|
Post by jmilton on Jun 12, 2014 14:46:10 GMT -5
Just read this and thought it worth passing on:
MQ-P – From a PCM master source 48 kHz/20 bit or higher; (typically 96/24 or 192/24 content) MQ-A – From an analog master source MQ-C – From a CD master source (44.1 kHz/16 bit content) MQ-D – From a DSD/DSF master source (typically 2.8 or 5.6 MHz content)
jm
|
|
DYohn
Emo VIPs
Posts: 18,357
|
Post by DYohn on Jun 12, 2014 16:27:57 GMT -5
DSD. Bah. Re-sampled and synthesized content. It's the new flavor of the day but if it's accuracy you seek it sucks.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jun 12, 2014 22:57:37 GMT -5
|
|
KeithL
Administrator
Posts: 9,961
|
Post by KeithL on Jun 13, 2014 11:04:19 GMT -5
I think this is a very well-balanced article - and well worth the read for anyone who is seriously "excited" about DSD. However, I would even go one step further with one thing he said..... The output of a DAC isn't PCM OR DSD: it's ANALOG AUDIO. Worrying about when and if the signal is, or gets converted to, DSD, or PCM, or 6 bits, or 1 bit, or 96 bits, is really pointless - unless you're designing DACs. Here's a philosophy question for you: You're at a museum; enjoying looking at your favorite painting, marveling at how well the painter handled light and object placement; this is a really excellent painting! As you're lost in reverie, a guard enters and informs you that the exhibit is being closed - because they've just discovered that the painting is a forgery. Does the painting somehow magically become less enjoyable? Is it now NOT a good painting any more? Have the colors changed? Is the perspective suddenly off? And, if you're one of a small but vocal group of people who are quite convinced that PCM does something dreadful to sound quality, while DSD is just perfect, will your favorite albums all become nasty digital-sounding crap when you find out that the majority of them were converted to the dreaded PCM somewhere along the way - then converted back to DSD? I certainly hope not. But I also hope that you might start wondering if they might sound even better if they'd suffered through one less conversion - and remained as PCM..... hmmmm..... As far as I know, technically, DSD is reasonably practical as a format for recording digital audio (it may even be slightly superior to PCM, all else equal, for that), or for converting analog source material to digital; DSD is currently somewhat inconvenient for delivering audio to the consumer (because devices that actually play DSD are still in a minority); and DSD is hugely impractical for mixing and mastering anything that requires any processing beyond minor level alterations and fades (and it's even a nuisance for that). In other words, DSD works OK for the equivalent of "direct to disc" recordings, and is a distinct nuisance for any that require more complicated "engineering". Converting anything to DSD for delivery - as a download - is pretty simple (just grab a free copy of Audiogate or spring for a - not nearly free - copy of Saracon), but playing it is going to limit your options and make setup a bit more complicated, and mastering physical SACDs is very complicated (because special writing equipment is required). As I've mentioned before, EVERY conversion between DSD and PCM will quite possibly produce some audible change in sound - because the conversion process is not "bit perfect" - there are digital filters and other such things involved. This doesn't specifically suggest that either is better, but it does make it nearly impossible to "directly compare" the two. Whichever way you convert, your converted file will sound different than the original. (And, if you were to convert the same analog source into both, then the audible differences between the two A-to-D converters you used to do the conversion, and in the D-to-A converters you used to play them, would probably be bigger than any possible differences in the results. So, again, you wouldn't be able to compare "two exactly equal files - one of each type".)
|
|
DYohn
Emo VIPs
Posts: 18,357
|
Post by DYohn on Jun 13, 2014 11:21:39 GMT -5
My problem with DSD - besides the fact that it is simply the audiophile flavor of the day - is the same problem I have with all sigma-Delta processes. It does not simply sample the original signal to create PCM, it uses a mathematical algorithm to derive an approximation of the original signal. Can it sound good? Of course it can. But is it as close as digitally possible to the analog original? No, it's not. There ARE bit-perfect ADC/DAC technologies available, it's just that they are difficult to implement and not cheap.
