back in the 80s and 90s due to the manufacturing challenges in achieving perfect trimming of the resistors down to the LSB, multi-bit converters suffered from poor linearity which is where the shift to PDM gave a theoretical advantage. However, the code/decode accuracy is highly dependent on the level of jitter on the reference clock signals. Additionally, coding a signal with a 1 bit converter results in residual quantisation noise due to the fundamental inability to adequately dither the signal compared to a multi-bit converter - 3 to 4 bits is required for perfect dither.
Therefore any claim of the superiority of DSD to PCM is pointless - the question should be which method gives the smallest error signal between the recovered signal and the original signal. Unfortunately that comes down to the equipment used.
Regards Anthony
"Beauty is Truth, Truth Beauty.." Keats
...should be on the same page
Regards Anthony
"Beauty is Truth, Truth Beauty.." Keats
brought this link over from Classical & thought it might be pertinent as the same subject is being chewed on:
best regards,
I find John Ellison's analysis of digital resolution relevant and useful.
The bit length x sampling frequency, which he listed for various formats in comparison to Redbook CD format, expresses the potential resolution of those formats, eg. the obtainable throughput (or bandwidth in its digital sense) of digital information (one can use either bits or bytes).
Insofar as digital audio entails conversion of analog signals to digital signals and then back to analog signals, and neither signal can be more than an approximation of the other, higher potential resolution allows closer approximation of the signals.
Of course what one hears is what matters. The ear has differential perception of frequencies much as the eye has for instance increased perception of tonal difference in shadowy verses brightly illuminated scenes.
To my elderly ears, higher resolution digital music offers audible improvement over lower resolution digital music at least up to, as John Ellison listed, 13.06 times CD resolution, ie that of so-called DXD. For me the audible difference inaudio format, DSD versus PCM, diminishes with increasing resolution such that at DXD or DSD256 the difference is scarcely if at all perceivable.
Let me conclude by citing Rushton Paul's newly released article in 'Positive Feedback' where he notes that several studios now record in DXD or DSD128. Having purchased a few albums in DXD format, and aware that recording technology continues to improve in respects other than digitization, I consider that money well spent.
Seventies
I'm not sure I understand it completely, but I'll study it some more.
Thanks again!
John Elison
/
Do you get the implications of Roseval's chart? This fetish for higher and higher sampling rates (to the exclusion of everything else) continues to amaze me.
(And, BTW, DSD128 pushes that noise up an octave, while DSD256 pushes it up another octave.)
.
.
.
Noise?
You can't hear it, I can't hear it.
I've seen hi-res bashing before, some think 24/96 is too much.
You seem to have a Thing about DSD noise.
You don't think the smoother wave form of DSD to be important either.
Hey, it works for me, I never think about this stuff, except here!I have done Live Recordings in 24/96, 24/192, and DSD 64.
DSD is so much more relaxed and natural.
If I could record in DSD 128 or 256, I would,
but like the Portable recorder format, not so much gear to lug and set-up.
Edits: 09/04/20
Furthermore, some folks say that DSD's UHF noise interacts with and interferes with audible frequencies too. I wouldn't know, but other listeners on this board have frequently commented on the unnatural "softness" of DSD and have theorized that that UHF noise may be a factor in this perception. If you like it though, I'm not going to argue with you. As far as the shape of the waves is concerned, do you listen with your eyes much? You wouldn't be the first to be duped by that.
Nevertheless, I'm very happy for you if you prefer DSD - Knock yourself out!
Do you like vinyl? If you do, it's because you listened to it and decided it sounded good. You certainly didn't measure it and discover it has so little distortion it had to sound good. Of course not! Vinyl has more distortion and noise than just about any other format I've encountered. Furthermore, the distortion and noise is nearly all in the audible range. Therefore, the only way anyone can like vinyl is if they use there ears to make that decision. If that's the case, you should listen to DSD256 and let your ears be the judge.
Good luck,
John Elison
When CD's came out, I couldn't unload my LP's fast enough. Regarding DSD256, I already agreed with you that it sounds great, even though my preference is for PCM. (See further below.)
All I'm objecting to are these DSD-Yahoos who think that, ooh!, 0-to-100kHz frequency response (or whatever) is the ultimate - to the naive exclusion of any other consideration.
Go back to making Babe Posts, that suits you.
I'm a firm believer in letting my ears be the judge. With respect to audible sound quality, there's nothing better than DSD128 and DSD256. At least that's what I hear in my system.
With respect to PCM, DXD would be my choice relative to sound quality.
Best regards,
John Elison
The few tracks I have in DSD 256 are the best sounding tunes I've ever heard. I prefer it to DXD. The problem is that most of what I find in DSD 128 or 256 is just audiophile stuff that I'm not interested in.
Well, if you like vinyl, you can always copy your favorite LPs in DSD128 with a TASCAM DA-3000 DSD recorder . That way you can have DSD128 recordings of music you like. However, I've found a lot of commercial DSD256 recordings from places like Native DSD sound good to me.
Best regards,
John Elison
I'd rather just play the LP instead of spending $900. That's 5x above what I paid for my entire digital setup.
I've downloaded all of the NativeDSD and SoundLiaison samples, plus I tried the Carmen Gomes in both DXD and DSD256. And that is the problem. I can't stand her voice and much of the other stuff on there is also recorded for audiophile recording's sake. It just isn't music I like and just think of it all as the "Norwegian Sound". And having recently been to Norway on a family vacation, that's what the music there sounded like to me even performed live. Always just a bit too laid back.
