Upsampling CDs - Storage space and format

From time to time I find it useful to upsample ripped CDs - predominantly such ones where mixing and mastering has done on 24bit/96kHz or so level, but the available version on the market is only 16bit/44.1kHz. Those tuned rips in my experience do sound better and provide less noise when stored in 24bit/305,8kHz flac.
The additional advantage to go up to 24bit/705,6kHz might be limited, but I tend to hear a bit of further improvement. Unfortunately I didn’t succeed in storing those files in flac format - the error message tells me the flac encoder couldn’t be activated, status 11. wav is possible, but needs huge amounts of storage space which probably makes this approach not really worth the effort.
As I didn’t succed to find something here in the forum so far, my question is whether there is a workaround for this problem? May I otherwise suggest that you add a respective feature to future versions of audacity? Also, I would be interested in sharing others’ experiences with such or similar upsampling approaches.
Thanks for kind replies!

If you were alive when CDs were introduced,
then you will now only be able to hear up to ~14kHz

i.e. for you a sample-rate of 28kHz is plenty.

Resampling can introduce artifacts, which can sometimes be pleasing.
Re-mastering the CD rip could also make it sound subjectively better.

But using a sample-rate of 700kHz is 20x more than you need.

Of course I’m aware of that fundamental physiological limitation. Therefore, I’m not after extending the frequency range to be heard, but want to avoid aliasing and filtering problems as well as lower SNR by spreading white noise across a broader and largely not to be heared frequency range.

Does your DAC or soundcard support these extreme sample rates? (Windows will happily/secretly re-sample and play them on any-old cheap soundcard/DAC.)

What is a blind ABX Test?

“CD quality” is generally better than human hearing, especially with music or other regular program material.

Sometimes you can’t even hear the difference between a high-bitrate MP3 (lossy compression) and a high-resolution original in a proper scientific, blind, ABX test. Or, you may have to listen very-carefully and compare to hear the difference.

There are 8-bits in a byte so you can calculate the file size of uncompressed audio as :
Bit-depth/8 x Sample Rate x Number of channels x Playing Time in seconds.
For CDs that’s (16/8) x 44,.kHz x 2 =176kB per second (about 10MB per minute, or a bitrate of 1411kbps).

FLAC is typically a little more than half of that. But if you up-sample to 24-bits (without dither) and leave the sample rate the same, no more information is added (the additional bits are filled with zeros) and the 24-bit file will be about the same size as the 16-bit file. WAV will store the useless zero-bits but FLAC is “smarter”.

(Dither is added noise that’s supposed to sound better than quantization noise and you’re supposed to dither noise when you downsample, not when you upsample.)

If you up-sample the sample rate, everything is interpolated and although there is no more useful information, the additional 8-bits are filled with non-zero data and the FLAC will be larger.

The filtered output from a DAC is continuous analog so it’s essentially “infinitely interpolated” and any intermediate interpolation doesn’t add anything. (Not all DACs/soundcards are filtered and you can’t hear that either, but that’s another story.)

This FAQ says:

Upsampling does not create or remove aliasing. And your CDs shouldn’t have any aliasing. Aliasing happens when you sample audio above the Nyquist limit (a signal over half of the sample rate) or when you downsample. But every audio ADC has an anti-aliasing filter and every proper downsampling algorithm includes anti-aliasing filtering. (Every DSP programmer knows that.)

If there is noise in the recording upsampling won’t remove it. And, you wouldn’t WANT to remove or reduce any sounds from the original recording (unless you want to apply noise reduction or other processing).

You can experiment with an 8-bit file. Upsampling won’t remove the quantization noise.

There is quantization noise that’s worse at lower bit-depts. You can hear it at 8-bits (it’s what low-resolution audio sounds like). At 16-bits you can’t hear quantization noise or dither noise under any normal-practical listening conditions.

Thanks for your kind thoughts indeed. Yes, my DAC would support such a rate, but as I learned from your comments it seems to be impossible with flac-type files.
Regardless of your arguments, I think I can hear quite a bit of noise reduction: My most prominent example is Beethovens Piano Sonatas performed by Wilhelm Kempff. Naturally, this was a deeply analog recording and I bought a 24/96 version with a bit of white noise in the background. Upsampling it to 192kHz and even further to 384kHz definitely improves the noise level. Same applies for remastered CDs of older analog recordings and - less obvious - to (some) very recent DDD-CD recordings.
I somewhere found the explanation (unfortunately lost the reference) that doubling the frequency range by upsampling leads to a factor of squareroot(2) in (white!) noise. As only a small portion of the noise will affect the spectrum we can hear, there in effect is a noise reduction to be perceived by human ears.

Although the philosophy of high-fidelity is to accurately reproduce the recording, there are effects that can “enhance” the sound. You can make a MUCH more dramatic/obvious change than re-sampling, depending on what you do and how far you go with it.

Equalization is super-common and it can be used to enhance the sound to your taste (boost the bass or highs, etc.) or as a corrective effect (to some extent) to make-up for speaker/room weaknesses or to fix-up bad recordings.

There are various “harmonic exciter” effects (derived from the old Aphix Aural Exciter) that add high-frequency harmonics. EQ can only boost what’s there, but these you can add highs that aren’t in the original, even into the ultrasonic range (usually not a good idea, but you could do it). Enhancer.ny is an optional Audacity plug-in.

Or you can “go crazy” and add reverb or something. I use a “hall” or “theater” setting on my home theater receiver to add some delayed reverb to the rear speakers.)

Or, different speakers/headphones will always sound different (better or worse).

Or, you can get a measurement microphone for about $100 USD, and there is free software to measure your speakers/room. Then you can correct/improve with EQ and/or acoustic treatment.

Thanks for your thoughts again. As I’m listening largely to Classicals like Beethoven or Bruckner symphonies, I’m very much after a ‘natural’ reproduction of the recording and don’t want to apply an ‘artificial’ effects.

Of course the room can be a problem and here I have (carefully) applied some of the built-in filters of my speakers to correct for asymmetries in the listening room layout.

Looking at the type of music I’m playing predominantly, the main challenge in my experience is to reproduce (or at least come as close as possible to) the sound of a large orchestra playing in a traditional concert hall.

For simple physical reasons no living room ever will be able to provide this listening experience close to perfection – so my approach was to also have a closer look at headphones. I came across Crosszone and the result of working with those very non-traditional headphones in that respect convinces me more than listening to my speakers.