Is there any way to prevent dithering noise in WAV files in Audacity 2.0.2? The original 16 bit WAV files I import from my Sony ICDSX1000 recorders have no dithering noise whatsoever, but when I edit and save as a new file, dithering noise appears. I have Audacity set to 32 bit float, and the dithering noise doesn’t appear until I save the file as a 16 bit WAV file. I don’t want to save my edited lossless files in any other format besides WAV or AIFF, because my iPod doesn’t support any other lossless formats. Can anyone please help me? I’m using Windows XP Home Edition.
Go to Edit → Preferences → Quality. There are two boxes where you can set dither to None.
because my iPod doesn’t support any other lossless formats.
Not related to dither, but I believe iPods can play ALAC (lossless compression).
There is a bug in Audacity 2.0.2 in which the default dither settings produce dither that is audible when exporting stereo tracks. You should update to the current version of Audacity which is available here: http://audacityteam.org/download/
Thanks DVDdoug. I got the dithering noise out by setting dither to none in both of the boxes. It was already set to none in the realtime conversion box, but the noise didn’t disappear until I also set it to none in the high quality conversion box.
Steve, I have a fiew questions before updating. How well does the latest version of Audacity work with JAWS compared with 2.0.2? Has the bug where the playback speed button doesn’t work with JAWS been fixed? In the latest version, do I still have to set the sample format to 32 bit float instead of 16 bit to get rid of dithering noise. I’d like to use 16 bit if possible since it saves CPU. Thanks.
I’ve not tried with JAWS, but I’d guess that accessibility for VI users should be improved in 2.0.5. I know there have been fixes for a number of accessibility issues.
I’m not aware of that bug and I don’t see it listed on bugzilla (the bug tracking system).
For both of these issues you would be best to ask on the audacity4blind mailing list (http://www.freelists.org/list/audacity4blind) as there are skilled users on that list that use Audacity with JAWS.
If the “playback speed” issue is a bug in the current version of Audacity, then it needs to be logged on Bugzilla so that it can be fixed, but the first thing is to determine if it is a current bug or not.
Unless you have an extremely old computer (more than 15 years old), processing in 32 bit float format is likely to be more efficient than 16 bit. Modern processors have a built-in FPU that takes care of 32 bit float operations, removing the load from the main CPU. Even if your tracks are 16 bit, all processing is done in 32 bit float format, for reasons of efficiency, precision and sound quality. Working in 16 bit is not generally recommended because effectively the data is converted from 16 to 32 back to 16 bit with every processing operation. This will reduce the overall sound quality (though usually not by a noticeable amount).
16 bit data will save disk space, but that should not be a problem with modern computers, especially considering the (low) cost of hard drives these days.
Thanks, Steve. If I export a file to the same bit depth as I started from with dither turned off, will it do any damage to my audio if I amplify it before exporting?
Theoretically, amplifying with dither turned off will introduce a tiny bit of harmonic distortion due to 16 bit rounding errors. The purpose of dither is to prevent that distortion by adding a little “randomization” to the rounding. In practice the difference between correctly dithered and not dithered at all is very small and virtually inaudible except when playing very quiet audio at high volume.
The problem with dither in older versions of Audacity was that the “shaped” dither, when applied to stereo tracks, was at a higher level than it should have been, so that during quiet sections it was clearly audible. This has been fixed in the current Audacity version.
Thanks again, Steve. I definitely don’t want any distortion introduced into my audio. I just want to get rid of the clearly audible dither hiss that is currently in my files. Are there any advantages of using shaped dither instead of another dither type such as rectangle?
“Dither” provides a trade-off between removing harmonic distortion caused by rounding 16 bit integer data, at the cost of introducing a small amount of hiss. Obviously the less hiss the better.
In Audacity 2.0.5, “shaped” dither prevents harmonic distortion as well as any other dither type, but when played at normal or even quite loud volume, the hiss is noticeably the quietest (virtually inaudible). Shaped dither requires a bit more processing power than other types of dither, so exporting may be slightly slower (but probably not much).
This all begs the question: why are you languishing on the now-obsolete 2.0.2 when 2.0.5 is available (with sevral bugs fixed over 2.0.2) ?