|
|
KeithL
Administrator
Posts: 9,961
|
Post by KeithL on Jun 13, 2014 13:58:25 GMT -5
I don't know how to tell you this, but ALL conversion to digital fits that description - "using a mathematical algorithm to derive an approximation of the original signal". (Of course, all analog audio could be equally accurately described as "producing an approximation of the original signal by analog means" - since nothing analog is "perfect" either). However, the goal of both, which you can achieve pretty well, is to achieve an approximation that is good enough that it sounds perfect to us mere humans. In general, both before you do the A-D on the analog original, and after you convert back to analog with the D-A, you need to include a filter that allows up to the required passband, yet totally cuts off everything just a few percent above that. (For CD audio, it needs to pass up to 20 kHz flat, but be down at least 70 dB or 80 dB at 22.5 kHz.) This is something you simply cannot do effectively with an analog filter - at any price. There are several "non-oversampling DACs" out there, some quite expensive, and all have horrible performance (although some do sound quite pleasant). When you compare a Delta-Sigma DAC to its NOS equivalent, the performance isn't "almost as good for a lot less"; rather it is "as good or better" for "a LOT less" - unless you prefer to put up with a frequency response that is -3 dB at 20 kHz in return for avoiding a tiny bit of ringing. Oversampling entirely eliminates all the nasty trade-offs and issues with filtering. In return, you end up with some errors in transient response (due to ringing and overshoot in the digital filters used to produce the interpolated samples). However, since the side effects of super-sharp analog filters are unavoidable, as are the problems with omitting filters entirely, it's a choice between compromises. I am simply not aware of any DAC or ADC technology, at any price, that avoids all the compromises. The best way to mitigate the problems with both is to use a higher sampling rate (96k is quite good). Delta-sigma is no worse (or better) than several other methods, but is the cheapest way to get to the same end, which is why it is so widely used. Incidentally, even if you had a digital signal which was "the most accurate approximation to the original possible", that STILL wouldn't give you a walk on issues like the digital reconstruction filter tradeoff - because you still have to convert it back to analog to listen to it. (And, yeah, even if you had a powered speaker with a digital input, you'd still have to convert it somewhere - because air pressure is still an analog phenomenon - so it's gotta get back to that somewhere.) The only specific "indictment" I've ever heard of Delta-Sigma, as a technology, is that it is sensitive to signal jitter (having jitter in your clocks at some points in the process will produce more distortion in a Delta-Sigma converter than in some other types), so using a bad quality clock will degrade the performance of a Delta-Sigma DAC more than some other types. But that can be mitigated by using good clocks at the appropriate stages in the process. I am curious what "bit perfect ADC and DAC technologies" you're referring to... (Honestly, every claim I've heard to that effect ended up being a slightly exaggerated - and somewhat self serving - bit of advertising propaganda about how one manufacturer claims to have "eliminated all the important tradeoffs" and traded them for others that are "totally insignificant" - once you read the fine print. Apparently this always necessitates using some wonderful new technology which, because they had to invent it themselves, justifies the astronomical prices of their products. Oddly, when tested, these products routinely have audio performance that is equal to, or even worse than, plain old cheap Delta-Sigma equivalents.) My problem with DSD - besides the fact that it is simply the audiophile flavor of the day - is the same problem I have with all sigma-Delta processes. It does not simply sample the original signal to create PCM, it uses a mathematical algorithm to derive an approximation of the original signal. Can it sound good? Of course it can. But is it as close as digitally possible to the analog original? No, it's not. There ARE bit-perfect ADC/DAC technologies available, it's just that they are difficult to implement and not cheap.
|
|
|
Post by Topend on Jun 13, 2014 15:56:41 GMT -5
Does recording in a higher bitrate and sample rate reduce the amount of approximation?
Dave.
|
|
|
Post by rogersch on Jun 13, 2014 16:54:05 GMT -5
Does recording in a higher bitrate and sample rate reduce the amount of approximation? Dave. Yes
|
|
DYohn
Emo VIPs
Posts: 18,357
|
Post by DYohn on Jun 13, 2014 17:49:17 GMT -5
Keith: well, an R-2R resister ladder network for one, a topology that is becoming economically feasible with SMD technology.
|
|
|
Post by Topend on Jun 13, 2014 19:11:41 GMT -5
Does recording in a higher bitrate and sample rate reduce the amount of approximation? Dave. Yes So HD Tracks is on to something.
|
|
DYohn
Emo VIPs
Posts: 18,357
|
Post by DYohn on Jun 13, 2014 20:12:18 GMT -5
So HD Tracks is on to something. Not if they just up-sample from a lower bit rate original. Then you gain precisely nothing. The digital original has to be high resolution.
|
|
|
Post by Topend on Jun 13, 2014 20:41:24 GMT -5
So HD Tracks is on to something. Not if they just up-sample from a lower bit rate original. Then you gain precisely nothing. The digital original has to be high resolution. Don't they get the master copy and reproduce it. It would depend on the master copy which could be either digital or analog. Dave.
|
|
|
Post by rogersch on Jun 13, 2014 23:22:29 GMT -5
Not if they just up-sample from a lower bit rate original. Then you gain precisely nothing. The digital original has to be high resolution. Don't they get the master copy and reproduce it. It would depend on the master copy which could be either digital or analog. Dave. I've read several items on different Hi-res audio forums where they proved that some HD audio tracks from the early days of "HDTracks.com" were just upsampled CD tracks.
|
|
|
Post by sharky on Jun 15, 2014 15:17:22 GMT -5
Not if they just up-sample from a lower bit rate original. Then you gain precisely nothing. The digital original has to be high resolution. Don't they get the master copy and reproduce it. It would depend on the master copy which could be either digital or analog. Dave. No. HDTracks is a shop, just like Amazon. They sell whatever the labels send them to sell. They only do their own files for Chesky label albums (Chesky owns HDTracks). The files you buy at HDtracks are the same ones that you buy at Qobuz or other HD music store. If the label just upsampled the existing 16bit/44.1KHz file then you gain nothing. A few cases of that have already been spotted, and obviously affect all stores, HDTracks included.
|
|