Surround sound, SACDs and DVD Audio

[Dave2002]

A recent interchange on another thread suggests some interest in SACDs and DVD Audio. http://www.for3.org/forums/showthrea...233#post167233

There may also be interest in surround sound.

Perhaps anyone who wants to discuss these topics could do so here.

I have been interested in surround sound for many years, and indeed helped Michael Gerzon to set up some microphones in an early experiment.
I believe that surround sound can work, but I don't know whether many surround sound recordings "do it properly". Some purists (I might tend to favour these) might simply
put up a single surround sound microphone (such as an Ambisonic soundfield microphone - see http://homepage.ntlworld.com/henry01...soundfield.htm for one example) and just extract the relevant number of channels to feed to appropriate loudspeakers. Others might use very many spot microphones, and somehow fabricate the recorded tracks into some form of whole. This may or may not give a cohesive feel to the resulting recording when played back.

Of course SACDs and DVD Audio are not only about surround sound. They may also support 2 channel sound - though hopefully in better quality than CD quality, and in some cases 3 channel sound - with a centre channel. Some of the Mercury issues (for example in the recent Mercury box) have previously been issued in 3 channel formats - I think I managed to acquire a few, and now wish I'd bought more.

The most recent SACDs and DVD Audio discs should have better audio quality, with more resolution (typically up to 24 bits) and a higher sampling frequency. Depending on various factors these may sound better than CDs, which are limited in bit depth (16 bits) and sampling frequency.

Perhaps I should also mention Blu Ray discs, as these should also be capable of delivering high quality audio, and can also do surround sound. Some enthusiasts for surround sound may be aware that sometimes surround sound makes compromises, and may use data compression. Often this may not matter (arguably) as the material is often the sound track for a film or may represent the sound effects for computer games, and there seems to be a feeling that even if there is compression, the results are better than using fewer channels. If the output of a DVD player or Blu Ray player is taken from the optical output (SPDIF) this is normally either reduced to 2 channels, or encoded using a multi-channel compressed format. However, more recent players may use the latest HDMI interfaces, and may be capable of driving multiple loudspeaker channels using uncompressed audio.

[ferneyhoughgeliebte]

Many Thanks for this, Dave.

[Gordon]

I have been interested in surround sound for many years, and indeed helped Michael Gerzon to set up some microphones in an early experiment.
I believe that surround sound can work, but I don't know whether many surround sound recordings "do it properly". Some purists (I might tend to favour these) might simply put up a single surround sound microphone (such as an Ambisonic soundfield microphone - see http://homepage.ntlworld.com/henry01...soundfield.htm for one example) and just extract the relevant number of channels to feed to appropriate loudspeakers. Others might use very many spot microphones, and somehow fabricate the recorded tracks into some form of whole. This may or may not give a cohesive feel to the resulting recording when played back.

I fear you are right on that one Dave. Many moons ago our lab did some work on surround with a view to finding a broadcastable format, sadly that never materialised. However the soundfield approach was theoretically very promising as several papers in that era showed [late70s early 80s]. A colleague, Jonathan Halliday, who later went on to Nimbus, was quite enthusiastic about this approach. However like other audio production techniques the practical value of multimics [“fix it in post”] and mixing the result sort of undermined it.

Of course SACDs and DVD Audio are not only about surround sound. They may also support 2 channel sound - though hopefully in better quality than CD quality, and in some cases 3 channel sound - with a centre channel. Some of the Mercury issues (for example in the recent Mercury box) have previously been issued in 3 channel formats - I think I managed to acquire a few, and now wish I'd bought more.

The most recent SACDs and DVD Audio discs should have better audio quality, with more resolution (typically up to 24 bits) and a higher sampling frequency. Depending on various factors these may sound better than CDs, which are limited in bit depth (16 bits) and sampling frequency.

When you have original tapes which are multitrack like those Mercurys it makes sense to try and preserve their original idea of what the sound was like in the venue. So by all means mix down to stereo for the CD level but put also the 3 channels in the DSD without any additional fiddling.

But consider the S/N of those early tape and film recorders that Mercury [and everyone else in the late 50s] used. Lucky if you get to 70dB weighted and those days no Dolby A until the mid 60s. You only need about 10 bits for that, the eleventh and subsequent bits are carrying the tape hiss in which a lot of audio detail like reverberation etc is buried. Later things did improve but multitrack tended to consume the gains of progress in tape machine design. The S/N of a tape track is inversely related to the track width so the more tracks on a given tape width the worse the S/N which explains why some of those early multimic’d and multitracked audio recordings of the early 60s were so noisy. Hence the move to 2inch tape width which had been used in the first video tape format in the late 50s. Remastering these old tapes needs lot of noise reduction.

I am not convinced about large bit depth in a consumer format. The 24 bits of the AES isn’t necessarily there for sound quality as such, it’s there for headroom in production too where erosion occurs through mixers and also to avoid clipping. Backing off 6dB for headroom loses a bit straight away and if a mixer fader is 12dB down on a given mic channel then it has also lost 2 more bits. Round off in the final adder could well lose another. Once the final recording is released you don’t necessarily need those 24 bits any more. The reason you need 24 bits is to keep the quantising step small.

Just think what dynamic range is implied by 24 bits [theoretically over 150dB unweighted: 20 x n x log2 + 10 x log12] and also the implications for accuracy in ADC and DACs not to mention amplifiers whose front ends make noise [look at the S/N spec of your amp which is usually rated against full output power]. I know from measuring the performance of ADCs many do not meet their claimed bit depth. A file syntax may contain places for 24 bits but the MSBs are often carrying noise or nothing at all. Oversampling is beneficial if only because it simplifies filter design and manufacture and also makes room to do noise shaping when reducing to CD.

Perhaps I should also mention Blu Ray discs, as these should also be capable of delivering high quality audio, and can also do surround sound. Some enthusiasts for surround sound may be aware that sometimes surround sound makes compromises, and may use data compression. Often this may not matter (arguably) as the material is often the sound track for a film or may represent the sound effects for computer games, and there seems to be a feeling that even if there is compression, the results are better than using fewer channels. If the output of a DVD player or Blu Ray player is taken from the optical output (SPDIF) this is normally either reduced to 2 channels, or encoded using a multi-channel compressed format. However, more recent players may use the latest HDMI interfaces, and may be capable of driving multiple loudspeaker channels using uncompressed audio.

The Dolby 5.1, 7.1 etc and similar MPEG audio compression systems may well compromise surround. But some of those formats were designed for cinema not for HiFi enthusiasts. As you say, the sound tracks of films offer the spatial experience that better sound in fewer channels might not and come into their own in blockbuster disaster movies. Where bit rate is not an issue of course one does not need compression but unfortunately bandwidth is always a factor.
Currently the world broadcasters and the cinema industry are looking ahead to very high definition TV using 4000 pixels by 2000 as a format ie 8 megapixels. At present full resolution HDTV is 1920/1080 [2 Mpixels] but there is also a standard at 1280 x 720 so HD is a bit variable.

This is pitifull compared to the average DSLR stills cameras which are offering 14 Mpixels at low cost in pocket “point and shoot” cameras and well over 20 in professional devices like Canon and Nikon. Flat panel displays to support the 4Mpixel level of resolution are not that far away, perhaps only a few more years before they become affordable. To go with this the audio systems might be 22.2 for example ie 22 channels of full bandwidth audio and 2 “earthquake” channels for the deep bass. Try and get that on a Blu Ray without using compression!! Needless to say much higher density carrier media are being developed. But where does one put those 24 speakers?

[Dave2002]

Gordon

Thanks for bringing us somewhat down to earth. You are of course right about 24 bits seeming like overkill. Having said that I'm sure I've heard recordings which for some reason sounded better when played through DACs with more than 16 bits resolution. I don't even mind particularly if a recording company/recording engineer wants to use floating point or even multiple length arithmetic, which many modern processors are now capable of, but I'll concede that most of the extra accuracy probably won't do much other than provide different noise but at very low levels. I have a hunch that some material which has been lossily compressed at lowish bit rates (e.g MP3s at 128kbps rate) may sound better if more bits are used in the calculations to recover the wanted signal and the output used to drive a DAC with higher resolution. I'm not arguing for low bit rate lossy compression, but merely suggesting that decoding can go better with more resolution used on playback.

Re surround sound and films, I have watched films recently which have been unbearable using surround sound because the bass levels have been far too high. Just because a sound system can do it doesn't mean the film makers should have exaggerated bass. The War of the Worlds was a movie which really showed this problem, and I gave up and watched it using the 2 channel TV speakers.

[Gordon]

.... I'm sure I've heard recordings which for some reason sounded better when played through DACs with more than 16 bits resolution.

I've not had that experience with DACs. What are you getting from the extra bits? There is no more information in an 18 bit version of an audio sample than in the original 16. The additional 2 bits are computed somehow. When you say “sounds better” what does that actually mean? There is an automatic assumption there somewhere that the “improvement” must be due to extra bits. What if it is just a better 16 bit DAC because the poorer sounding one isn’t actually getting up to 16 bit performance in the first place? A good DAC with more bits might be more linear than a poor 16 bit one but it would have to be engineered appropriately. There is no free lunch.

A similar argument exists for oversampling in DACs. There is no more bandwidth available in an upsampled version of a 44.1/16 CD file because if there was more it was filtered out at the encoder. All an upsampler does is compute interpolated samples from the samples it has been given; it can’t create real bandwidth let alone magically recover the harmonics that the coder filter removed. A smidgeon of distortion however - from a slightly non linear DAC say - and you get a few harmonics that give an impression of a brighter sound. Engineer that non linearity carefully and you could make a DAC sound like a valve amp!!

A very clever fast processor in an upsampling DAC could do spectral analysis and infer dynamically what those missing harmonics might be and compute an approximation of them but that is not what a normal upsampler does. However the HE AAC version of DAB+ does do that analysis [it does it as part of the compression algorithm] using a technique called Spectral Band Replication [see Wiki] and so the decoder can add back some inferred high frequencies that were never sent.

With all this processing going on you wonder what it is you are listening to.

I have a hunch that some material which has been lossily compressed at lowish bit rates (e.g MP3s at 128kbps rate) may sound better if more bits are used in the calculations to recover the wanted signal and the output used to drive a DAC with higher resolution. I'm not arguing for low bit rate lossy compression, but merely suggesting that decoding can go better with more resolution used on playback.

I'd agree with that. I'd guess that most people think an MP3 player is an MP3 player. If the designer is trying to save money by using the slowest processor he can then he will also have to make compromises in the arithmetic. That may well mean rounding off calculations less than adequately with the result that the audio is affected. How does any consumer get to know this? Similarly an MP3 encoder can be poorly implemented so not all MP3 files are created equal either. A poor coder may require more bits/second to give the same audio quality as a good coder [remember the old DAB MP2 coders that the BBC replaced some while ago?]. The value of compression is obvious and most consumers, who are not HiFi enthusiasts, will accept the benefits just as they accept CD as adequate for their purposes. MP3 like CD never was specifically designed as a HiFi medium.

If you must compress then FLAC [or WinZip etc etc, all based on LZW or Huffman techniques] and the like give protection against losses but the degree of compression is not as much and so file sizes, particularly when you start at say 192/24, are larger. If you look at the thread about a free version of the Goldbergs the source site offers an MP3 version coded at around 500kb/s amounting to 133 Mbytes whereas the FLAC file is 1.4Gbytes. We'd all rather have no compression but economics etc sometimes demands it. Neither DAB nor Freeview would offer listeners and viewers so much choice without it. Whilst broadband speeds are increasing steadily the source coding standards are driving up the starting bit rates which probably means that lossy compression will always be with us at least for live streaming.

[PJPJ]

.......I am not convinced about large bit depth in a consumer format. The 24 bits of the AES isn’t necessarily there for sound quality as such, it’s there for headroom in production too where erosion occurs through mixers and also to avoid clipping. Backing off 6dB for headroom loses a bit straight away and if a mixer fader is 12dB down on a given mic channel then it has also lost 2 more bits. Round off in the final adder could well lose another. Once the final recording is released you don’t necessarily need those 24 bits any more. The reason you need 24 bits is to keep the quantising step small.......

I don't doubt the science, but streaming a piece of orchestral music [BIS Kalinnikov] using 16/44.1 and 24/44.1 (the lowest of high resolution quality) produces for me an easily heard improvement in sound-stage focus. A good sound becomes an excellent one.

As for surround, my latest purchase is:

http://www.mdt.co.uk/MDTSite/product//MDG9031598.htm

24 bit PCM (I don't believe pure DSD recording produces a better result by default) and the stereo sound is of a quality far superior to CD, the multichannel producing that focus of the orchestra and soloist and sound-stage depth which makes listening to these superb performances a pleasure.

The Aho piece is for cello and oboe - excellent definition, no blurring, no flat perspective, all in comparison to the CD layer, of course.

[Ariosto]

I agree about 24 bit sound, which of course is excellent when recording because of the additional headroom and lower noise floor. I'm actually talking of my own live recordings where the mics go into a high quality pre-amp as well. The down sizing to 16 bit after editing means that the CD sounds extremly good without the addition of any unecessary processing.

[Gordon]

I don't doubt the science, but streaming a piece of orchestral music [BIS Kalinnikov] using 16/44.1 and 24/44.1 (the lowest of high resolution quality) produces for me an easily heard improvement in sound-stage focus. A good sound becomes an excellent one.

If you have the means to decode and then render a full 24 bit file then of course it is better than 16/44.1. Do we know whether each version has been produced from a 24/96 original and carefully noise shaped? If so a carefully prepared 16 bit file should perform at better than 16 bits.

I'm not sure what "sound stage focus" is. If it means clear and natural placing of instruments in an acoustic then I understand. Such a feature is usually attributable to oversampling at source [thus avoiding those anti alias filters with sharp cut off] and good clean clocking of the DAC but in this case the clocks are the same. I have heard high resolution files where the "sound stage" makes the instruments so clear they seem to be in different places altogether. I believe that that was down to multimics and tricks in the mix. BIS are reknowned for their minimalist production methods which would mean that the mic set up would be simple and would lead to a much more natural sound.

As for surround, my latest purchase is:

http://www.mdt.co.uk/MDTSite/product//MDG9031598.htm

24 bit PCM (I don't believe pure DSD recording produces a better result by default) and the stereo sound is of a quality far superior to CD, the multichannel producing that focus of the orchestra and soloist and sound-stage depth which makes listening to these superb performances a pleasure.

The Aho piece is for cello and oboe - excellent definition, no blurring, no flat perspective, all in comparison to the CD layer, of course.

I'm not sure what you are saying here. Is your first clause a general statement or one about the MDG disc? DSD isn't 24 bit PCM is it? Where have you got the 24 bit PCM version of a CD from? Are you using the 24 bit DAC you mention above to upsample the CD?

[PJPJ]

If you have the means to decode and then render a full 24 bit file then of course it is better than 16/44.1. Do we know whether each version has been produced from a 24/96 original and carefully noise shaped? If so a carefully prepared 16 bit file should perform at better than 16 bits.

The original is 24/44.1

I'm not sure what you are saying here. Is your first clause a general statement or one about the MDG disc? DSD isn't 24 bit PCM is it? Where have you got the 24 bit PCM version of a CD from? Are you using the 24 bit DAC you mention above to upsample the CD?

No, the MDG disc is an SACD, recorded in 24 bit PCM, then converted to DSD for SACD reproduction. (There isn't a 24 bit CD.) The point I failed to make was the stereo SACD programme sounds to me noticeably superior to the CD programme on the same disc.

The sentence should have read:

As for surround, my latest purchase is SACD MDG9031598 Strauss Oboe Concerto et al, 24 bit PCM (I don't believe pure DSD recording produces a better result by default), and the stereo sound is of a quality far superior to CD, the multichannel producing that focus of the orchestra and soloist and sound-stage depth which makes listening to these superb performances a pleasure.

[Gordon]

No, the MDG disc is an SACD, recorded in 24 bit PCM, then converted to DSD for SACD reproduction. (There isn't a 24 bit CD.) The point I failed to make was the stereo SACD programme sounds to me noticeably superior to the CD programme on the same disc.

OK thanks for the clarification. I'd expect that the hybrid DSD layer working in 2 channels would preserve more of the original 96/24 than the CD layer. This is valid comparison because you have the same stereo source material offered in two formats reproduced through much of the same equipment.

I assume that your CD player uses an oversampling DAC that will avoid filtering issues so what you are hearing is almost all due to bit depth. With modern noise shaping methods which can be surprisingly good I am a bit surprised at how much difference you hear. One of the easiest ways to muddy up sound in ordinary CD replay is by having poor DAC clocking. I'm not sure how DSD clock stability affects the decode but the same principle applies - jitter displaces correct amplitude reconstruction - it may be that DSD is less susceptible. Of all the technical papers on the subject I don't remember one looking at comparing clocking of CD and DSD.

Whatever these technical musings might mean your ears are telling you that the DSD is better than the CD. SACD playback is worth the extra I suppose because it frees one from the limitations of standard CD but SACD has not really taken off that well probably because most people are not that interested in HiFi sound quality. My collection of about 3,500 CDs plus innumerable CD-Rs is almost entirely standard CD. There can't be more than about 20-30 SACDS in there so changing my CD player just for those doesn't make sense. I stopped collecting avidly some while ago and download 24 FLACs nowadays. Perhaps I need to invest in a file player.

[PJPJ]

I use a Linn Unidisk SC for SACD/CD playback and a Musical Fidelity CLiC for stereo files. Sound comparisons using the CLiC introduce no variables - I'm not sure whether the Unidisk makes as good a job of playing CDs as it does with SACDs, i.e. CDs don't sound as good as they might using this player rather than another. The CLiC doesn't give an unfair advantage to one or other of 16 and 24 bit, i.e. the 24 bit versions sound better because they are better, and not because the CLiC does a better job with them.

The CLiC also allows one to dispense with all those CDRs.

I discovered a 24/96 recording of the Goldberg Vars this morning - I've not had a chance to listen yet so can't comment on either the performance or the recording quality. You said "SACD has not really taken off that well probably because most people are not that interested in HiFi sound quality." I have to agree. This discussion, for example, has been banished to the outer limits possibly in case it frightens the horses on a music thread.

As for high resolution downloads, the necessity for some sort of storage device also seems to put people off. There's a limit for how many hours' worth of 24 bit recordings one can keep on one's computer hard drive, the problem of transferring them when a new computer is bought and so on..... The Bach Goldberg Vars download is nearly 1.4 GB.

http://www.opengoldbergvariations.org/

[Dave2002]

I've not had that experience with DACs. What are you getting from the extra bits? There is no more information in an 18 bit version of an audio sample than in the original 16. The additional 2 bits are computed somehow. When you say “sounds better” what does that actually mean? There is an automatic assumption there somewhere that the “improvement” must be due to extra bits. What if it is just a better 16 bit DAC because the poorer sounding one isn’t actually getting up to 16 bit performance in the first place? A good DAC with more bits might be more linear than a poor 16 bit one but it would have to be engineered appropriately. There is no free lunch.

Gordon

I may have misled you slightly, though I believe the effects are perhaps similar. You may remember the BBC downloads of Beethoven symphonies. They were done at 128 kbps MP3.
I downloaded them and burnt them to CD. I was disappointed with the SQ at the time.