WC
I am not sure if you mean this, but from 2.0.3 onwards you can activate toolbar buttons in Audacity by pressing ENTER on your keyboard.
Gale
I went on ahead and updated to version 2.0.5 last night. I’m a little slower updating programs than most people because I want to make sure that the updated programs are still accessible for blind people like myself before updating. This hasn’t been much of a problem with Audacity, but a lot of times the developers of other programs aren’t so careful to make sure their updates are accessible.
Thanks, Steve. If I export a file to the same bit depth as I started from with dither turned off, will it do any damage to my audio if I amplify it before exporting?
Digital amplification or attenuation less damaging than analog amplification/attenuation. Volume adjustment is the most common thing done by recording, mixing, and mastering engineers, and nobody worries about the teeny-tiny digital rounding errors. A recording may go through several generations of volume adjustments between the time it’s recorded and the time is’s released… It’s just not an issue.
Amplifier circuits (analog) always add some noise. And, if you turn-down your volume control there is usually noise generated in your power-amp (or somewhere after the volume control) and since that noise doesn’t go down when you turn-down the volume, your signal-to-noise ratio is reduced.
Thanks again, Steve. I definitely don’t want any distortion introduced into my audio. I just want to get rid of the clearly audible dither hiss that is currently in my files. Are there any advantages of using shaped dither instead of another dither type such as rectangle?
I’m sure different people have different preferences. I thought you had turned dithering off? Since you are not down-sampling your bit-depth, you shouldn’t be dithering.
The answer is, use whatever sounds best. If it sounds best without it don’t use it at all! But, you shouldn’t hear the dither under any normal/reasonable listening conditions anyway… I believe Steve when he says it was too high in previous versions.
Dither is (or should be) around 90dB down at 16-bits (near the “quite-end” of the 16-bit resolution). If you don’t have a feel for how quiet that is, take one of your files and use the Amplify effect to reduce the volume by 80 or 90dB starting at the 30-second point. Crank up the playback-volume as loud as you can stand it (or for as much distortion as you can stand) and then listen to how quiet it gets at the 30-second mark. You’ll probably just hear the analog noise from your system.
There is still a case that you should dither in a 16-bit project if you are processing the audio by running effects (if you are more worried about harmonic distortion than dither noise).
Shaped dither in a 16-bit 44100 Hz project peaks at about -72 dB as opposed to about - 85 dB for Triangle dither.
Gale
I have Audacity versions 2.0.2 and 2.0.5 installed side by side on my system. Out of curiosity, I did a test on exported WAV files to see which dither type seemed loudest, and to compare the shaped dither on stereo exports in 2.0.2 and 2.0.5. I amplified each file to 50 DB after re-import, and the shaped dither in 2.0.2 is definitely much louder than the shaped dither in 2.0.5. To my ears, in 2.0.5, triangle is louder than shaped, and rectangle is slightly louder than triangle.
I now have my default sample format set to 32 bit float to prevent excess dither noise and loss of sound quality when processing. My machine is a 2006 Acer running Windows XP, and unfortunately, I don’t think it has a dedicated FPU for 32 bit float, because the main CPU usage is higher when I use 32 bit float than it is when I use 16 bit PCM.
A bit of care is needed here.
“Shaped” dither is designed to be (virtually) inaudible, but assumes that playback is at a normal level.
The “shaping” is based on the “equal loudness contour” (http://en.wikipedia.org/wiki/Equal-loudness_contour). Higher frequencies are good at preventing quantize noise and at low “SPL” levels, hearing is not very sensitive to very high frequencies .
The “Equal Loudness” curve changes depending on how loud it is, so if one sound is louder than another when both are played quietly does not necessarily mean that it will sound louder than the other when both are played at an amplified level.
Having said that, I totally agree; the shaped dither in 2.0.2 is louder than the shaped dither in 2.0.5, triangle is louder than shaped, and rectangle is louder than triangle.
I expect that your machine does have an FPU, but your computer is doing more work than just the number crunching. It also has to cope with reading and writing data to/from disk and many other tasks. 32 bit audio has twice as many “bits per second” as 16 bit audio, so disk read/writes are handling double the amount of data. Depending on the computer, that could be putting a lot more work on the CPU.