24bit vs 16bit, the myth exploded! - Page 77

I guess you listened to that htg podcast. I tried too but closed the window after he said that his special stuff is also full of tubes.. oh yeah and that his studio is the best he's ever seen (or something like that).

Thanks a lot for the help .

Wow, I did not realize this thread was still going. I totally want to be a part of this. Has anyone brought up ripple distortion on the digital reconstruction of the signal? I couldn't read through all the comments but in what I saw, it didn't seem like there was anything mentioned. 

Ripple distortion?

Perhaps he is talking about the Gibbs effect (http://en.wikipedia.org/wiki/Gibbs_effect).

 

It was discussed already, just not called "ripple distortion". Basically, with any reasonable implementation of filtering at 44.1 kHz, the effect is too short and is at too high frequency to be audible. Also, human hearing does not really care about what waveforms "look" like, since it senses bands of frequencies, rather than a simple time domain signal like in a WAV file. There is an example of a real reconstruction filter in a DAC here, which shows that the filtering only significantly affects frequencies above 20 kHz, and that the pre- and post-ringing have a duration of only about +/- 1 ms. Additionally, there were some ABX tests on the forum that could be used to compare (among others) different reconstruction filters, but the positive results are lacking so far.

Or perhaps he's talking about ripple in the frequency domain:

 

 

In which case there's no need to worry either, because the ripple is usually too low to make an audible difference.

xnor, yes that was exactly what I was talking about. The ripple in the passband. I think that is one of te main arguments for 96khz recordings; to try to reduce the amount distortion. It's been a couple of years since my DSP class so I'm trying to refresh myself. 

Frequency response ripple in the passband is only an issue in outdated or very low end DAC chips (such as older onboard codecs, some portable players, etc.). It is easy to reduce to well below the threshold of audibility with a reasonable FIR filter design; very low ripple and very high stopband rejection can be achieved by using the right window function, at the expense of requiring a somewhat longer impulse response. The DAC example linked above does not have any significant ripple either.

Edited by stv014 - 5/13/13 at 6:13am

 

Barry Diament has made similar observations about recording at 24/192 and his Soundkeeper Recordings is built on delivering his hi-res recordings. He has full songs in different resolutions available for download and comparison:

 

http://www.soundkeeperrecordings.com/format.htm

At least one of the 16/44 wavs has a bad header or corrupted wav container so won't play in some players. I downloaded the Maria samples from Americas. I remember the same thing a few months ago with the sample. I can't remember if I tried the other offerings. Anyway if your player chokes on the 16/44 wav you can just "convert" it losslessly to a new wav file which will have identical pcm audio but will actually play in any player/device/app that supports wav. ffmpeg will do it with "ffmpeg -i input.wav -c:a copy output.wav". You could also use foobar2k's convert utility.
Edited by julian67 - 5/13/13 at 9:12am

Not sure if this has been discussed yet or not. Apologies if it has. So, I listened to my first hi-res 24bit/88.2kHz recording the other day. At the time to be honest it sounded pretty normal to me. Then I realised that I was listening to it on my secondary system, which the DAC only has 3 sample rates, 32 kHz, 44 kHz and 48 kHz. I first thought, oh that must be why I can't hear any noticeable difference in quality. What I really should have been wondering was, why is this working on this DAC? Yes the sample rate was a multiple of 2 of 44.1 kHz and indeed that is what the DAC seemed to lock on to (a red LED told me so). But is this it? Did the DAC just average the two bits instead or one? Any explanation on why this worked would be really appreciated. I can see how it might be possible, but assumed the DAC would be more fussy than that.

My guess is you're using Windows with DirectSound. If the sample rates don't match the sound engine will resample it accordingly.

Sorry, I should have explained my setup. I'm streaming the music over ethernet from my music server to a squeezebox, then from the squeezebox to the DAC. I'm pretty sure that the Squeezebox server is not resampling on the fly, but I suppose it could be.