|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
71.195.153.197
In Reply to: Chuck Ainlay , engineer for Dire Straits and Mark Knopfler , compares PCM and DSD , and he prefers ... PCM . ... posted by ZS KEKL on November 22, 2005 at 18:05:29:
...PCM recording/playback, he is more responsive to 96/24 pcm because it still contains it at a slightly, barely audible level.The "softening" that many find objectionable to in DSD is actually closer to the sound of live acoustic music, IMO.
But no matter, both systems sound excellent compared to CD, and both offer multichannel which is the really big difference.
Harry
Follow Ups:
the equipment he was using is reacting against the ultrasonic noise present in DSD, creating intermodulation artefacts in the audible band.there's been some discussion about this in various forums - and quite a lot of studio equipment (not to mention consumer equipment) is apparently sensitive to out of band noise.
one solution is to filter out the out of band noise during monitoring but retain it in the recording.
this problem is not just in DSD, it's inherent in any sigma delta ADC design. Usually, for PCM, either the ADC applies a filter or it's applied a bit later in the processing chain.
It strikes me that having to deal with noise in either a recording or playback chain that is due to a "flaw" in the recording format itself makes very little sense.
where i stated "this problem is not just in DSD, it's inherent in any sigma delta ADC design."in other words, 99% of all PCM ADCs in use today.
If you like, I can supply you references (links to various papers and application notes) published by TI/BB, Cirrus Logic etc.
There's a recent AN from BB that specifically talks about this noise and what to do with it. The most common solution appears to be to apply noise shaping plus a low pass filter that is the inverse of the noise curve in the ADC (in other words, frequency response above 20kHz will therefore be attenuated).
The only ADCs that do not suffer from ultrasonic noise are the resistive ladder type designs which unfortunately suffer from low level linearity issues.
that you never mention is that these DACs generate noise, but there is no signal whatsover at these frequencies so that the noise can be removed with no worries. And as you say, it is.With DSD, the high frequencies by design carry signal as well as noise, where S/N declines dramatically and eventually falls to zero, but somewhere above 50 kHz. Presumably, this high frequency signal is audibly important, otherwise why would it be designed into the system. Therefore, filtering higher frequencies costs information in DSD and is not a good approach.
not correct. sigma delta modulation generates quantization noise throughout the whole frequency spectrum. noise shaping in the converter then determines where to shift this noise. so your explanation is overly simplistic.*** Presumably, this high frequency signal is audibly important, otherwise why would it be designed into the system. ***
I think you presume too much. From what I've read, Sony/Philips seems to be implying that the advantage of DSD is the avoidance of decimation filter in the A/D stage, and the avoidance of oversampling or brick wall filters in the D/A stage. As far as I know, no one has conclusively proven that reproduction of frequencies beyond the range of human hearing has any benefit. Indeed, many experiments, both formal and informal, suggest otherwise (for example, read Griesinger's article).
Although it's not about frequencies outside human hearing range.Filtering the DSD bitstream after da conversion leads to less noise and higher in band resolution at the cost of lower bandwidth.
However every 1bit digtal system stores information about the signal across it's entire passband.
Filtering throws away information.
The trick is that according to the Philips engineers this lost information falls outside the human hearing capabilities.
(Frequency range and amplitude range)
.
When you mentioned "ultrasonic noise present in DSD", I thought you were talking about the rise in HF noise that results from the 1 bit noise shaping, the noise that players have filters designed to cut to prevent downstream components having problems, not converter noise.Every comparison I read between DVDA and SACD talks of the ultrasonic noise of DSD as against the linear response of PCM, is this not the case, or are you talking about something different?
*** When you mentioned "ultrasonic noise present in DSD", I thought you were talking about the rise in HF noise that results from the 1 bit noise shaping, the noise that players have filters designed to cut to prevent downstream components having problems, not converter noise. ***It's actually the same thing. Again, read the application notes from BB and CL. A typical PCM sigma delta ADC does noise shaping on the 1 bit modulated signal, resulting in an ultrasonic noise curve very similar to DSD (I think the latest BB app note even have figures of theoretical curve vs actual measured curve).
*** Every comparison I read between DVDA and SACD talks of the ultrasonic noise of DSD as against the linear response of PCM, is this not the case, or are you talking about something different? ***
PCM derived from a sigma delta ADC does NOT have a linear response. Check actual response curves of real world ADCs, they are NOT flat, because the ultrasonic noise have been filtered away. And if the ultrasonic noise is NOT filtered, then PCM will also have the same noise characteristics as DSD.
PS - Ultrasonic noise is also generated by delta sigma DACs. Again, check real world response curves of actual DVD-Audio players - typically they are NOT flat above 20kHz. For example, on my panasonic DVD-S97, the response is down -0.5dB at 20kHz and rapidly declines after that - again, this is to offset out of band noise generated by the DAC. Hopefully you'll see the actual curve once I publish the review of that player (along with a review of a modded Marantz DV-6500 - that one has an even steeper curve)
The 'very similar ultrasonic noise curve' might look very similar but you fail to mention that the curve rises higher up in the frequency range. A simple analog filter after the dac output easily takes care of this.Note also that this isn't a fault in the pcm storage format. This adc noise floor is generated after the recording and has nothing to do with pcm.
In stark contrast is that with dsd it is a drawback because the generated noise is captured in the recording.
You are dissing up certian player related implementation issues to wrongfully 'prove' imaginary flaws in the pcm storage format.By your logic that would make an additional dsd flaw because some players convert dsd to pcm before the dac.
.
and bent theories about digital technologies.Your 'expertise' is coming from a Sony Philips marketing blog and misinterpreted datasheets.
... it is you who is spreading "wrong assumptions" and "bent theories"?I could refute your points, but "frankly", why bother? It's the greater fool who tries to argue with one.
But let me ask a simple question: who is more likely to be right? My statements are backed by application notes and data sheets from multiple manufacturers, and supported by cold, hard, measured data (which anyone can replicate). So far, you haven't provided any substantiation for any of your tortured rationalizations and assertions.
It's just you misinterpreting them.You fail to mention that the 1 bit adc technologie used for many PCM recordings could have the rising noise floor due to noise shaping well beyond the upper frequency limit of let's say a 24/96 recording.
You keep pointing out outdated technolgies like brickwall filters and imply that these are PCM drawbacks.Your "cold, hard, measured data" should clearly show you the disadvantage of DSD.
*You* want to make this into a PCM vs DSD discussion, but I don't.PCM derived from sigma delta (which by the way is 99% of PCM recordings these days) and DSD share many characteristics - good and bad - because fundamentally they are based on the same underlying technology and principles. That's all I'm trying to say.
If you want to turn this into a discussion about a value judgement about the two, then i hope you'll excuse me if I don't join in.
This is link to the BB paper describing out of band noise for PCM delta sigma DACS:http://focus.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=slea048&fileType=pdf
Notice how similar the noise curve is compared to DSD. It's because they are both based on the same underlying converter technology.
As for the noise curve and application of noise shaping on a sigma delta ADC, refer to this oldie but goodie from Analog Devices:
http://www.analog.com/UploadedFiles/Application_Notes/432842357460935458165408691AN388.pdf
I think Cirrus Logic has written a really good paper too, but I'll leave it to you to search for it.
Even Analog tape which is free of Digital artifacts but has tape hiss, 60Hz hum. pre-echo, crosstalk and other physical artifacts many of which cannot be filtered out without hurting the music.I do not believe we will live to see this perfect format you describe which is free of all noises or artifacts.
The engineers who work with PCM, DSD and Analog on the most part do a very good to excellent job of elimating these non-musical noises which plague our formats.
And formats don't have noise and artifacts, but the medium is susceptible to both of these annoyances. Be careful with your terminology.But yes, analog tape recorded with audio does have its issues.
...and the flaws must be dealt with in either the the recording or playback chain (or both). That is reality. You pick what you're willing to deal with, and live with it.
This post is made possible by the generous support of people like you and our sponsors: