Please explain SACD to me...

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi guys.

I'd like to know more about the SACD recording format. Could someone explain it concisely? Or refer me to a good source of information on this?

I understand the copyright-based restrictions on SACD players with digital-outs. But is it possible to modify an SACD player to have a digital output and then build my own DAC/LPF?

Thanks
 
The DSD format used in SACD is entirely different from the usual PCM in almost every other digital audio gear. Basically it's the raw output of a delta sigma modulator (after which usually a decimation filter converts the data stream to PCM). Of course you can tap the signals right before the DAC chip in your player, build a line transmitter and get them out to an external DAC.

However, there are few pure DSD DAC chips on the market today, and quality is not that different. Providing the internal DAC of your SACD player with better operating conditions (new, separate PSU, clean clock, clean ground) and swapping the analog signal conditioning circuitry seems the easiest and best approach.

As of today, many SACD players internally convert the DSD format to PCM, which defeats the purpose of having DSD in the first place, but leaves a backdoor to simply tap the I2S lines, add a S/PDIF transmitter and use any standard commercial or DIY external DAC.

It is no secret that I'm not a fan of SACD at all (it actually sucks) and generally recommend buying a decent DVD-audio capable player for tweaking.
 
AMT-freak said:

It is no secret that I'm not a fan of SACD at all



I am not a fan too.I used for a period a sony scd xb9 (xb 940) and "innovation" and the more simplicity in the treatment of digital signals is not for nothing clear. I have been able to count at least 4 smd crystals for clocks in the main board.

I have tried the Pass D1 output stage balanced, but in along term I listen of, this player was revealed to be much disappointing one.

I am also a lot more interested in the possibility to modify a DVD A too.
 
Which ones?

If you expect a complete list, I will disappoint you. Whenever recently I had a look at high rez photos of SACD players in audio mags, all I could see were current Analog Devices or Burr Brown PCM input DAC chips. The only one I remember right now was an Accuphase, but there are others.

I guess they are simply re-using their DAC backends from CD or DVD players. I expect this to change as DSD devices become more popular, but what I don't expect to change is the :bawling: way of implementing PSUs, clocks and analog output stages. And DSD sucks a whole lot more if you don't do it right.
 
Sacd

AMT-freak said:
It is no secret that I'm not a fan of SACD at all (it actually sucks) and generally recommend buying a decent DVD-audio capable player for tweaking.

Hi AMT-Freak,
My exposure to SCAD is limited to the Sony SCD-1 of a friend of mine. My back is also remembering the Sony as this machine is really too heavy to lift.
I installed the Allen Wright prints for him and we did a clock shootout. I feel the Allen Wright mod is much better than Sony's circuit but my friend disagrees.
We tried of course the standard SMD clock made by KDS, the LCaudio XO-2, The Tent clock on the small Allen Wright board and the KWAK-CLOCK in its 45 MHz incarnation.
With the latter clock and Allen Wrights prints a stunning realism and incredible depth of soundstage is obtained. SACD has more definition and much more dynamics as subjectively experienced.
I learned that SACD is MUCH clock sensitive than normal CD. I did not like the Sony at all in the original state.
The problem with DVD Audio is that even less titles are available than for SACD.:bawling:
 
I'm quite aware of the software problem ( :bawling: ) and I don't debate that SACD can beat CD in some respects, but the theoretical/objective performance of the DVD-Audio standard is much better than 2.8 MHz DSD.

There are some nice sound-only DVD-Video discs (yep, it sounds stupid). Chesky records for example. Plus a small selection of DVD-Audio discs that can only be played in one of the few high-qulity DVD-Audio players.

Soon there'll be a DVD-Video player in almost every household in Europe, and I wonder why the industry once more completely screwed it up. :mad: (I think it shouldn't have been that much a problem to implement DVD-Audio in the DVD family right from the beginning and urge the manufacturers to build players that accept the complete family of discs.)
 
I have an entry-level SACD player (Sony DVP-NC650V) and the only reason for buying it is that the analog stages are better than similarly priced CD players. That said, I've never believed that SACD or DVD-A will ever become the mainstream format. The general public just doesn't discriminate the quality of audio that well. In fact, the rising popularity of MP3 only indicates that the typical consumer cannot possibly hear what "we" can discern.

One question I've always had is: How does the data rate translate into theoretical sound quality between the two formats?

SACD => 2.8Mbit/sec @ 1-bit 2.8MHz clock
DVDA => 4.6Mbit/sec @ 24-bit 192kHz clock

Yet, my basic digital theory tells me that freq resp for SACD should be 1.4MHz while only 96kHz for DVD-A.

:)ensen.
 
The problem with this question is that conversions from the analog to the digital domain and vice versa cannot be understood intuitively. You can easily look at the corresponding mathematical model, but this doesn't tell you the relation between data rate and "psycho acoustically relevant" audio information stored.

PCM is actually easy to understand, but still people repeatedly (and intuitively) claim that a 10 kHz signal is reproduced more accurately at a higher sampling frequency. They claim that more points of the wave are sampled and stored digitally, so the reproduction should be better. This is not true. A 10 kHz signal is by definition a sign wave, and this is perfectly reproduced with a sampling frequency of 20 kHz. Things will (in theory) not improve with a higher sampling rate. Thus, the sampling rate in PCM defines the maximum frequency range we can transmit. Then there is the resolution of each sample, i.e. the number of bits with which the amplitude value is stored. The resolution (and dynamic range) doubles for each added bit. 24 bits (144dB) is 256 times better than 16 bits (96dB). Please keep in mind that in theory the resolution(dynamic range) of PCM is equal in the complete frequency range (0 Hz ... 1/2 fs).

Personally I don't worry that much about the frequency range above 40KHz, so to me a sampling rate of 96KHz, let alone 192kHz is perfectly adequate. Also a resolution of 24 bits is more than plenty and I guess in most signal chains the last 6 to 8 bits of a DVD recording will play below the noise floor.

Now, let's look at DSD. DSD is the raw output of a delta sigma modulator running at 2.8 MHz. That means we can transmit very high frequencies up to 1.4 MHz, but the resolution will decrease with increasing frequency. By a stupid coincidence, we will find that most human created music has an energy spectrum which is exactly the opposite of that. Your tweeters get much less energy than your woofers. In other words, DSD lacks resolution where we need it (subtle low level high frequency signals) and offers plenty where it is not that important (bass drum near -0dB). I will not go into detail here (unless the forum software is updated so that I can type tex formulas ;) ), but experts claim that above about 5KHz the resolution of CD (16bit PCM) is better than 2.8 MHz DSD. So the questions is whether the high frequency range of DSD is actually usable.

I clearly say that I didn't have the opportunity to make a real comparison between the two systems (sound wise), I'm only talking about theory.

BTW, almost all audio A/D converters are delta sigma converters running at above 6 MHz, so the question which system wins sonically might come down to whether a reduction to 2.8 MHz DSD or 96kHz/24bit PCM is losing more relevant audio information.
 
DVD-A is only 24/96? I thought it ran at 192kHz?

AMT-freak said:
Now, let's look at DSD. DSD is the raw output of a delta sigma modulator running at 2.8 MHz. That means we can transmit very high frequencies up to 1.4 MHz, but the resolution will decrease with increasing frequency. By a stupid coincidence, we will find that most human created music has an energy spectrum which is exactly the opposite of that. Your tweeters get much less energy than your woofers. In other words, DSD lacks resolution where we need it (subtle low level high frequency signals) and offers plenty where it is not that important (bass drum near -0dB). I will not go into detail here (unless the forum software is updated so that I can type tex formulas ;) ), but experts claim that above about 5KHz the resolution of CD (16bit PCM) is better than 2.8 MHz DSD. So the questions is whether the high frequency range of DSD is actually usable.

That explains why bass notes are noticeably smoother with SACD vs CD but then the experts create a paradox since with DSD the highs are also smoother.

Then there is that paper by James Boyk in which he measures ultrasonic energy >35kHz for string and wind instruments and as high as 102kHz for cymbals. The microphones

http://www.cco.caltech.edu/~boyk/spectra/spectra.htm

Unfortunately he did not repeat his measurements in different venues to see if the ultrasonics are indeed different. Nor did he experiement with different microphone placements. He may have well proved that there is a "sound" to what we all keep describing with spatial terms such as air, soundstage, separation and placement.

My feeling in this regard is that the difference between "live and Memorex" are those ultra-highs. It may be why we know someone is calling from the the living room and not from the bedroom down the hall. I've found that to be kind of strange, every adult can discern it, but we have no real explanations for why we can do it nor does anybody really ask.

AMT-freak said:

BTW, almost all audio A/D converters are delta sigma converters running at above 6 MHz, so the question which system wins sonically might come down to whether a reduction to 2.8 MHz DSD or 96kHz/24bit PCM is losing more relevant audio information.

Could it be because the 1-bit is easier for the DAC to handle than dealing with conversion of a 24-bit word. Is the 24-bit word "slurred" by the slew rate of the following analog stages?

:)ensen.
 
Then there is that paper by James Boyk in which he measures ultrasonic energy >35kHz for string and wind instruments and as high as 102kHz for cymbals.

I don't think anyone denies that music instruments disturb air at high frequencies. But energy specta in music have a 1/f characteristic, and the ear has a limited band of sensitivity. Thus you certainly have to draw the line somewhere.

I'll just remind everyone that you can still get multibit R2R DACs running all the way up to 768KHz fs.

Thanks for the interesting link.
 
I think everyone here is missing one of the major advantages of DSD. That is, since there is no word length, there is NO DIGITAL CLIPPING! How cool is that? Basically, with DSD, a 1 bit means up and a 0 bit means down. Now do that 2,800,000 times every second, and you can get a pretty close aproximation of the original wave form. With PCM, you wind up with much rougher steps between each sample, and you have to use some pretty serious high-slope filters to get rid of extranious HF information. Also with PCM, you are limited to a word-length.....16 bits for CD and 24 bits for DVD-A. Theoretically, you can get more dynamic range with DSD than with PCM.

Now, don't get me wrong. I have a great DVD-A set-up at home. I really like it, but it is still PCM. When I listen to professional DSD equipment from Sony and Tascam, I like the sound much more than 24 bit PCM. To me, DVD-A just sounds like 5.1 channels of PCM. DSD, to my ears, sounds more like real audio.

Cheers,
Zach
 
That is, since there is no word length, there is NO DIGITAL CLIPPING! How cool is hat?

Well, digital clipping should not be a problem. It has occured in some productions, but it can only happen during recording, and any audiophile recording engineer will know how to avoid it. Period.

Basically, with DSD, a 1 bit means up and a 0 bit means down. Now do that 2,800,000 times every second, and you can get a pretty close aproximation of the original wave form.

Although it's not exactly like that, your model has some validity to explain in a simple way what's happening. You will agree that the higher the frequency, the lower the accuracy will be. Take a signal of, say, 14 kHz. In your model there will be time for only 200 steps of "up" or "down" during the whole period which is not much. If the amplitude of the signal is high, there will be a series of ones but you still can't keep up with the slope of the signal.

With PCM, you wind up with much rougher steps between each sample, and you have to use some pretty serious high-slope filters to get rid of extranious HF information.

As long as the 2*f < fs condition is met, 24 bit PCM still resolves every sample in 16,777,216 steps. The coarse resolution in the time domain doesn't matter as explained in the previous post. The only consequence is that the signal is bandwith-limited. As to the HF crap, DSD behaves much worse than PCM in that respect. With oversampling you can separate audio information and HF crap in PCM, so you don't need a steep filter. DSD in turn has very high HF crap levels in the higher frequency range, and care must be taken to get rid of it. In fact a DSD "DAC" is nothing but a filter.

Theoretically, you can get more dynamic range with DSD than with PCM.

No, this is not true. 24 bit PCM has a constant dynamic range of 144dB over the whole frequency spectrum. DSD is better at lower frequencies (do we need to get better than 144dB anyway?), but loses against CD (96dB) above about 5kHz. In the ultrasonic range, dynamic range is down to a ridicilously low -20 dB.
 
My feeling in this regard is that the difference between "live and Memorex" are those ultra-highs. It may be why we know someone is calling from the the living room and not from the bedroom down the hall. I've found that to be kind of strange, every adult can discern it, but we have no real explanations for why we can do it nor does anybody really ask.

I'd account that to the fact that almost every human has two ears and a brain to processes phase and level differences, "calculating" the direction a sound comes from. ;)
 
AMT-freak said:


Well, digital clipping should not be a problem. It has occured in some productions, but it can only happen during recording, and any audiophile recording engineer will know how to avoid it. Period.



You are correct here. It is not a problem with playback. But, digital clipping sounds horrible when it occurs during the recording. One thing I hate doing when I record orchestras using PCM is setting the recording levels at -20 just so I don't clip at that one point when the orchestra gets really loud. When I'm recording at -20, I'm not actually recording at 24 bits anymore. I'm actually recording more like 17 bits. This, to me, is a terrible comprimise in dynamic range when recording highly dynamic music such as symphonic pieces. This, for me, is where DSD really shines. I can run the input signals as high as my analog inputs will allow me, and never have to worry about that one big peak the orchestra is going to play 2/3s of the way through the program. Another nice thing about using DSD during the initial recording is, I don't have to limit the dynamic range of the inputs by using a limiter. A limiter is almost a neccesary evil when recording highly dynamic music in a live situation because a limited signal will always sound better than a digitally clipped signal.

Cheers,
Zach
 
When I'm recording at -20, I'm not actually recording at 24 bits anymore. I'm actually recording more like 17 bits. This, to me, is a terrible comprimise in dynamic range when recording highly dynamic music such as symphonic pieces.

At -20dB, it's more like between 20 and 21 bits ;) This is about equal to the dynamic range of the analog circuitry before the ADC (mic amps). And the resolution / dynamic range in DSD is anything but unlimited. Recording levels do matter.
 
AMT-freak said:


I'd account that to the fact that almost every human has two ears and a brain to processes phase and level differences, "calculating" the direction a sound comes from. ;)

Sounds received at two points can only determine a near-360 deg circle of directions. Theoretically, we'd need a third ear to determine angle of attack, yet we can do this with only two. Plus, humans can actually determine that a sound that comes from outside bedroom door begins just around the corner or farther down the hallway, or even all the way around into another bedroom. That cannot be possible unless there is another way the "signal" can be processed. The question is how, and I'm suggesting that ultrasonics can play a part in some type of biological SRS algorithm.

:)ensen.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.