|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: Reduced S/N ratio? posted by Jim Pearce on January 30, 2003 at 11:24:36:
I don't remember specifics. I would need to look over what testing they are doing to see what they are measuring. Reduced S/N versus 16-bit CD? Hard to believe. Don't forget there are many types of noise. The quantization noise of Sony's DSD is claimed to be well below the quantization noise of 16-bit CD in the audio range, that is up to 20 kHz, and I see no reason to question that. But there are other noise sources.
Follow Ups:
Just look in the equipment archives. And yes, the S/N ratio is below CD. JA generally comments that there should be no subjective consequences. I have trouble reconciling these measurements with Sony's claims, but perhaps there is a simple explanation.
and it showed the SACD noise level below the CD noise level up to 11 kHz and above it after that. As you say, JA says this is probably not important. This conflicts with the specs on the machine, which say the SACD dynamic range is higher than CD dynamic range up to 20 kHz. I can't explain this discrepancy. But this is above 11 kHz only.
I have trouble following Frank's modulation argument.
it is not at all clear to me what information is being presented here. The difference between the top and bottom chart is the presence of a dithered 1 kHz tone. This has effects I do not understand. At 10 kHz, the "digital black" is about -110 dB for SACD and about -120 dB for CD (CD is better by this measure). With the tone, the 10 kHz numbers are still -110 dB for SACD but up to maybe
-108 dB for CD (now SACD is better at this frequency). I don't know what this means. JA seems to refer to the chart with the tone when he discusses "dynamic range". I will need to figure out what tests they are doing and what they mean.
It's easy to spot the rise in noise if a signal is played back.
Wat will happen if you lower the signal to 'digital zero' The noisefloor will drop and if the signal goes up it rises.I call that modulation. Now it's only noise but what if a 10kHz signal is present next to the 1kHz tone?
It gets also modulated.
Frank,
I think that it is likely that JA stopped printing the digital black test because he was having trouble with different muting schemes and because he discovered that the presence of a signal was affecting the noise. Perhaps we'll know more when he writes about these tests. What is pretty clear to me is that DSD performs much better in the digital black tests than in the 1 kHz tone tests. Is that modulation? What else would it be.
An even more fun test for SACD/DSD would be to run a series of tests,
with say 100, 1K and 10K test signals and look at the effects of the noise as a function of frequency.I can't prove this, but I believe that the degree of noise modulation will increase as a function of frequency.....
I might not be expressing this well, so this is a 2nd explanation.
I suspect that the higher the frequency tested, the worse the noise modulation gets above the tested frequency.
Regards,
On this page you see both graphs for digital black and a test signal at -60dB at the same page. Setting for both measuremens are the same.
At 10kHz the noise floor at 10kHz rises by 10dB.Clear evidence that the noise is modulated by the signal.
The level of noise isn't really important.
What is important that the so called dsd noise is in fact distortion that doesn't belong in the signal.Frank
For the regular CD player, the noise at 10 kHz is about 10 dB higher with signal compared to without signal.Maybe they put the signal generator too close to the player- who knows.
It's got nothing to do with DSD or the noise being modulated by the signal.
from a signalgenerator next to the player.The rise for the cd signal is less severe than the sacd's modulation but it indicates 1 bit dac technology is used.
The signal modulates the noise there's no doubt about it.
see link. I wonder if you could find any CD player of any design that does not.
"This is due to the high-order noise-shaping used in the Crystal delta-sigma DAC to achieve as high a resolution as possible below that frequency. "It's evidence of 1 bit dac technology and noise shaping
it can be found in the latest Stereophile, in the review of the Ayre D1 DVD/CD player. The plot for the spectrum given the dithered 1 kHz signal is substantially identical to that of the Meridian 508.24. It rises to -108 dB at 10 kHz.They say in the write-up that the noise floor in this graph is the same as the noise floor of the dither in the input signal.
That is your answer. This has nothing to do with noise shaping. They are inputting a signal with a rising noise floor, so that is what they are getting at their output.
This is why all the Stereophile graphs of this test output look pretty much the same.
Certainly, the noise shaper is not moving any signal energy. Nothing of the sort happens.
Normal dither will have the effect that the noisefloor rises across the entire bandwith and with a signal a 1.5..3dB rise should occur.With dsd and noiseshaping you see the rise in noise only above a certain frequency.
Yes, the graphs are generated by a dithered 1 kHz signal. Yes, they all look similar. No, they're not all the same and in fact they can differ considerably. It is the differences in the noise floor at various frequencies that JA draws attention to: they are significant. Frank is right about one thing: noise-shaping leads to a much steeper slope.
this has kind of gotten off on a tangent. The behavior Frank noticed led him to conclude that SACD noise shaping also shapes the signal. I pointed out that the same behavior occurs on CD player graphs. He said this was also due to noise shaping.Well I looked at these output charts for about 5 different CD players. They had slightly different wiggles in slightly different places. But they were all basically the same shape. And then I read the text in the latest issue.
It is very obvious that the change in the noise floor between "digital black" and the dithered signal is primarily due to the noise floor in the input signal due to the dither. It has nothing to do with whether the DAC is single bit or not.
OK this applies to CD players. Now we all know that there is a noise floor that increases in SACD and it appears in the Stereophile graphs to increase rapidly above 8 kHz. This is part of the design.
It is still not the case that the noise shaping shapes the signal. Frank's logic, which was not true for CD, is no more true for SACD.
To test sacd performance with a 1kHz dithered testtone as dsd signal is not logical as dsd should be able to cope with an -90dB testtone at 1kHz on it's own. If dither on a -90dB testtone is needed for good testresults with dsd then sacd would be in big trouble.I think the dither is applied to the 16bit pcm test signal.
And of course the noise is in the dsd signal. And my point is that the dsd noisefloor in that signal is modulated by the signal.
The digital black is just a dsd signal with a signal just high enough to keep the muting circuitry from activating. In effect this is the bottom noise floor of dsd (on that partical player).The question is where the test signal is derived from. From a signalgenerator with noise in it's output or is it a math derived 'ideal' testsignal. In the latter case the noise level should be the same as the bottom noise curve.
If it's from a signal generator the rise in noise level should be the same on different machines asuming the same test disc is used.
Frank,
I'm pretty sure it is the same test signal JA uses for the linearity test.
Except that the SACD players clearly show less dynamic range than CD players above 10 kHz or so, and that clearly is a result of noise shaping - as JA has often remarked. The question remains: why does the increase in the noise floor show up an octave below the ultrasonic level?
is too low to realize the promised specs.The rise in noisefloor is just showing that the noise shaper is becoming less effective beyond a certain frequency.
It also shows that distortion rises and that resolution is dropping.
As I said I still don't understand how these tests relate to the stated specs of the SACD players. They seem to be a lot worse than would be expected. But I need to look into what the specs really say-it is hard to believe that Sony would quote specs that are so completely out of line with test results (or that Stereophile would not note this discrepancy). I'll try to figure this out. To tell the truth, I've never paid much attention to their graphs until now.
Firstly, as JA has pointed out, very few Hi Rez players can deliver even 18 bits of resolution across the audible spectrum. Secondly, that SACD's resolution and dynamic range is frequency dependent in a way that Sony would rather not discuss. However, this may have no subjective consequences (again according to JA).
I've been digging around the Stereophile charts and the Sony specs. The spec for the SCD-1 is 105 dB dynamic range in the audio band, presumably up to 20 kHz. So I looked at the review of the SCD-1 and here is the link to the graph, not with a 1 kHz digital signal but with digital silence. This shows the rising noise characteristic, but it also shows that it stays below 105 dB all the way up to 20 kHz. It is just below this at 20 kHz. Seems to me this is the way the noise spectrum is designed to look. This would agree with the spec.OK, why does the 777 graph not look like this? I would think it should. I would think that machine is operating out of spec and should be repaired. I don't think the noise is designed to be as high as that shown in the graph.
My assumption would be that the 1 kHz dithered tone is a tougher test, and more indicative of actual resolution.
it's nice to see someone taking the time to establish facts rather than conjecture
input, and have nothing to do with the components being tested. JA is paid by the square foot, so he just keeps printing the same graph.
it's not that funny anymore, and not very becoming either
This is also why Meridian changed over to Analog Devices Multi-bit Delta Sigma Architecture DACs in the 5xx and 8xx series.Regards,
This post is made possible by the generous support of people like you and our sponsors: