[theaudiohobby]

Why is it that when I ask you guys to be specific you say that I am nitpicking. For my side, I have already provided a reference that associated the SNR with the resolution. From my understanding of your posts, you contend that is not so. If it is not so, then provide some concrete substantiated reasons why it is not, Is that too much to ask?

Jazz, with due respect, all I am trying to do is get a firm handle on your point. I hope you understand.

PS: SNR <= {n.20.log10(2)+10.log10(12/8)} dB and n is the number of bits

therefore for a SNR of exactly 120dB if we have n = (120 - 10.log10(12/8) ) /20.log10(2)
= 120 - 1.7 / 6.02
and this is approx 20 bits.

 

[Wodgy]

Jazz is asking about the time-domain performance of SACD in reproducing high frequencies. The mental graph he's thinking of is amplitude versus time, and he's saying that SACD cannot vary the amplitude quickly enough to represent signals near 20kHz with decent dynamic range.

 

[JaZZ]

Thanks for the translation, Wodgy!

Quote:

Originally posted by theaudiohobby Why is it that when I ask you guys to be specific you say that I am nitpicking. For my side, I have already provided a reference that associated the SNR with the resolution. From my understanding of your posts, you contend that is not so. If it is not so, then provide some concrete substantiated reasons why it is not, Is that too much to ask?

Well, I've asked you a relatively simple question, and you still refuse to give an answer. In turn you are lamenting not to get an answer from me. I can't answer your question because I can't reproduce if what you stated is valid or not, although it would be a gross antagonism to my own calculation. It really would be interesting to find a solution for the problem, but that would imply that you give up your snootiness and (please) do the proposed calculation.

As far as I'm concerned, the corresponding resolution values for 20 kHz and the different formats are as follows:

CD..............16 bit
DVD(-A).......24 bit (up to)
SACD...........7.17 bit (equivalent)

2,822,400 : 20,000 = 141.12 = 2^7.17

[corrected: 241.12 --> 141.12]

 

[theaudiohobby]

Quote:

Originally posted by Wodgy Jazz is asking about the time-domain performance of SACD in reproducing high frequencies. The mental graph he's thinking of is amplitude versus time, and he's saying that SACD cannot vary the amplitude quickly enough to represent signals near 20kHz with decent dynamic range.

Rubbish..we just spent a couple of pages of this thread discussing time-domain performance that is amplitude vs time which is the impulse response. He should read up on some of the links and previous pages on the thread. As I said, if he disagrees with my assertion, he should provide some concrete evidence to that effect.

 

[JaZZ]

Quote:

Originally posted by theaudiohobby ...if he disagrees with my assertion, he should provide some concrete evidence to that effect.

The evidence is right in front or you eyes (and not for the first time!) ... see above!

 

[theaudiohobby]

Jazz,
This part of the thread is at an impasse, when I asked for concrete information, I was asking for any published information that formed the basis of your assertion. If you cannot produce any, then let's call it a day, there is no need wasting time going round in circles.

 

[JaZZ]

Quote:

Originally posted by theaudiohobby Jazz, ...when I asked for concrete information, I was asking for any published information that formed the basis of your assertion. If you cannot produce any, then let's call it a day, there is no need wasting time going round in circles.

I've already offered an important source (IAR) which has influenced my opinion. But why not form your own opinion just on the basis of facts you can easily reproduce!? Do the proposed calculation and you are enlightened!

I really wonder what holds you back from doing so.

 

[theaudiohobby]

Jazz,

Give me a break!! what you have there is at best called conjecture. If that is all you can provide for published information, I withdraw from this part of the thread.

 

[JaZZ]

Well, I agree, Im not completely convinced by the article, but parts of the claims can't be denied -- and one of them is the poor dynamic resolution at high frequencies, easily calculable BTW.

 

[Wodgy]

I wish there was a document somewhere describing the rationale behind SACD from an engineer's perspective (i.e. assuming a solid understanding of traditional signal processing). I'll confess I don't have a good handle on why JaZZ's point is not correct, though I admit that there are many counter-intuitive results in engineering that are no longer counter-intuitive once they're explained. If anyone has any good references, I'd be interested in reading them.

To pose JaZZ's question another way, consider this thought experiment: if you were trying to transmit amplitude-modulated data over a 20 kHz carrier wave through a channel with SACD's properties, what would the maximum data rate be? Would it be higher or lower than Redbook CD?

 

[Joe Bloggs]

Quote:

Originally posted by theaudiohobby Rubbish..we just spent a couple of pages of this thread discussing time-domain performance that is amplitude vs time which is the impulse response. He should read up on some of the links and previous pages on the thread. As I said, if he disagrees with my assertion, he should provide some concrete evidence to that effect.

theaudiohobby, let me reiterate what Jazz is saying here:

A sine wave at 20kHz has a period of 0.00005s. It goes from the top peak to the bottom peak in half a period, 0.000025s.

The space between two sampling points in SACD is 1/2.82MHz = 3.54308....e-7 s.

There is thus only 71 sample points covering the time between this top peak and bottom peak. (rounded up to the nearest integer. I'm being generous here)

If we want SACD to be able to encode a full-scale signal at 20kHz, that means 71 down bits should take us from the highest voltage signal to the lowest.

Which would mean that there are only 71 possible steps an SACD waveform can take--compared to 65536 steps in CD and 16,777,216 steps for 24-bit DVD!

Or if we want SACD to go up to 22050kHz full scale

CD is 16 bits and DVD is 24bits. By this analysis SACD would only be the equivalent of 6.15 bits!

(my numbers differ from Jazz because of two things: firstly, he made a typo: 2^7.17 is 141.12, not 241,12

secondly, he forgot that a sine wave goes from peak to peak twice, not just once, so the number of samples between two peaks should be half of the number he calculated.

)

Note that this is not in contradiction to the impulses we were looking at earlier: for example if you want SACD to have a rise time from bottom to top of 6.122us=0.000006122s (as shown earlier), you just need to reduce the number of steps between top and bottom to 6.122us / 3.54308e-7s(space between two samples) = 18 samples = 4.2bits! (again, rounding up

)

Now, there is something fundamentally flawed about the analysis above, but seeing as you deem yourself to have such a superior understanding of the ins and outs of SACD, I leave it as an exercise to you: oh, pray tell, what is wrong with this conclusion?? I challenge you to explain this yourself instead of referring us to some paper you don't understand

 

[theaudiohobby]

sigh...

 

[theaudiohobby]

Quote:

Originally posted by Wodgy I wish there was a document somewhere describing the rationale behind SACD from an engineer's perspective (i.e. assuming a solid understanding of traditional signal processing) ?

Why not just get a book like the one below from the library

Oversampling Delta-Sigma Data Converters : Theory, Design, and Simulation: James C. Candy (Editor), Gabor C. Temes (Editor)

it will tell you all you need to know

If Jazz's position was correct, there will have been at least one AES/IEE paper doing the rounds parading DSD/SACD woeful high frequency performance. Such an opportunity would be too good to miss.

 

[JaZZ]

Joe...

...you're right, and I thought about halving the number myself, in view of the positive and negative half-wave issue, but decided to leave it at the original calculation, seeing the dynamic range as the whole range the imagined curve can move in -- and then 141 steps is low enough to impress in a negative sense. To make it even worse, a sine wave couldn't even benefit from the full 71 steps (this is reserved to triangle waves), so it comes down to approx. 60 steps. But this just at maximum signal level, so realistic levels would only cover some modest 5 to 20 steps...

And BTW the number (171) was only wrong in the last calculation -- where it is corrected now.

 

[JaZZ]

After all let's not forget that the reason for the need for noise shaping with DSD is a poor original S/N ratio -- and why else than due to the calculated poor dynamic range at high frequencies!