i for one prefer to get my music for free. if the quality is a smidge less, so what, all the more money for me to pour into my system to recoup that lost sound and further enhance movies... multichannel music does very little for me in the traditional format, live it is acceptable, but on the eagles dts hell freezes over disk, the bonus track "bridges" or something with bridges in its title is exceptional with each of them singing from a different speaker. other than that ill take my music loud and from the front.

Some basic and probably ignorant questions, but questions that I believe will help many people like myself understand digital signals in sound reproduction...

... Could someone describe the basic digital format attributes (or data structure) for sound encoding?

For a single audio track, assuming a typically wide frequency range (10Hz - 30kHz), with lots of instruments playing at the same time, how is this information stored digitally?

For example, if a typical CD is recorded in 44.1 kHz/16-bit, is there a complete picture of the audio spectrum being created 44,100 times a second? Does the does the bit depth describe amplitude only, frequency, both, or even more?

I believe that 44.1, 88.2, 96 & 192 kHz are sampling rates, but what are the attributes being sampled at this rate? Frequency; amplitude; both, or something else? Soundhound refers to megahertz sampling rates (1,000 kHz); how are these comparable? Are higher sampling rates useful for oversampling/error correction, to be used during the read/processing of the digital signal? It would seem reasonable that high sampling rates might be especially helpful in accurate description of high frequencies, but is it true?

... And I believe that a 24-bit recording can provide 256 x (times) the incremental information than a 16-bit signal... but what attributes are being stored here? ... And of what value are those extra 256 x (times) incremental steps? I understand that oversampling can be useful in error correction, is oversampling information contained in the extra bit depth? ...Does the greater bit depth provide just more information, or a way of verifying accuracy of the information, or both?

...And the attributes themselves... are discrete frequencies described digitally, or are there narrow frequency format slots (or bins) for frequencies throughout the audio spectrum, with each slot (bin) being so narrow that it closely approximates a smooth frequency/amplitude curve? I would guess that each frequency slot (bin) would have an amplitude assigned to it.

Or is the sampling rate high enough so that the bit depth just describes amplitude (or what else)?

Would a good analogy be a tape recording machine... where tape speed is comparable to sample rate, and tape density (dynamic range) is comparable to bit depth?

I would greatly appreciate a short primer, or a reference to a novice friendly source for this information.

Thank you...

Allan

Wow, that is a tall order that would be impossible to fill in any reasonable post-length! I would recommend the excellent book by Ken Pohlman (sp?) called Principles of Digital Audio. It covers it all.

There have been numerous primers in Stereophile, Sound & Vision and I'm sure others that cover the basics of digital audio. The "Home Theater secrets" website probably also has some good information. A Google search will undoubtedly find more information that you'd care to read.

Just be wary of technical descriptions from equipment manufacturers themeselves - these will of course be extremely biased to their own pet approach. Make sure you are getting information from a 3rd party that has no financial or marketing stake one way or the other.

Briefly, the sampling rate in PCM defines the high frequency limit of the system. The bit depth defines the maximum signal to noise ratio. 16 bit PCM has a theoritical limit of 96db: 24 bit has a limit of 144db (this is so quiet that it cannot even be realized on any practical electronic system, short of using liquid nitrogen cooling).

DSD encoding (used in SACD) is based on Delta-Sigma modulation which has been around for decades (I worked with this type of equipment in the late 1970s). The sampling rate issue is much more complex in this system, especially when more than one channel is being included in the same bitstream. The sampling rates used in DSD cannot be directly compared in effect to the sample rates used in PCM. The DSD technology is vastly different than linear PCM, however as I stated before, many PCM systems use 1 bit technology for A/D and D/A conversion - only the storage method is different.

Both DSD and PCM are truly excellent systems. Even "lowly" 16 bit/44.1K PCM used on CD is much better than most people realize - so good in fact that in a blind comparison with a 24 bit PCM (or DSD) master, it would be next to impossible to pick out the difference in a real world listening situation.

Keep in mind that the electronic noise from microphones, mixers etc (not to mention the acoustic ambient noise from the recording room and the musicians themselves) that cannot be avoided in any real-world recording situation will limit the noise floor to roughly -60db below full scale digital - at the most. This would be with a classical recording in an extremely quiet room. Most studio recordings of popular music have signal levels that do not get much lower than -30db or so below digital full scale.

If you look at the measurements of a good DVD or CD player in any review, you will see that the noise floor is equavalent to about 18-20 bits at the best.


[This message has been edited by soundhound (edited April 08, 2004).]

I think a good producer can make a recording sound good in any format. People like Hugh Pagdham or Doug Sax regularly produce excellent sounding recordings regardless of whether they are lps, cds or sacds. (I think Sax works with Allison Krauss if I'm not mistaken.) I'm surprised more people don't notice how variable the software is. For me, that's the greatest limitation on how good the sound/video is. BTW, I recommend "Rainbow Body" as a superb orchestral recording. The Copeland is sublime, IMHO!

Jay

There certainly are huge variances in the quality of recordings themeselves, and of the sonic choices the engineers and producers make in the tracking and mixing of music albums. I have numerous recordings where the lead vocals were recorded in different studios, and in some instances by different engineers, with the difference in timbre clearly evident from track to track.

Even albums recorded by the same engineer have variances from track to track because of instruments/vocals being reorded and/or mixed on different days over a period of weeks. The mastering phase is supposed to even out these variances, but it is impossible to correct them all, such as different sounding lead vocals when the rest of the ensemble was recorded seperately.

Many home systems are not set up ideally, with compromises made in speaker placement and room acoustics to acommodate room decor and furnishings. These compromises will make subtle differences in a recording more difficult to hear than in a situation where a room is set up for music listning only, with no placement or acoustic treatment restrictions.

After taking a day to go over some technical literatures found on Google search, I have come to the conclusion that I must have a very revealing system!!!!!

After reading up on some PCM, DSD, Delta Sigma modulation, Nyquist limit, decimating filters, high order error correction, Fourier transformation equations (I thought I would never see this after college) and much more theoretical stuff, I dicided to do a little test.

I originally stated that I heard a noticable improvement on the sound of SACDs that are supposed to have been mastered with DSD technology through out the recording process over some remastered SACD's from an earlier master (tape?). Furthermore, the improvement in sound was noticable even when these hybrid discs were played on a CD player. From this I concluded that the improvement must be from the production side (ADC and digital manipulation).

I also read that 24 bit recording sampled at 48KHz with current ADC know-how will produce "almost" perfect reproduction of the anolog sound waves.

From this I have 2 possibilities.
1. DSD ADC process improves on the "almost" perfect reproduction of 24 bit, 48KHz PCM process.
2. The engineers on the Kraus and Taylor were very good and produced better sounds from the digitized sound waves of same quality.

To prove whether the improvement is due to the DSD upsampling or post ADConversion process, I listened to James Taylor's Best CD, a CD of "Oh where art thou," one track of which is sung by Alison Kraus, along with the afrementioned SACD's on four different CD players: Sony XA 777es, Adcom GDC 700, Sony 300 cd changer and a Panasonic DVD Q50.

Here is what I found. SACD's played as CD clearly sounded better then the CD's When played through XA 777es and GDC 700, both of which are part of my 2 channel system. Through Sony cd changer and the DVD player connected to my home theater setup, there were no discernable difference.

Conclusion:

DSD up-sampling may or may not be responsible for the sound improvement although I think it may prove to be so with further testing. The improvement is only discernable if the listening equipment meets the revealing threshold. Furthemore, from the minute difference of quality between XA777es and GDC 700, I would say the quality of the software is more important than the quality of the DAC's in one's system.

Aside from the CD players the following equipments were used for this test.

Home theater
950/770, Polk SDA II(front), Aperion 5.1 with 12" sub (the rest for 7.1), Outlaw ICBM, outlaw interconnects, 14 guage copper speaker wire (Home Depot) terminated with gold plated banana plugs

2 Channel
Audio Research SP9 MkII preamp, Audio Research VT100 MKII amp for mids and highs, Adcom GFA 555 MKII amp modified with bigger and more power supply for bass, VMPS RM 40 speakers, DYI attenuator between pre and amp for mids, Outlaw interconnects, DH labs Q10 speaker cables.

The only way you are going to "prove" anything is to take a live microphone feed using the identical front end (recoding electronics) and encode one using DSD and another using PCM at whatever sampling rate and bit depths you choose, and do a live blind comparison. I have participated in such comparisons - once with a live orchestra as the reference. Barring this, you are just hearing differences in the artistic choices made by the engineer, the sonic differences of transfers of what could easily be different generations (backups) of the master tape, differences in the outboard equipment used for the transfer, and of course differences in the sound of your SACD player verses your DVD-A player.

Comparing two different recordings of different music, or even of the same recording but different transfers made at different times (especially when one is multi-channel and one is stereo) is meaningless. Beyond the fact that you may prefer one version or the other, it says nothing about the capability of the underlying technology.

I cannot stress this fact enough - PCM uses the same conversion technology (A/D & D/A) as DSD in many cases. Many, many SACDs are originally recorded using PCM. This is done because it is very difficult to perform any digital signal processing on the DSD bitstream, and there are other practical considerations that enter into the picture (project portability being one of them). The SACD you are listening to, especially if it is of popular music, may very well have had it's master recorded in PCM and then later converted to DSD for release on SACD. Of course they are not going to tell you this fact on the album sleeve. Classical releases are more likely to have been natively recorded in DSD because there is relatively less post-production signal processing involved.

The lineage of a recording from studio to finished disk can be very complex, and not as "pure" as you might think. There are many steps of production and post-production along the way, and many people have influence on the finished product. Most of these steps are unknown to the final purchaser. Choices are made for economic as well as marketing reasons. Things are not so black and white as they may seem to be, and as manufacturers may wish you to believe. This is true of any field, not just the music recording industry.

I guarantee that a good engineer could produce an equally outstanding recording regardless of whether the medium was PCM, DSD, or analog tape.

If there were "vast" differences in the sound of the actual encoding processes (PCM and DSD), don't you think that everybody in the professional recording industry would migrate to the one that "obviously" sounds better and abandon the one that sounds inferior? Don't you think that any manufacturer of professional recording equipment would ensure that what goes into their recording equipment sounds identical to what comes out? If this were not the case, that manufacturer would not be selling equipment to professional users for long. In a professional situation, the recording engineers have the constant ability to directly hear the live musicians, the live musicians through a microphone feed into the control room, and the result of recording that live microphone feed on any type of equipment (PCM, DSD, analog) of their choice. In a studio situation, I have never heard the "vast" differences in the sound of the two encoding processes that seem to be so "obvious" on consumer's homes. The only obvious differences I have heard is when comparing any digital format to analog tape, and the analog tape is obviously colored in comparison to a live microphone feed, more or less so depending on how heavily the tape is modulated.

[This message has been edited by soundhound (edited April 10, 2004).]