|
Propeller Head Plaza Technical and scientific discussion of amps, cables and other topics. |
|
|
|
Propeller Head Plaza Technical and scientific discussion of amps, cables and other topics. |
|
In Reply to: "I've found the depth of image to be largely a function of the recording, especially the capture techniques. " posted by kurt s on March 1, 2011 at 18:43:37:
All the best ADCs today are using very high (megaHertz) sampling rates, doing digital filtering and then decimating down to the desired sampling rate. When I was using a real cheap consumer ADC the results at 44.1 kHz were pretty horrible, but running it at a higher sampling rate and then using (much higher quality) software filtering produced much better results. The same thing applies in the reverse direction, with a cheap DAC there is much to be gained by upsampling in the computer and sending the higher resolution digital signal to the DAC.
However, using the very best available sample rate converters and starting with high resolution digital, one can observe that a high sampling rate signal can not be down converted to 44.1 kHz and then converted back to the original format without audible degradation. This is true even if all of the signal processing is extremely accurate (e.g. 64 bit floating point) and for a huge range of possible filters. There is always some kind of a problem if the original music had any significant high frequency content. If one uses a narrow filter it rings. If one uses a wide filter then if it is set too high there is significant aliasing distortion. If one uses a wide filter and it is set too low there is significant dulling of the upper octave. The problem is related to the width of the filter banks in the inner ear, if the digital filters have a longer time constant their ringing is going to affect the imaging. If you record at 192/24 format, then most of these problems have gone away. However, those with access to first rate equipment that runs at even higher sampling rates report better results at even higher rates, e.g. 384/24. Keith Johnson was running at twice that rate.
Your comment about AM distortion is well known to the designers of audio compression equipment and limiters. There the needed amplitude modulation increases the bandwidth with resulting aliasing problems. The better equipment first upsamples the input to a higher rate, removes any images created by the upsampling, does the required modulation, refilters down below Nyquist and then decimates back to the original sample rate. Unfortunately, this isn't the best way to do things if there are going to be multiple stages of digital processing as the filtering associated with each sample rate conversion imposes a certain amount of degradation. This can be avoided by using analog equipment to perform processing (at the cost of noise and distortion), or by keeping the entire digital signal at a very high resolution through all desired processing, and reducing to the final CD format only as the last step. (This is what I do with the cassette tape restorations that I work on.)
When playing back 44.1 kHz material the optimum filtering (unfortunately) depends on the original recording. If the original material was properly recorded it has no energy close to 22050 Hz, perhaps none above 21000 Hz. In that event the best results come with a brick wall filter. One can have as narrow a transition bandwidth if one wants, even 1 Hz, and there will be no ringing associated with such a reconstruction filter if there was no energy in the input waveform. However, most CDs are not properly filtered, so then one needs to use more forgiving filters. If one creates test waveforms it is clearly possible to see the interaction between record and playback filters associated with sample rate conversion. I use iZotope RX 2 Advanced for sample rate conversion and Soundforge Pro 10 for generating various test signals. The theory and practice match up quite well.
It is important to remove unnecessary ultrasonic noise, whether images or otherwise, otherwise it will reappear in the sound. Even if the noise isn't directly audible it still intermodulates with the audible signal and the resulting difference frequencies can be heard. I suspect this is one of the reasons why people have commented that material upsampled from 44.1 to 96 kHz don't sound as good as upsampled from 44.1 to 88.2 kHz, namely there are more possible combinations of beat tones to offend. If the filtering has been done sufficiently well then this won't happen, but there is always the question of how much is "sufficient".
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- RE: "I've found the depth of image to be largely a function of the recording, especially the capture techniques. " - Tony Lauck 10:21:14 03/02/11 (1)
- A very informative look at what you do behind the scenes on the pro audio side - kurt s 17:15:42 03/02/11 (0)
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2024, Audio Asylum, all rights reserved ]
