24 bit 48khz 32 bit rendering (Teaching an old dog new tricks)

I'm gonna go with the assumption that something in your test was flawed, because there shouldn't be a difference, especially not one you can hear through the car stereo. 24 bit doesn't add more 'detail' to your sound, it only extends the dynamic range. It's like if you have a garage that is 20ft long and you put a nice car in it. Then you have a 40ft long garage and you put the same car in it. In the 20ft one, you can fit more cars, but at the end of the day, only the car at the roller door at the front is accessible and usable. It's the same car. The same sound. The car at the back you can't get to. Of course this all changes once you talk mixing/recording (compression and increasing noise floor, not clipping etc), you need that extra room to work on the car (the analogy starts falling apart here). Even when you're talking dither. Have you done a test comparing dithering? It's really a very minute difference, in reality. It's so quiet. So far back in the mix. People get so hung up about it when a slight change in an eq can make a much more noticeable difference.

44 vs 48, theoretically we shouldn't hear the difference. If you can hear it (and it is possible), it's due to the filter algorithms involved. This gets mighty complex fast, and I don't claim to be an expert in it. At the end of the day, if the filters are good, double blind tests (run by Bob Katz) don't show a noticeable difference (great argument there haha).

By the sounds of it, you are hearing a big change. I'd suggest looking more closely at your mixdown of these two samples cause there is just no way that the change you are hearing (and I don't deny you are hearing a change) is caused by simply running 24/48.

At the end of the day, if there were a bit of a difference, try not listening to either tracks for a week, then listen to one track. Wait 24 hours. Then listen to the other track. Guess which one was which. You may hear a difference when compared side by side, but no listener does that. They put on a track and listen to it. The differences you're talking are so small, you're not going to have any less of an experience in real world listening.


Just one man's opinion. I'm sure plenty out there will disagree! 

I'm in the same boat as you Rimshot. I can tell the difference between 16/44 and 24/48 on one of my systems and can't tell on the other. Even though the science may totally prove me wrong or maybe it's something with the converters or whatever...I've never failed a blind test on my own material and I've done this soo many times. On someone else's material, I might not be as accurate. But for my stuff, I pick the 16/44 out all the time. Now 24/96 and above, I can't tell a difference on when comparing it to 24/48 unless I record a bunch of acoustic instruments that are non-sonic like electric guitars with over-drive and basses with a DI etc.

Now, I think some of the things we hear that we think are better may be due to converters. For example, when I use my old Layla 24/96 card, I can tell the difference between 16/44 and 24/48. With my RME Fireface, I hear no difference. Then again we're talking Layla converters that came out in what...1999 or whatever vs. the RME converters of today. So that could play a role in things...I don't know for sure. But I definitely hear what you hear exactly the way you described it. I definitely agree with Matt on this...but I can definitely hear what you mentioned on my Layla for sure. And like I say, what I hear on this end is the way you explained it. It's gotta be a converter thing or something.

-Danny

mattplaysguitar


I'm gonna go with the assumption that something in your test was flawed, because there shouldn't be a difference, especially not one you can hear through the car stereo. 24 bit doesn't add more 'detail' to your sound, it only extends the dynamic range. It's like if you have a garage that is 20ft long and you put a nice car in it. Then you have a 40ft long garage and you put the same car in it. In the 20ft one, you can fit more cars, but at the end of the day, only the car at the roller door at the front is accessible and usable. It's the same car. The same sound. The car at the back you can't get to. Of course this all changes once you talk mixing/recording (compression and increasing noise floor, not clipping etc), you need that extra room to work on the car (the analogy starts falling apart here). Even when you're talking dither. Have you done a test comparing dithering? It's really a very minute difference, in reality. It's so quiet. So far back in the mix. People get so hung up about it when a slight change in an eq can make a much more noticeable difference.

To add to this: If you are dealing with properly dithered signals, the only difference you will hear between 16 and 24 bits, even at ridiculous volumes, is added noise. At normal/real-world/won't-get-hearing-damage-in-seconds levels this noise is almost always completely inaudible (for 16bits). The only way to hear it is generally to either use material that doesn't use the most significant of the 16 bits (i.e. lots of headroom or a very few abnormally high level peaks), or wait for a quiet point and crank up the volume. But in either of these cases, you aren't really talking about 16bit audio any more!

If you have audio properly dithered down to 16bits and hear anything other than noise (and in the real world you won't even hear that), it's either in your imagination or due to something other than the bit reduction. If someone tells you otherwise, I'd suggest you find someone else as your source of information regarding how digital audio works.

44 vs 48, theoretically we shouldn't hear the difference. If you can hear it (and it is possible), it's due to the filter algorithms involved. This gets mighty complex fast, and I don't claim to be an expert in it. At the end of the day, if the filters are good, double blind tests (run by Bob Katz) don't show a noticeable difference (great argument there haha).

By the sounds of it, you are hearing a big change. I'd suggest looking more closely at your mixdown of these two samples cause there is just no way that the change you are hearing (and I don't deny you are hearing a change) is caused by simply running 24/48.

At the end of the day, if there were a bit of a difference, try not listening to either tracks for a week, then listen to one track. Wait 24 hours. Then listen to the other track. Guess which one was which. You may hear a difference when compared side by side, but no listener does that. They put on a track and listen to it. The differences you're talking are so small, you're not going to have any less of an experience in real world listening.


Just one man's opinion. I'm sure plenty out there will disagree! 

One way of doing controlled testing of different sample rates is to take a higher sample rate original, downsample it to the lower rate, and finally upsample it back to the original (using a reasonable quality sample rate converter). This results in two tracks both at the same higher sample rate, but one has been downsampled/upsampled. Since they are both running at the same rate, any differences due to the converter (DAC) running at different rates is eliminated, and the only thing remaining is the sample rate conversion (SRC) itself.


And I also should add that unless you do a proper double blind (or ABX) test, any results are severely suspect, no matter how obvious you think they are. It's surprisingly often that people's ability to hear what they consider "obvious differences" go away completely when they really have absolutely no way of knowing what they're listening to beforehand.