sorry to belabour this discussion but i found this here
http://www.sa-cd.net/showthread/18108//y?page=first
if these are observations are factual, i don't understand why the inherent problem is not a 'problem' to some people, including MEs such as Bruce Brown. i'm truly confused. is this HF noise caveat just negative propaganda? if it isn't, why isn't this noise audible or at least measurable?
Quote:
Post by DAD-Digital Audio De December 20, 2006 (3 of 22)
reply
Dear Claude
Thanks for telling me. I was not aware of that, however the main reason for my Tread is not the Mozart recording itself. My issue is to bring some expert knowledge to this forum about the technology used for SACD production.
We are the manufacture and developer of the Sphynx 2 converter (and our own AX24 converter) used for the Mozart recording. Together with Merging Technologies we have developed a solution for DXD recordings.
As you can see I have recommended everybody to read:
http://www.digitalaudio.dk/technical...ion%20v3.5.pdf
This article is explaining the benefits of DXD for SACD production.
Allow me a short and popular explanation of the reasons behind our technologies:
1. In a multi bit (PCM) A/D audio converter all frequencies is mirrored around half sampling rate.
2. As a consequence multi bit (PCM) audio sampling can not reproduce higher frequencies than half the sample rate.
3. The mirrored frequencies will loose their harmonic relationship to the original signal. Therefore and A/D converter has an Anti Aliasing Filter, typically a filter starting at 45 % of the sample rate with full attenuation at 55% of the sample rate.
4. A perfect square pulse has an unlimited frequency band, however with less amplitude of the higher frequencies. (A perfect square sound pulse do not exist in nature however many type of attacks will contain parts which is close to a square pulse when analyzed in the analog domain).
5. Many claims that a sampling rate at 192 kHz/24 bit should be enough (can reproduce frequencies up to 96 kHz), however at 192 kHz sampling rate an Anti Aliasing Filter is still needed.
6. At DXD 352.8 kHz/24 bit (or 384 kHz) an anti aliasing filter is not needed since the frequencies at more than 176.4 kHz are very weak by nature.
7. A downsides of the anti aliasing filter is that some energy is lost in pre/post ringing. A smooth filter will give a better impulse response a less ringing than a stiff filter.
8. When analyzing a 3us perfect pulse we can reproduce 49% of the amplitude at 192 kHz and 88 % of the amplitude at 352.8 kHz (due to the lack of anti aliasing filter and the wider frequency band).
9. The impulse response is very important, since the brain is using the small differences in time from one ear to the other ears in order to re-calculate an image of the room.
Therefore a digital recording at 352.8 kHz (DXD) is sounding real analog. You are able to capture the ambience around the instruments at 352.8 kHz.
192 kHz, 96 kHz and 44.1 kHz/24 bit are all sounding digital. Of course the higher resolution is better that the lower resolution, but these formats are all sounding digital.
10. DSD is a one bit format (SACD format) and do not have an anti aliasing filter. It has a band wide up to 1.3 MHz and can therefore reproduce the amplitude of a perfect pulse 100%.
11. The downsides of DSD is that the format can not be edited since it is a 1 bit format, and the quantizes noise of the format is significant (-80 dBfs without noise shapers)
12. With a noise shaper we have been able to move some of the noise to a higher frequency band; however the energy will always be there. We can keep the noise below -120dBfs up to 24 kHz but then the noise will increase. At 100 kHz we have -22dBfs noise in our DSD implementation.
13. If you need to edit a DSD file you have to convert it to some kind of multibit format.
When you then again want the DSD format you will ad quantizes noise once again.
14. Therefore DSD (SACD) is a consumer format. It is very good if you only ad quantizes noise once, due to the perfect impulse response, but it should not be used for production.
Best regards,
Peter
|