Is the dynamic range of CD sufficient?

[BlueFox]

The real question is; Who cares? There is always something better, and striving to make things better is what makes us human.

 

[Julf]

None of this has anything to do with the original post that kicked it off by one member mentioning graining and "a stair-step".

Indeed. We all agree that there is no stair-step on the output of a properly functioning DAC.

It is because we are mostly arguing technical semantics and scope/context.

We can of course start to argue about that too

Anyway, I can't really add anything more until you reply to my question:

What part don't you agree with in "Pulse width modulation modulates pulse width. Pulse density modulation modulates pulse density. Thus PWM is a subset of PDM with the constraint that all pulses related to one sample are contiguous. PCM (and DSD) is a fixed-frequency, fixed-pulse-width data stream (in the case of standard DSD, the data is 1 bit wide)."?

 

[Orb]

Julf a stair step does exist in both theory and sort of from alias artifacts, sorry but Monty actually accepted this when I debated this some time ago; we both agree it is not "the stairstep" but "a stair or step".


In fact to quote Philips from a paper discussing DSD and the other application note:

PCM: Pulse Code Modulation. A digital format, used for example in CD, whereby a digital signal is represented by an accurate representation (e.g., 16 bits, meaning that the range -1,+1 is subdivided in 65536 sub-intervals) of the wave form at equidistant points in time (for example, in CD 44100 times per second a 16-bit approximation of the wave form is stored).
Pulse Density Modulation: A form of pulse modulation where a large positive signal is represented by a long series of positive pulses; a zero signal is represented by alernating positive and negative pulses.

In a PDM signal, specific amplitude values are not encoded into pulses as they would be in PCM.
Instead it is the relative density of the pulses that corresponds to the analog signal's amplitude.

In other words they do not work the same from the pulse-train/binary code perspective and why all digital manufacturers differentiate between DSD and PCM, and why it is not called 1-bit PCM conversion to PCM conversion but DSD to PCM conversion, and require radically different signal processing.
Going by what you say you would expect the bitstream output to be identical between DSD and PCM (talking of PCM NOT using a SDM DAC, and even if going through modern SDM one notes there is an internal conversion process).
You agree with that yourself yes?
If you feel that strongly maybe you should correct Miska (I think involved in the development of HQPlayer) over on Computeraudiophile and his own thoughts on difference between PCM and DSD-PDM.
Orb

 

[Julf]

Julf a stair step does exist in both theory and sort of from alias artifacts, sorry but Monty actually accepted this when I debated this some time ago; we both agree it is not "the stairstep" but "a stair or step".

I admire your debating skills, but the stair step does not exist "in theory". What you have are discrete time points.

In fact to quote Philips from a paper discussing DSD and the other application note

Could you please answer my earlier question:

What part don't you agree with in "Pulse width modulation modulates pulse width. Pulse density modulation modulates pulse density. Thus PWM is a subset of PDM with the constraint that all pulses related to one sample are contiguous. PCM (and DSD) is a fixed-frequency, fixed-pulse-width data stream (in the case of standard DSD, the data is 1 bit wide)."?

To add another angle to that (and these questions are not just for Orb - anyone feel free to think about it) - forget for a minute what you think DSD is. Let me instead ask "What is your definition of PCM? And what part of that definition doesn't DSD satisfy?". And to turn it around, if you were to construct a 1-bit, oversampling PCM system based on delta-sigma, how would it be different from DSD?

So instead of saying what DSD is (and thus trying to imply that it isn't PCM), please explain in what way it isn't.

Going by what you say you would expect the bitstream output to be identical between DSD and PCM (talking of PCM NOT using a SDM DAC, and even if going through modern SDM one notes there is an internal conversion process).
You agree with that yourself yes?

This is getting silly. Multibit PCM is of course not "identical" to 1-bit PCM - they have different word lengths.

 

[j_j]

As Julf and JJ mention SDM is integral to 1-bit stream in general (never disagreed with that, and more technical info has already been discussed in many threads here), however DSD by Sony utilises PDM encoding.

SACD is 5th order Delta-Sigma. No more, no less. They made up a name for it.

The fact that somebody made up a name for it does not change what it is.

There is a fundamental difference between PCM systems (including delta-sigma, SACD, etc) and PWM, as well.

PCM is a linear system with noise (either multibit or not). PWM is not a linear system, it has both distortions and noise level that varies with output level.

PDM does not specify how the pulses are created. How it's done is very important.

SACD creates the pulses by using 5th order delta-sigma, which makes it a form of PCM.


Orb: Please read the paper by Candy et al. It will clear up some of your misconceptions.

And, of course, there is no "stairstep" anywhere in digital. If you see a stairstep, it is a misrepresentation or a mistake. Many plotting algorithms in the digital arena show stairsteps, that's because they show individual IMPULSES as rectangles instead of impulses. It's easier to see on a plot, but it does not represent the mathematical meaning of the samples. If you have "stairsteps" on output after filtering, you don't have filtering, and the system is not a proper PCM system.

 

[j_j]

If you feel that strongly maybe you should correct Miska (I think involved in the development of HQPlayer) over on Computeraudiophile and his own thoughts on difference between PCM and DSD-PDM.
Orb

Why don't you invite your expert over here.

DSD and SACD are both forms of PCM. DSD isn't even 1-bit, of course. I assume you knew that, right?

 

[Orb]

OK I just managed to find one of the other Philips papers I own on the net.
http://www.icacommission.org/Proceedings/ICA2004Kyoto/pdf/We4.E.3.pdf
2. Characteristics of Direct Stream Digital.
Page 2 has the graph representing both PCM and DSD (bear in mind this is written by engineers involved with DSD).

Figure 1s hows a sine wave both in PCM format and DSD format.
The only visible resemblance between the two wave forms is that where the sine has a large positive value, the DSD stream has a relatively large number of ’+1’ values, whereas in the areas where the sine wave has a large negative value, a lot of ’-1’ values are present in the DSD stream.
The DSD signal is thus akin to a pulse density modulated version of the PCM signal.
Typically, DSD is generated with the use of a Sigma Delta Modulator (SDM - see Sec. 3).
However, this is not mandatory, since the Scarlet Book does not prescribe how a DSD bitstream should be generated [3]. This extreme freedom makes it possible to use other and future techniques for bit-stream generation, e.g., the techniques described in [4, 5, 6, 7].
Rather irrespective of the number of bits, high sample rates in the digital world are desirable because the larger the sample rate, the less the audio artefacts introduced
by the time quantization. We will review a few examples, which show that through the use of SA-CD and DSD, signal distortions due to the time quantization are virtually absent.

Again they are keeping PCM to the classical definition that I have mentioned earlier, due to the fact 1-bit must switch to PDM/PWM type of encoding rather than explicit defined binary code values for amplitude.
Cheers
Orb

 

[Orb]

I admire your debating skills, but the stair step does not exist "in theory". What you have are discrete time points.

Yes it does Julf, it is integral to the transfer function with quantization.
I provided all the information for Monty in the past who conceeded the point, but feel free to disagree with engineering (not DSP) degree level university lectures on this and various other engineers that show the transfer function in detail.
Orb

 

[Julf]

OK I just managed to find one of the other Philips papers I own on the net.

Could you please stop quoting papers and actually answer my questions?

Page 2 has the graph representing both PCM and DSD (bear in mind this is written by engineers involved with DSD).

...
The DSD signal is thus akin to a pulse density modulated version of the PCM signal.
...

Note the word "akin". Do you think there perhaps is a reason why they specifically use that word?

due to the fact 1-bit must switch to PDM/PWM type of encoding rather than explicit defined binary code values for amplitude.

1-bit "must" not switch into anything. 1-bit is a special case of multi-bit. Nothing magic.

 

[Julf]

Yes it does Julf, it is integral to the transfer function with quantization.

As JJ pointed out, there really is no stair step. There is a collection of regularly spaced impulses.

 

[Orb]

As JJ pointed out, there really is no stair step. There is a collection of regularly spaced impulses.

Yes, all those degree engineering courses are wrong in how they present in detail transfer function and its process......
Beat my head against the wall.
But then this is not necessarily DSP focus.....
And btw remember I said it exists in technical theory not physical representation as such, this is why dither is required you know as that theoritical step is what causes the quantization error-distortion, but anyway not going to argue about something that requires many pages to explain to stop both of you arguing as we have been down this path several times in the past with others and I get bored of presenting links showing engineering theory and application of certain aspects of digital.
You do understand that alias artifacts (look via NOS DAC) when looked closely DO look like steps though rather than sinewave, and you seem to forget the Stereophile link earlier showing what a non-dithered low signal looks like
But of course that is a perfect sinewave rather than with flat "steps" [sarcasm], however this cannot be computed by some because digital must never be shown how a natural signal looks before a further artificial noise source is added
Orb

 

[Orb]

Could you please stop quoting papers and actually answer my questions?



Note the word "akin". Do you think there perhaps is a reason why they specifically use that word?



1-bit "must" not switch into anything. 1-bit is a special case of multi-bit. Nothing magic.

I never said 1-bit "must" switch, I said PCM MUST SWITCH TO PDM when going to 1-bit.. not the same thing and now you are twisting my responses by carefully selecting a brief portion of the sentence/paragraph -naughty; there has to be a change in operation because of the way PCM is defined, as has clearly been shown with the previous papers I linked or quoted from.
If you want PCM down to 1-bit, then you MUST go through a SDM (or any other modern 1-bit equivalent thinking of Trellis), but doing so converts PCM to PDM-PWM format and is no longer PCM.......

Anyway.
Funny because all those papers answer the fundamental argument that you still cannot let go; you have stated DSD and PCM are the same, each of those papers explain why they are NOT the same and in fact two of them are by research engineers AT PHILIPS AND SPECIFIC TO DSD
They show what the classicial definition and structure is (fundamental difference in definition and presentation of PCM binary code and PDM-PWM pulse train and how they operate as clearly shown in last document that validates with all the others I have posted), anyway I think I will side with Philips and the application notes by various digital manufacturers

Furthermore they are posted for the benefit of other members who will get more value out of them than reading our arguing.
Definitely recommend reading for those confused with the arguing.
Orb

 

[Julf]

OK I just managed to find one of the other Philips papers I own on the net.

If we are saying something looks like a duck, swims like a duck and quacks like a duck, but that people often keep calling it a pigeon, it doesn't really help if you keep digging up more examples of people calling it a pigeon. What would help is you explaining in what way that something *isn't* like a duck.

 

[Julf]

And btw remember I said it exists in technical theory not physical representation as such, this is why dither is required you know as that theoritical step is what causes the quantization error-distortion
...
But of course that is a perfect sinewave rather than with flat "steps" [sarcasm], however this cannot be computed by some because digital must never be shown how a natural signal looks before a further artificial noise source is added.

I really don't have anything to add to that wonderfully articulate posting.

 

[Orb]

I really don't have anything to add to that wonderfully articulate posting.

Certain words come to mind at the moment Julf that are much better for articulation
Look just type in a search engine "transfer function" and "stair" and "digital"......
Not linking the various degree engineering couse notes I have just to "win" (seems that is how you are approaching this thread to me) an argument as that would be naughty, but it is very easy to find relevant papers on the web.
In fact to help you out just done that and look:
http://www.eecs.umich.edu/eecs/courses/eecs373/labsW14/lab7/index.html

The second case illustrates the affect of quantizing resulting in a stair case like transfer function. In this case the quantization error is from 0 LSB to 1 LSB. The last graph illustrates an ADC model that shifts the transfer function to center the quantizing error over the conversion intervals. In this case, the quantization error is +/- ½ LSB. In both cases the quantizing error is 1 LSB, but the relationship of the conversion value to the input voltage is shifted by ½ LSB.

Wow matches the various degree digital engineer lecture notes I have (they go into much more detail and have 5-10 pages just covering transfer function with digital conversion but validate with that quote)...

And I am sure I can find ones to match the degree lecture notes mentioning "a stair" with alias artifacts.....
Before you respond remember I am stating a) transfer function "stair" is a theoritical step, b) the alias artifacts "looks like" a step or stair especially with NOS DACs due to no reconstruction filter, c) a non-dithered very low 16-bit signal will also "look" like a step structure, try it again at 24-bit and it will be smooth as shown by Stereophile HOWEVER in reality an additional artificial noise source (dither) is added to the original signal so this is not an issue.
Orb

 

[Alrainbow]

I have a very simple a basic test to prove or disprove this argument.
Take a dsd 128 of a really good recording and down sample the recording to different levels . 24/192'. 24 /176. Down to 16/44. And see at what point it matters to the sound. I have some ref class music I use. It's only 16/44. But it is more then enough to show a systems good and bad. For me anything above 24/96 I really cannot discern accurately . I am willing to bet most all here cannot either.

 

[Julf]

Not linking the various degree engineering couse notes I have just to "win" (seems that is how you are approaching this thread to me) an argument as that would be naughty, but it is very easy to find relevant papers on the web.
In fact to help you out just done that and look:
http://www.eecs.umich.edu/eecs/courses/eecs373/labsW14/lab7/index.html

Hmm... Lab notes about measurement accuracy of a sample-and-hold ADC?

Before you respond remember I am stating a) transfer function "stair" is a theoritical step, b) the alias artifacts "looks like" a step or stair especially with NOS DACs due to no reconstruction filter, c) a non-dithered very low 16-bit signal will also "look" like a step structure, try it again at 24-bit and it will be smooth as shown by Stereophile HOWEVER in reality an additional artificial noise source (dither) is added to the original signal so this is not an issue.

Haven't we gone through this a number of times already?

Yes, we all agree you will see stair steps out of a NOS, filterless DAC - because such a DAC lacks the reconstruction filter. But that is an artefact of how that kind of a DAC works. You won't see them at the output of a DAC with the appropriate filter.

We are definitely going round in circles. I am more than happy to declare you the winner of whatever you want to be a winner of. Congratulations!

 

[Julf]

I have a very simple a basic test to prove or disprove this argument.
Take a dsd 128 of a really good recording and down sample the recording to different levels . 24/192'. 24 /176. Down to 16/44. And see at what point it matters to the sound. I have some ref class music I use. It's only 16/44. But it is more then enough to show a systems good and bad. For me anything above 24/96 I really cannot discern accurately . I am willing to bet most all here cannot either.

Indeed. In a double-blind test I did on CA back in the days of the old site (unfortunately now inaccessible) people could not tell 24/96 from 16/44.1, and often even favoured the mp3 I threw in as a control. By far the winner was the mid-resolution version that I had amplified by 1 dB (another control).

 

[DonH50]

I am losing the terminology here, want to interject a few questions to j_j & julf about my assumptions and for clarity (mine and others):

1. A conventional DAC's output (pre-filtered) - or S/H output, not the same as a T/H (real circuit) - to me is a "stairstep" represented mathematically as an impulse and ZOH, true?
2. Are you saying the stairsteps are removed by the image filter, and that is why there is no stairstep visible in the final output?
3. In the practical (circuit) world, impulses are hard to come by, so the output of a delta-sigma DAC is still a pulse (not impulse) train, yes?
4. I am not used to seeing a delta-sigma converter referred to as "linear PCM" (implied in post 467) but rather still a non-linear system due to the quantization. You can make a linear D-S loop, is that the implication? Or do you consider a D-S converter a linear system (which would confuse the heck out of me)? I think you are referring to the model as a linear loop plus (quantization) noise, yes?
5. Perhaps it would be worthwhile to define PCM, PWM, and PDM explicitly for everyone else trying to follow this debate. (I would provide mine but rather it came from the principals.)

Sorry if this is too far off, can take to PM. Perhaps a new thread should be opened to discuss sampling methodologies and architectures.

Curious, thanks - Don

 

[jiannone]

Don, you seem to have a grasp on this beyond my own. I did find this page and its included app to be informative for understanding the stairstep issue further. Here's a bit of wisdom yanked from the mouths of The Matrix babes', there are no stairs.

http://www2.egr.uh.edu/~glover/applets/Sampling/Sampling.html