> I'd rather just play the LP instead of spending $900. That's 5x above what I paid for my entire digital setup.
Okay! Different strokes for different folks!
To me, $900 is downright cheap.
My turntable with arm and cartridge cost more than $16,000 and my phono-stage cost $4,200. My speakers combined with Rythmik servo subs cost over $17,000. I just bought a $3,000 Mytek Brooklyn Bridge and a pair of $6,000 PS Audio 1200-watt monoblock amplifiers . Therefore, $900 is downright cheap to me.
To each his own!
Good luck,
John Elison
I doubt I could afford the turntable mat for your system ;)
And certainly not the electric bill!
If art interprets our dreams, the computer executes them in the guise of programs!
Which format do you like best?
I have more than 40 DSD256 albums and I think DSD256 really sounds awesome. I also own a TASCAM DA-3000 DSD recorder and I've made 65 DSD128 recordings from my vinyl records. Recently, I've heard some DXD recordings and I think they sound very good, too. Actually, I like the sound of most hi-rez digital formats including some of the newer 16/44 CDs. I've been collecting hi-rez digital recordings for several years now.
Best regards,
John Elison
I'm not questioning your preference for DSD playback vs PCM. Personally I have settled on 24/176 for convenience in editing and format conversion.
However, out of interest, have you converted your DSD files back to PCM using the Tascam software and compared the result? If so, do you still prefer the DSD version? Similarly have you converted a redbook file to DSD and compared back to the original and had the same preference? If so, then what you are hearing has more to do with the decoding phase rather than any inherent advantage of recording in DSD. dCS introduced DSD conversion in the Purcell (IIRC) and reviews were mixed between DSD and PCM upsampling of redbook. This could well have been due to changes in jitter through the additional set of transmission lines. WIth the Tascam software, this issue is eliminated.
The ADC in the DA-3000 samples the signal identically whether you select DSD or PCM - the difference is the whether the bitstream is fed to an additional decimation plus HPF stage or not.
Although PDM/PEM/PWM DACs were introduced in the 90s to overcome the linearity issues with the multi-bit DACs, in theory, 1 bit coding is fundamentally flawed due to the inability to adequately dither the signal without saturating the modulator - a multi-bit modulator is needed to properly dither the signal and avoid saturation. Lipshitz published a paper with the AES (2001) providing the proof to his assertion.
Regards Anthony
"Beauty is Truth, Truth Beauty.." Keats
I generally feel the same way you do about the hi-rez formats, although my own general preference is for PCM. I believe I was early on the bandwagon with hi-rez, as I started around 2000 with both DVD-Audio and SACD (two separate players at the time). As time has gone on though, I've come to feel more and more that the MCh experience makes a huge difference in my own enjoyment (in addition to the various hi-rez formats), and I feel that I'll soon make the move to a Dolby Atmos capable system. And, like you, I still enjoy many plain old CD's and CD-resolution files.EDIT: Actually I thought I'd list a couple of my faves in each of the highest rez formats.
DSD256: Bach Cello Suites (Eijlander)
Rimsky-Korsakov et al Band Music (Netherlands Navy Marine Band)
Mahler Third Symphony (Fischer, BFO)DXD: Berlioz Symphonie funebre et triomphal, etc. (Staff Band of the Norwegian Armed Forces)
Mahler Seventh Symphony (Jansons, Concertgebouw)
Bruckner Ninth Symphony; Wagner Parsifal Excerpts (Gatti, Concertgebouw)(I guess this is pretty conventional "sonic blockbuster" repertoire, except for the Bach Cello Suites.)
Edits: 09/04/20
The DSD sampling rate is 64x that of redbook but the amplitude resolution is actually 1/32768 that of redbook. Multiply those together and DSD 512x worse than redbook - but at least it does so 'with astonishing regularity' :)
13DoW
ps. DSD overcomes that deficiency by using a lot of noise shaping and so it does end up somewhat similar to 24/96 PCM - as shown in the picture posted by Roseval.
Yes, I like DSD very much. I own a TASCAM DA-3000 DSD128 recorder and I now copy all my vinyl records in DSD128. I also own at least 40 or more DSD256 albums and they sound absolutely spectacular. I definitely like DSD.
Thanks!
John Elison
I'm afraid you most of all made a file size calculator.
Not to be mistaken for resolution.
IMHO resolution should be expressed in dynamic range and frequency range.
If you do, you will notice that DSD64 isn't such a big (4000!) improvement at all
The Well Tempered Computer
The number is four -- not four thousand. However, I think you're right about it representing the difference in file size instead of the difference in resolution.Thanks!
John Elison
Edits: 09/03/20
2,822,400 / 705,600 = 4.000-times Higher Resolution than RedbookFor me (Europe) 4000 is written as 4.000
LOL
The Well Tempered Computer
Edits: 09/03/20
> 2,822,400 / 705,600 = 4.000-times Higher Resolution than Redbook
>
> For me (Europe) 4000 is written as 4.000There's another valid method for using commas and periods when writing numbers. I use a period as the decimal separator and I use a comma for the thousands separator or digit group separator. This should have been apparent when you saw the number 2,822,400 written with commas to separate each three-digit group.
Best regards,
John Elison
Edits: 09/03/20
/
The lay listener should also not think of 'resolution' in this context as something qualitative, it is simply a representation of the dynamic range (though, as Roseval points out, you cannot compare PCM & DSD on bit rate because of the DSD noise shaping). And, as redbook PCM already has more dynamic range than anyone can reproduce is 'resolution' even meaningful in the quantitative sense?
Regards,
13DoW
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: