Audio Myth - "DSD Provides a direct stream from A/D to D/A."

[accwai]

[...] At first glance, it doesn't look "real" to me, but that could have several other explanations besides photoshop.

 

Well, that photo is a focus stack. Basically, you take a series of photos with slightly different focus between frames. Then you run the whole pile through a piece of stacking software, which extracts the sharp parts from each frame and assemble them into a single sharp frame. One rather serious limitation of macrophotography is that as magnification increases, depth of field decreases, real fast. To get the depth of field back, you need to stop down hard, which will then make sure you hit the diffraction limit hard. So there is normally no way to make a really sharp macro photo. But focus stack completely sidestepped this limitation. What you get is photos that are blazing sharp all over, *far* sharper than what the eyes are used to. What makes above example even more awkward is that the magnification is 1x. That means the sensor image is the same size as the subject. The subject is big enough for the eyes to see its overall structure, but not big enough to clearly see its texture. Then the stacking process took the texture and stuff it in your face.

 

So from a scientific standpoint, the photo follows a rather precise mapping process and should qualify as being very accurate. But does it feel real and convincing?

 

Thanks - that was a very entertaining post, and I look forward to your further commentary on it. -Paul

 

You're welcome. In fact, we still have that unfinished business from a number of years back. Will have to look into it when I got a chance

[Paul R]

Well, that photo is a focus stack. Basically, you take a series of photos with slightly different focus between frames. Then you run the whole pile through a piece of stacking software, which extracts the sharp parts from each frame and assemble them into a single sharp frame. One rather serious limitation of macrophotography is that as magnification increases, depth of field decreases, real fast. To get the depth of field back, you need to stop down hard, which will then make sure you hit the diffraction limit hard. So there is normally no way to make a really sharp macro photo. But focus stack completely sidestepped this limitation. What you get is photos that are blazing sharp all over, *far* sharper than what the eyes are used to. What makes above example even more awkward is that the magnification is 1x. That means the sensor image is the same size as the subject. The subject is big enough for the eyes to see its overall structure, but not big enough to clearly see its texture. Then the stacking process took the texture and stuff it in your face.

 

That it does! I guess I rather equate to some of the systems that are hyper detailed - it is a view into a reality we cannot normally experience. Of course, that is true of many things, not just photography and music. It is a core underlying argument in high end audio of course. If a modest system is capable of reproducing music pretty much exactly the way we would hear it normally, then the super high end systems are possibly best equated to a scanning electron microscope.

 

 

 

So from a scientific standpoint, the photo follows a rather precise mapping process and should qualify as being very accurate. But does it feel real and convincing?[/Quote]

 

Not to me, no. It is too good (though I would not have guessed the correct explanation) and therefor triggers a suspicion that it is either not real, or not natural.

 

Which brings us all the way back to PCM vs. DSD. DSD sounds natural to me, but it often comes off as too "soft" for people used to the very unnatural hyper detailed sound that many high end audio systems produce.

 

Not that there is anything wrong with that, I enjoy that photo for the detail in it, and the clarity of the view my "natural" senses would not provide.

 

-Paul

 

 

 

 

 

You're welcome. In fact, we still have that unfinished business from a number of years back. Will have to look into it when I got a chance

[徐中銳]

Which brings us all the way back to PCM vs. DSD. DSD sounds natural to me, but it often comes off as too "soft" for people used to the very unnatural hyper detailed sound that many high end audio systems produce.

 

Paul R, please provide details.

 

As I said yesterday NZDT :

The term DSD often is but a generalisation. If enthusiasts or critics are indeed serious' date=' start with [u']specifics[/u] in product engineering then examine its execution upon...

[Paul R]

Paul R' date=' please provide details.[/font']

 

As I said yesterday NZDT :

 

Well yes, DSD is actually a trademarked name that has been absconded with by the general public. Like Kleenex or Escalator. I am sure that both Sony and Phillips are not all that happy about that, so I see where you are coming from.

 

As I meant it however, I was talking about the format we commonly use, with DSD being DSD64, as found on any number of SACD releases. It just sounds more natural than 16/44.1 PCM, which is the CD format it competes with.

 

In Audiophile terms, I find the sweet spot to be DSD128 (Double Rate), as it is easy for just about any processor to manage. This is especially true when transcoding 16/44.1 PCM to DSD128.

 

I tend to think more of the format of the data than the engineering gear that turns that data into music. The data is "forever" while the engineering tech seems to change on a fairly regular basis about ever 18-24 months.

 

-Paul

[徐中銳]

Back to Chris' opening Post :

Another interesting application note from Benchmark's John Siau. This myth goes something like this:

 

"DSD provides a simple and direct digital path between the A/D and D/A."

"DSD is simpler than PCM."

"DSD is not PCM."

 

While DSD can provide spectacular audio performance, all of the statements above are false. There are many wonderful DSD recordings, but the quality is not due to any virtues of the DSD format. Direct Stream Digital (DSD) seems like a simple and attractive system, but it absolutely fails to deliver a "direct" path between the A/D and the D/A.

Audio Myth -"DSD Provides a Direct Stream from A/D to D/A" - Benchmark Media Systems, Inc.

 

Why would Siau say that ? Who's he ?

 

Well, does one DAC model sound different to another ? Under what conditions ?

 

Paul, it's not you personally, the objective science of the matter concerns specifics in practice (whether recording *in DSD* or its reproduction).

[Hiro]

Well' date=' does one [i']DAC[/i] model sound different to another ? Under what conditions ?

 

 

As Ken Ishiwata said:

 

‘DSD can by-pass certain processing within those [delta-sigma] D-to-A converter chips, so you …. get a less processed signal with DSD compared to PCM, which of course will influence the sound quality.’

 

https://andreweverard.com/2013/09/24/review-marantz-na-11s1-a-very-good-thing-worth-the-very-long-wait/

 

Interestingly, as the article reports, "Ishiwata is also an advocate of upconverting existing CD-quality files to the DSD format in the computer, and then playing them back through a DSD DAC such as the NA-11S1."

 

As you've probably heard, his upcoming reference SA10 SACD player/DAC upconverts PCM to DSD 11.2MHz...

[mordante]

As Ken Ishiwata said:

 

‘DSD can by-pass certain processing within those [delta-sigma] D-to-A converter chips, so you …. get a less processed signal with DSD compared to PCM, which of course will influence the sound quality.’

 

https://andreweverard.com/2013/09/24/review-marantz-na-11s1-a-very-good-thing-worth-the-very-long-wait/

 

Interestingly, as the article reports, "Ishiwata is also an advocate of upconverting existing CD-quality files to the DSD format in the computer, and then playing them back through a DSD DAC such as the NA-11S1."

 

As you've probably heard, his upcoming reference SA10 SACD player/DAC upconverts PCM to DSD 11.2MHz...

 

Better solution would be to stop using delta sigma dacs it would seem.

 

DSD seems to be a cure for a disease you can vaccinate against.

[Hiro]

Better solution would be to stop using delta sigma dacs it would seem.

 

DSD seems to be a cure for a disease you can vaccinate against.

 

Native DSD/SDM doesn't have ringing, aliasing and crossover distortion like PCM.

 

DSD is a good way to avoid these maladies of PCM dacs.

 

As for stopping using SDM, that would be rather difficult, given that 99,999% of the DAC market and 100% of modern ADC market uses delta sigma modulation.

[Paul R]

Paul' date=' it's not you personally, the [i']objective science[/i] of the matter concerns specifics in practice (whether recording *in DSD* or its reproduction).

 

I get that, I think.

 

Honestly? The fundamental data is the same, whether it is PCM or DSD or even Analog. In comparison to analog, the fidelity of digital recording is much higher, and the difference in fidelity between DSD and PCM is really difficult to get objective results for.

 

In practice, DSD is a simpler data format, so it is easier to convert to and from analog. PCM came along more because of the restrictions on processing power and data storage than any inherent superiority in the data format. When you objectively look at the data formats and data transforms, DSD is much simpler than PCM. That makes it preferable for several technical reasons.

 

That's of course, from my (professional) point of view. A practicing digital design engineer might feel differently, because the off the shelf options open to him or her with PCM may be greater. Or because he or she is not accustomed to or doesn't totally understand noise shaping, or how to filter ultrasonics. Or any of a dozen more reasons. And from my hobbyist point of view, I definitely feel that DSD provides a more natural and enjoyable listening experience.

[mordante]

Native DSD/SDM doesn't have ringing, aliasing and crossover distortion like PCM.

 

DSD is a good way to avoid these maladies of PCM dacs.

 

As for stopping using SDM, that would be rather difficult, given that 99,999% of the DAC market and 100% of modern ADC market uses delta sigma modulation.

 

Also 100% of all music sold, recorded, mixed and edited is PCM.

[Hiro]

Also 100% of all music sold, recorded, mixed and edited is PCM.

 

There are record labels doing native DSD recordings and direct analog-to-DSD transfers.

 

What's more, I've read on this forum that new mixing capabilities for DSD are coming soon...

[Paul R]

There are record labels doing native DSD recordings, and direct analog-to-DSD transfers.

 

What's more, I've read on this forum that new mixing capabilities for DSD are coming soon...

 

Not to mention those hundreds of thousands of analog recordings out there...

[Hiro]

Not to mention those hundreds of thousands of analog recordings out there...

 

Yep. To say that all recordings are PCM is certainly not accurate...

[徐中銳]

Quickly, let's remember another engineer of DSD :

Recalling Koch's explanation to the flow chart of my previous Post' date=' extracted from his [/size']DSD - the new Addiction :

« The term Direct Stream Digital (DSD) was coined by Sony and Philips when they jointly launched the SACD format. It is nothing else than processed Delta-Sigma modulation first developed by Philips in the 1970’s. Its first wide market entry was not until later in the 1980’s when it was used as an intermediate format inside A/D and D/A converter chips.

 

Fig. 1:

 

Figure 1 shows how an analog source is converted to digital PCM through the A/D converter and then back again to analog via the D/A converter. The A/D internally contains 2 distinct processes:

 

1. Delta-Sigma modulation: the analog signal is converted directly to DSD with a very high sampling rate. Various algorithms are in use depending on the application and required fidelity. They can generate 1-bit DSD or multibit DSD oversampled at 64x or 128x compared to regular CD rate.
   
2. Decimation filter: the DSD signal from the previous step is downsampled and converted to PCM. Word length is increased (for instance 16 or 24 bits) and sample rate reduced to CD rate or a low multiple of it for high resolution PCM formats.
   

 

The D/A process is very similar where:

 

1. the PCM signal is upconverted to a much higher sample rate.
   
2. then converted to DSD via the Delta-Sigma modulator (to reduce word length)
   
3. then converted to analog.
   

 

This technology was chosen because of its improved linearity and consistent quality behavior across physical components, as most of the heavy duty signal processing was shifted to the digital domain where it was not susceptible to variability of electronic components. It was quickly adopted in most converter systems and we can say that since about the late 1980’s we have been listening to some form of DSD without even knowing it.

 

As science progressed as well as our experience with digital audio, we started to realize that the algorithms for the DSD-to-PCM and PCM-to-DSD conversions can have a profound impact on the sonic performance when they are developed according to classic formulas. These are relatively complicated algorithms and they introduced a new phenomenon that we describe as “digital sound” or ringing effects. Hence the motivation by the engineering teams of Sony and Philips to remove these steps altogether from the conversions between analog and digital. This simplified DSD path that bypasses the PCM path is shown in Fig. 1 above. As is usually the case most simplifications in the signal path lead to sonic improvements and so it didn’t come as a surprise when first listening tests were so astonishing that this format was considered as an archiving format for recording studios. That alone says something about its sonic fidelity. At the time no recording studio was even considering using any PCM format to archive its analog recordings. »

[Paul R]

Quickly' date=' let's remember another engineer of DSD[/url'] :

 

Which a wonderful explanation. The download speeds section is quite dated though - a 512mb download here is done in a few tens of seconds - well under a minute.

 

That kind of widely available speed really changes things.

 

-Paul

[徐中銳]

A brief analogy to the audiophile—of appreciating quality' date=' consider what [b']Magnum[/b] Photographer Elliott Erwitt said :

« Quality doesn't mean deep blacks and whatever tonal range. That's not quality, that's a kind of quality... »

 

Have readers seen before the following photo, North Carolina (1950) by Erwitt ?

 

 

Simply (for now), is DSD and PCM a segregation-of-sorts through myth, ignorance, prejudice ?

 

Before others say so, does the less-processed water on-the-left taste better ?

[semente]

Yes, multi-bit SDM. The only way to get DSD out of them is by noise-inducing quantisation. The only way to get PCM is by downsampling. Pick your poison.

 

Isn't it possible to make a multi-bit SDM file format?

Why wasn't SACD conceived as multi-bit in the first place?

 

R

[Miska]

Isn't it possible to make a multi-bit SDM file format?

Why wasn't SACD conceived as multi-bit in the first place?

 

Lot of typical DSD recording gear (Korg and TASCAM) uses TI PCM4202/PCM4204 ADC chip, which I'm also using on my ADC. And that is true "1-bit" ADC converter.

 

Yes, it is simple and I've been using such for my own purposes. But using multi-bit SDM doesn't practically change anything. It is as much noise-inducing and just takes more space.

 

Like I said before, if it doesn't utilize all the bits all the time, then it becomes equal to DSD everywhere else except waveform peaks of loudest points. If it does utilize all the bits all the time, the difference is very small apart from disk space consumption.

[mansr]

Isn't it possible to make a multi-bit SDM file format?

 

It would be possible. For whatever reason, nobody has done this, at least not for wide distribution. Of course nothing stops you padding to 8 bits and storing it as WAV or any other standard format. I do that with 1-bit DSD when I need to analyse it with tools that don't support DSF.

 

Why wasn't SACD conceived as multi-bit in the first place?

 

I don't know for sure. It was developed in the 90s with 1-bit converters prevalent. Perhaps that had something to do with it.

[Hiro]

Lot of typical DSD recording gear (Korg and TASCAM) uses TI PCM4202/PCM4204 ADC chip, which I'm also using on my ADC. And that is true "1-bit" ADC converter.

 

Yes, it is simple and I've been using such for my own purposes. But using multi-bit SDM doesn't practically change anything. It is as much noise-inducing and just takes more space.

 

Like I said before, if it doesn't utilize all the bits all the time, then it becomes equal to DSD everywhere else except waveform peaks of loudest points. If it does utilize all the bits all the time, the difference is very small apart from disk space consumption.

 

EMM Labs ADC8 MKIV is another ADC that comes to mind, it can record directly to 5.6MHz DSD.

[徐中銳]

As Ken Ishiwata said:

 

‘DSD can by-pass certain processing within those [delta-sigma] D-to-A converter chips, so you …. get a less processed signal with DSD compared to PCM, which of course will influence the sound quality.’

 

https://andreweverard.com/2013/09/24/review-marantz-na-11s1-a-very-good-thing-worth-the-very-long-wait/

 

Interestingly, as the article reports, "Ishiwata is also an advocate of upconverting existing CD-quality files to the DSD format in the computer, and then playing them back through a DSD DAC such as the NA-11S1."

 

As you've probably heard, his upcoming reference SA10 SACD player/DAC upconverts PCM to DSD 11.2MHz...

 

Well, supplementing your interest, recalling (and my underlining that of) the following paragraph by Andy Boxall :

« The player doesn’t stop at CD and SACD playback, and will recognize discs burned with hi-res files, or those taken from a USB source. Marantz calls the SA-10 a “clean design” product, preferring to build it from the ground up, rather than re-use established techniques, including a new way of converting digital signals over to analog. All digital signals, whether they’re from a disc of file, are converted over to DSD256 — four times the sample rate SACD uses — then passed through a DSD converter before reaching the amp. Other [DACs may] down-convert to PCM, but Marantz says its way results in the purest sound. »

 

[Miska]

I don't know for sure. It was developed in the 90s with 1-bit converters prevalent. Perhaps that had something to do with it.

 

Most digital and analog PWM systems are still two level (some call it 1-bit although it is inaccurate and ambiguous way to talk about it) today. In fact it is far more popular now, since class-D amps are widely used throughout, down to mobile phone headphone amps and such.

 

DSD is just digital PWM system, simple.

[Marco]

While we're at the high sampling frequency thread, I've been scratching my head over this for a while:

 

Let's take an ADC, and it samples at, e.g. 40000 samples per second. You use 2 channels and both of them are connected to a single microphone(so now you have a circuit that just makes a stereo recording).

 

Let's now assume, for easy understanding, that we are going to record a sine tone, produced by a single tweeter.

Let's say this sine tone is a 20KHz signal. Shannon now states it is easily and fully mastered into an electric signal, with no loss of information(amplitude and frequency information is as good as can be: "transparent").

 

Now, do we agree that in order to record the signal, both converter channels must measure the signal at exactly the same time?

