“A lot”? Come on … it’s below -120dBFS. There’s no way you can hear this noise. Simply not possible. Yes, it’s more than with some other SR converters, but still not “a lot”. I would rather point out Cubase’s relatively poor anti-alias filter. That’s something you MAY hear with SOME material if you have extreamly suberb ears. OTOH, gradual filter slope gives you good impulse response.
This site is great, if you know what you are looking at and understand the graphs. But looking at the graph without any understanding about what it tells you and crying out: A LOT OF NOISE takes you nowhere.
Reminds me of a Monty Python skit…
“…if we increase the size of the penguin until it is the same size as a man, we find that it’s brain is still smaller - BUT! and this is the point, it’s larger than it was…”
I’m wondering if that was not a pure sine wave that they used. The peaks at what you are perceiving as noise seem to follow the overtone series. Could they be overtones of the sound source? If this is the case, then recording at 96K picks up more of the overtones and translates them back when passing to 44.1K. That could be a good thing, especially if recording the same sound at 44.1K does not produce the same overtones.
Recording at 44.1k gives you a frequency response up to 22K, but recording at 96K gives you a frequency response up to 48K. If even some of this response can translate back to 44.1k, the resulting CD or audio file should be brighter and cleaner sounding than it would be if all of the tracks had been recorded at 44.1K. I think this is what we could be seeing in that graph.
I recently began recording my tracks at 96k 24 bit, and my finished mix was done at 44.1K 16 bit. I did notice a slightly brighter and cleaner sound in the finished product than when I recorded my tracks at 44.1 16 bit.
As soon as I get a computer system that is capable, I am going to start recording at 192K. That will give me a frequency response up to 96K. I know that will be something that only dogs can hear, but although our hearing is limited to 20K, anything above that can be felt if not heard. I believe that it is that feeling that influences how a listener perceives a piece of music.
There have been a few studies on the subject, and none of them have shown that higher sample rates are “felt” or perceived in any way. If there IS a perception, it’s probably due to perception bias.
Also, there’s no such thing as a pure or impure sine wave, and that tone they used doesn’t have any overtones. The additional content you’re seeing in the graph is all noise produced within whatever device(s) is being used. Many have asserted over the years that using higher sample rates in order to be able to capture and reproduce higher frequency overtones above Nyquist = fs/2 has a noticeable effect on the audible band, but again, there’s no evidence of this, and in fact it contradicts the basic laws of physics; and in any case, if the overtones did have a noticeable effect, that effect will be captured and reproduced accurately, given fs > 2F
You obviously have more knowledge on the subject than I do, so my question to you would be “Why then should we even consider recording at 96K or higher, or even as low as 48K for that matter, when 44.1K can reproduce frequencies up to 22K and we can’t hear anything over 20K?” Is this just for the purpose of taking up unnecessary space on our hard drives, or is there a legitimate reason for these higher resolutions?
To answer your question, the one argument for using higher sample rates that I find tenable is that it allows for much gentler filtering, which may have a beneficial effect on the audible range. Another good argument (but which I admit is a bit over my head) is that using higher sample rates will minimize peak calculation errors, which at certain frequencies can be as much as +/-3dB @ 44.1kHz.
I used to make the argument (because I read Dan Lavry saying it ) that as one goes up to really high sample rates, like 192kHz, you start to lose accuracy, because these after all are physical machines that have real-world physical limits, and the op-amps etc. can only take measurements so fast, until at one point each measurement becomes less and less accurate since the machine is operating so quickly. Now, I’m not so sure about that argument, since there are medical devices that use sample rates that are exponentially higher than the one we use for audio, and they do this with great precision.
Once upon a time, I was a crusader on this subject. But now for me it’s so much pedantry. Nobody, not even the most ardent audiophiles, ever listened to a song and thought, “That would sound much better if they had used a higher sample rate.” Because none of it really matters if your song sucks