DSD and aliasing?
I gave up on SACD a long time ago, mostly because the multi-format players I owned (Esoteric X5, Sony 9000 & 5400) didn’t do that well with redbook imo. The proponents of SACD and high rez seem to all say the same thing - that high rez has more “air” and I was wondering if this was actually due to aliasing. Audio science review and I believe Goldensound both measured high rez and the measurements showed a large amount of ultrasonic energy in the music, well above what appears to be the audible range. I was very involved in synthesizer and drum sampling and programming in the early 80s. I recall that I, as well as the people I was working for, often preferred the sample sounds without the anti-aliasing filters because of the shimmer it seemed to put on the samples. Much like tape hiss. When we would mix the 24 track down to 15 ips 2 track, the subtle layer of hiss actually improved the sound in some ways. It seems people like some noise mixed in with their music. So I wonder if the alleged “air” in SACD is actually a small amount of ultrasonic aliasing reaching down into the audible range, if that’s even possible.
It's interesting you point out the players you were listening to were bad at Redbook. I've heard DAC's go through a marked change before/after about 2005 in their PCM performance. Older DAC's just really sucked at low res playback while new DAC's play Redbook almost as well as high resolution files. The slight difference in playback quality now between 44/16 and 96/24 can be found IMHO just by looking at the frequency response, without adding extra shimmer. |
I guess my post wasn’t clearly stated. It was not meant to be about whether SACD players can do good redbook. They obviously can, although some designers, Audio Note among them, believe that multi-function transports cannot extract as much from redbook than a dedicated CDs transport. Be that as it may. What I was asking was whether the “air” that people seem to consistently attribute to high rez, is actually ultrasonic digital aliasing creeping into the audible range. Given my past experiences with digital sampling, I found that people sometimes preferred having a bit of noise mixed in on top of the music. I don’t expect anyone to have a definitive answer, but it seemed interesting after seeing the amount of ultrasonic energy present in DSD recordings. |
The higher the recording sample frequency the higher the low pass filter can be. DSD does not need a low pass filter. A low pass filter causes a phase shift at frequencies that may confuse the playback sound. Some DACs are better at Redbook playback than others, exactly why can be complicated and secret. |
Just a technical note - the ultrasonic noise is quantization noise from the 1 bit sample size, not noise from aliased images.
From some of the graphs I've seen of the noise shaping for DSD64 it starts to rise right around 25 kHz. Nobody is going to hear that as much as they try.
Here's a way to test the upper limit of your hearing - Audiocheck Extended Online High Frequency Hearing Test. Note that you'll experience aliasing if you use a NOS DAC or a DAC with a 'leaky' low pass filter.
Here's a test to see what upper frequency limit you can reliably hear - Audiocheck Blind Testing a 16 kHz Upper Hearing Limit. You can change the target frequency using the links at the top of the page.
If you're curious, check out this example of dynamic range, dither, and noise shaping. |