Rather later I listened directly to the MP3s and found these to be better. At that point I realised that MP3s can in fact generate more than 16 bits as output. There are of course too many variables in this anecdote to be conclusive evidence, but my overall feeling was that having more than 16 bits available can give better results in some circumstances.

To counter this you might also come back and suggest that my mapping from MP3 to CD wasn't done well, wasn't done with "correct" dithering, and that the DACs used were different etc., but for me it seemed to make the difference between recordings which were poor and ones which were at least acceptable. The DAC on my CD player is, I believe very good, or at least at the high end of what most of us would consider affordable.

Also, this is just one example - not sure if anything else is more than just hunches.

[PJPJ]

You don't say how you listened to the mp3s.

Those BBC downloads were of dire sound quality whether written to CD or not, certainly far removed from that of a decent CD let alone a better-than-CD quality SACD or other 24 bit recording.

[Gordon]

For PJPJ:

I use a Linn Unidisk SC for SACD/CD playback and a Musical Fidelity CLiC for stereo files. Sound comparisons using the CLiC introduce no variables - I'm not sure whether the Unidisk makes as good a job of playing CDs as it does with SACDs, i.e. CDs don't sound as good as they might using this player rather than another. The CLiC doesn't give an unfair advantage to one or other of 16 and 24 bit, i.e. the 24 bit versions sound better because they are better, and not because the CLiC does a better job with them.

The CLiC also allows one to dispense with all those CDRs.

Those items have good pedigree so there is no reason to doubt much their performance. I am interested by the MF CliC device because I’ve been thinking about such a thing for a while, it seems to do all the right things. It’s a bit pricy but I suppose nothing that is well engineered is cheap. The reviews seem very positive except for the WiFi sensitivity.

I discovered a 24/96 recording of the Goldberg Vars this morning - I've not had a chance to listen yet so can't comment on either the performance or the recording quality……

As for high resolution downloads, the necessity for some sort of storage device also seems to put people off. There's a limit for how many hours' worth of 24 bit recordings one can keep on one's computer hard drive, the problem of transferring them when a new computer is bought and so on..... The Bach Goldberg Vars download is nearly 1.4 GB.

I’ve seen that other thread [see Platform3] and have left a comment there. I did not have patience to download the full 1.4Gbytes FLAC. The MP3 is about 133Mbytes and is coded at 500kBit/s or more but does have its limitations. Sounds OK but there are aspects of the performance that are not up there with the best so I’ll not lose sleep over the technical quality in this case.

For DAVE2002:
Yes those Beethoven MP3s were a bit dire at 128 although the MP3 spec is quite clear in stating that it thinks that 128 MP3 is equivalent to CD.

As regards the conversion to CD you will have a decoder from MP3 to WAV first and then a burn to CD which will limit those files to 16 bits and garbage in garbage out applies. I doubt that the decoder will attempt to recover more than 16 bit depth and will limit its internal number crunching to that. I don’t think that MP3 coders have an option to code from 24/96 [it can do 16/48] and from memory the spec assumes that CD standard WAV is an interface of choice.

The conversion to WAV if done in your computer almost certainly was to 16 bit WAV unless you have the means to go to a later version. Even then when you burn to CD that WAV will have to be truncated and at best may have some noise shaping. With computer programs who knows what they do?

[PJPJ]

.......I am interested by the MF CliC device because I’ve been thinking about such a thing for a while, it seems to do all the right things. It’s a bit pricy but I suppose nothing that is well engineered is cheap. The reviews seem very positive except for the WiFi sensitivity......

My router is downstairs, the CLiC upstairs and it functions well wirelessly but it's perhaps expecting a lot to get perfect 24 bit streaming wirelessly over such a distance and through walls.

Despite that I have been considering wiring it to the router through the mains if it remains upstairs - I use it wired downstairs.

http://www.amazon.co.uk/Solwise-85Mbps-Homeplug-Adapter-Twin/dp/B002XL6IUU/ref=sr_1_6?s=computers&ie=UTF8&qid=1338666605&sr=1-6