2 hour 96 kHz track only showing 31 minutes in Win 7 64-bit

OK… this is a weird one. I originally used iZotope RX’s SRC to upsample a 44.1 kHz track to 96 kHz for editing. But after reopening the file, I noticed it only had the first 31 minutes of a 2:03:00 concert. I thought it was just iZotope, so I opened the file directly in Audacity (2.1.0). The same issue… only 31 minutes of file. I then tried upsampling through Audacity itself, without iZotope’s SRC… the same damn thing. Curiously though… I successfully worked on a 2 hour concert @ 96 kHz a few years ago… but that was in XP 32-bit.

Is there a limit in Windows 7 (64-bit), or am I missing a driver of some sorts?

Any ideas appreciated.

Thanks.

only 31 minutes of file.

Did it play? Was the Duration the only thing wrong with it?

Koz

It plays fine… just the last 92 minutes are physically missing when reopened in a DAW. It plays start to finish before saving… or at least spot checks throughout. Even Media Info shows it as 02:03:14, so it’s all there.

I run a dualboot Win 7 / Mac OS X, which I’m in right now, and even Mac only shows 31 minutes… but that was encoded/resampled in the Win 7 environment… maybe I’ll try a resample in Audacity while in OS X to see if there’s a difference.

It’s fine with the same file resampled to 48 kHz though. Just 96 kHz is giving the issue.


4800:
Screen Shot 2015-07-17 at 2.26.08 PM.png
9600:
Screen Shot 2015-07-17 at 2.22.11 PM.png
Media-Info:
Screen Shot 2015-07-17 at 2.16.34 PM.png

There is a file size limit for WAV files of either 2Gb or 4GB, depending on which spec you read.* With 2 hours at 96kHz, you are over 2GB (assuming 2-channel stereo and 16 bits or more), and if your bit depth is 24 bits or greater, you are over 4GB. The byte-count field in the WAV header probably “rolled-over” giving a value that seems random.

Try FLAC (if Rx supports it), or BWF, or RF64, whatever Rx and Audacity with FFMPEG both support.



\

  • The byte count (file size) field in the WAV file header is 32-bits, which means you can “count to” 2GB with signed integers or 4GB with unsigned integers. There’s no logical reason for signed integers since you can’t have a negative file size, but from what I understand the original Microsoft WAV spec said 2GB.

I thought Broadcast WAV retained the old file size limits for maximum compatibility. It just added all the tags and metadata header INFO missing from the original WAV standard.

Koz

Thanks, DVDdoug. Turns out it’s 4gb’s. But the 2hr4m 9600 file is only 5.6gb… shouldn’t I be getting more than 31 minutes? It’s only 25% over limit, but it’s affecting 75% of the file.

Either way… I guess I’ll be working at 4800 instead. Either that, or split the individual songs, then process separately. Didn’t want to do that, as it’s a concert.

I can’t understand why I was successful a few years ago. It was the Metallica set from Woodstock '99… a full 2hr set if I remember correctly. I even had it resampled to 192 kHz at one point, just messing around. I swear it was 10gb’s or so. Maybe it was split up, ands I just don’t remember.

There’s 2 ways of upsampling… changing the header to run at a certain speed, or… actually resampling to a higher speed, which results in more headroom, as the noise floor is a lot lower. You can see the difference in spectral view. The 9600 file will have a huge black space at the top of the spectrum, while a 44.1 file will have virtually none.

shouldn’t I be getting more than 31 minutes?

When you exceed the capacity of a file system, you don’t automatically get more or different rules. You get a damaged recording with ambiguous rules.

Koz

That makes sense. That would be why a 8820 kHz file is showing less than the 9600 file… it’s trying to make sense of something it’s not capable of handling in the first place.

Thanks. :slight_smile:

I’m wrong on this… sorry. The black space is a higher frequency range now available. 44.1 kHz cd’s top out at 22.5 kHz in the EQ, while 96 kHz allows up to 48 kHz. Or something like that… :stuck_out_tongue:

I just know upsampling gives better results when doing something like EQ matching.

ok… memory coming back. Last time I worked with 96000 Hz files, I did split the concert… I was playing with a DVD-Audio authoring app, so I needed to split the songs. :stuck_out_tongue:

I wound up splitting this show at the intermission, then upsampled to 96k. I just remember having to reset my filters when jumping from 1 file to the next sometimes. I’ll just save my setting as a preset before moving from the 1st half to the 2nd.

Now that I rember about DVD-Audio… I’m thinking of outputting at 24/96 in the end. The cymbals just sound so much better. :slight_smile:

44.1 kHz cd’s top out at 22.5 kHz in the EQ, while 96 kHz allows up to 48 kHz. Or something like that…

…I’m thinking of outputting at 24/96 in the end. The cymbals just sound so much better.

If you want to go 24/96, I’m not trying to stop you, but…

You can’t hear to 48kHz, or or even 22kHz. Even if you can hear up to 20kHz or so in a hearing test, hearing tests are done with pure tones against a dead-silent background (and they are heard faintly because our sensitivity drops-off as we approach the upper frequency limits of our hearing). In the context of music, these highest frequency sounds are masked (drowned-out) by lower frequency sounds (not so much by low-frequency sounds, but by low_er_ frequency high-frequency sounds).

Upsampling alone doesn’t add any information. You can use a harmonic exciter to add high frequency information, but a good upsampler should not alter the sound at all!!! It’s like copying a VHS tape to DVD or Blu-Ray… You still get VHS quality (or maybe worse with the extra analog-to-digital and digital-to-analog conversions).

The guys at the [u]HydrogenAudio Forum[/u] who’ve done scientific, level-matched, blind ABX tests have pretty-well demonstrated that there’s no audible difference between a “high resolution” original and a copy downsampled to 16/44.1. “CD quality” 16/44.1 is better than human hearing… If the audio is already better than human hearing, higher resolution can’t sound better.

There is information on the HydrogenAudio site if you want to do your own ABX test. But, you have to commit some time to it because multiple trials are required to get a statistically significant result. (But, if the difference is obvious with a quick listen, you can do multiple trials very quickly.)

Another analogy would be watching a (regular size) TV from across a football field… Increasing the resolution makes no difference because our eyes are not that sharp at that distance.

I just know upsampling gives better results when doing something like EQ matching…

That could be the case with some EQ’s but that shouldn’t be the case. And if a higher sample rate makes a better EQ, the DSP programmer should upsample, process, and downsample automatically as part of the algorithm so the user doesn’t have to worry about it.