 

Now, we go a step further.

 

We now change the distance from the tweeter to one of the microphones (or vice versa!).

 

Question:

What is the amplitude of both recorded signals if the change in distance to one of the mics has shifted exactly a quarter of a wavelength of 20KHz?

 

Let me know what you all think,

 

Marco

[mansr]

While we're at the high sampling frequency thread, I've been scratching my head over this for a while:

 

Let's take an ADC, and it samples at, e.g. 40000 samples per second. You use 2 channels and both of them are connected to a single microphone(so now you have a circuit that just makes a stereo recording).

 

Let's now assume, for easy understanding, that we are going to record a sine tone, produced by a single tweeter.

Let's say this sine tone is a 20KHz signal. Shannon now states it is easily and fully mastered into an electric signal, with no loss of information(amplitude and frequency information is as good as can be: "transparent").

 

Did you mean for the sampling rate to be exactly twice the signal frequency? The sampling theorem requires the sampling rate to be strictly greater than twice the signal frequency for an unambiguous result to be obtained.

 

Now, do we agree that in order to record the signal, both converter channels must measure the signal at exactly the same time?

 

To get a proper stereo recording, the ADCs should use synchronised clocks, yes.

 

Now, we go a step further.

 

We now change the distance from the tweeter to one of the microphones (or vice versa!).

 

Question:

What is the amplitude of both recorded signals if the change in distance to one of the mics has shifted exactly a quarter of a wavelength of 20KHz?

 

Moving the microphone will change the sound pressure according the inverse square of the distance to the speaker. A quarter wavelength at 20 kHz is 4.25 mm. If the distance was originally 1 m and the microphone was moved away from the speaker, this gives a drop in sound pressure by about 0.84%. The sound will also be delayed (phase-shifted) by one quarter wavelength (90 degrees).

 

With proper sampling, the captured waveform will have amplitude and phase per above. Sampled at exactly twice the signal frequency, the phase is indeterminate and the amplitude depends on where on the cycle the sample points happen to fall. It could be higher or lower than before moving the microphone.

[Marco]

Did you mean for the sampling rate to be exactly twice the signal frequency? The sampling theorem requires the sampling rate to be strictly greater than twice the signal frequency for an unambiguous result to be obtained.

 

To get a proper stereo recording, the ADCs should use synchronised clocks, yes.

 

Absolutely, while being purely theoretical, this was the easiest way to describe "the problem".

 

 

Moving the microphone will change the sound pressure according the inverse square of the distance to the speaker. A quarter wavelength at 20 kHz is 4.25 mm. If the distance was originally 1 m and the microphone was moved away from the speaker, this gives a drop in sound pressure by about 0.84%. The sound will also be delayed (phase-shifted) by one quarter wavelength (90 degrees).

 

I didn`t even think of the level drop because of the distance, while being obvious, thanks for pointing that out mansr.

If I understand correctly (and what I was hoping for to be honest), was confirmation of the fact that there are various frequencies that won`t (can`t!) be recorded properly. This can happen because the exact sampling moment is too early or too late (which will result in a recorded level change), or the distance from e.g. instrument to the microphone is changed (oh no, moving musicians), or any combination.

 

With proper sampling, the captured waveform will have amplitude and phase per above. Sampled at exactly twice the signal frequency, the phase is indeterminate and the amplitude depends on where on the cycle the sample points happen to fall. It could be higher or lower than before moving the microphone.

 

Thanks mansr,

What I am trying to figure out is: if there are clear relations between signal (given by an instruments/voice base tone, it`s harmonics and its placement) and the low sampling frequencies, to the point that amplitude, phase (and with that also localisation between the speakers) of it are just a result of chaotic behaviour (or mathematics, but very hard to predict in real life), how come no high res/dxd/dsd record label advocates this?

 

It just shows that it`s not perse the bandwidth that is of importance, but the "phase fidelity" and the levels of the stereo channels because of it as well. If i`m not mistaken this all looks like a data related modulation circuit.

 

How does one find reliable distortion, phase and level related figures (especially with high tones, cymbals etc) with these techniques, or am I exaggerating the whole thing? I`m guessing the use of static and low frequency (1KHz) signals is just a way to avoid the whole thing.

 

Isn`t DSD by this nature (and keeping it above 2 MHz from ADC to DAC), the better recording technique?