there are no bit depth options with JACK. It uses the largest bit depth the hardware will support when reading and writing to/from the hardware. It also offers the option to dither when converting back to hardware format, since the internal format used by all JACK clients (and almost all audio software in this day and age) is 32 bit float, which has an implicit 24 bit depth. You do not dither when converting to a deeper bit format, only when reducing the bit depth.
You do not dither when converting to a deeper bit format, only when reducing the bit depth.agreed.
This video provides a good explanation of all things to do with A/D or D/A conversion, including dithering etc
I agree with LinuxDSP on that video, excellent and becoming part of my coursework for my students to watch it and the companion to it.
Seablade@linuxdsp - well, that is what I appeared to notice the one time that I re-imported a file that I’d rendered to 16-bit with noise-shaped dithering. I wouldn’t say it became louder as such, but definitely became audible, in the sense that parts that sound silent when played back on a CD-player had a quality that I’d describe as sounding like “tape hiss” when imported to 24 bits. And it seems to me perfectly reasonable that it should do so, since true digital black is 48dB quieter than the dither level. I don’t think we’re disagreeing about anything, I was just suggesting it as a possible explanation that nobody seemed to have considered for the OP’s observation. And of course, the effect of dithering /is/ noticeable when compared with an un-dithered version of the same file (if not audible as noise or hiss) - otherwise, why would you bother to do it?
@Matt Francomb: I don’t think we fundamentally disagree. Dither would be expected to manifest as (very) low level noise, like tape hiss (as described in the link I posted). I guess I just interpreted your original comments wrongly.
I have the same problem with an mbox1 but i think my problem is clock related. burts of white noise every 5 seconds that last 10-15 seconds.
No combination of settings help for me, digidisign driver support though is non existant and there is only a patch that makes it work well it used to apperently but now only mbox 2 does
Why is it you have to have jack set to the sample rate set to the same sample rate as what your project is running in? I dont remember having to do that back in windows (mind you this was over 10 yeras ago) I have a couple of tracks that were recorded at 48khz but most as 44.1 so opening them up when jack is at the wrong sample rate causes playback speeed to be wrong.
When a DAW does not have JACK what it does is, more or less, load “JACK” when you open a project and close “JACK” when you close the project. Obviously you don’t need to tell it what rate, but you also don’t have JACK’s functionality (software interconnection).
I’m not sure if that’s what you’re getting at, but if you meant that on Windows DAWs you can have files of multiple sample rates in the same project without being asked to resample them, that means the resampling is being done on the fly. That’s unprofessional. It works, but it’s unprofessional and it is so for a variety of reasons. This is correct on playback applications only (ie. media players, PulseAudio).
Dithering /is/ unnoticeable. Most of the time, anyway. We do it out of correctness, as a lot of things are in the multimedia engineering world. The problems here have nothing to do with dithering. If any quality changed when you imported a 16bit file then something is broken, there’s no reason for anything to change. Same as in the case of the interface that works better in 16bit mode - it’s just broken (the driver or the interface).
OK, maybe dithering is unnoticeable, but it’s more than just correctness - lack of dithering is definitely noticeable! I demonstrated this for a client who wanted “as little processing as possible” done on the final mixdown and wanted to hear the difference because she was very skeptical of my explanation that “adding noise” could possibly make it sound better - rendered the same mix twice onto a CD, once with dithering and once without, and let her listen to the two on her stereo without telling her which was which, and she chose the dithered one. I was listening very carefully, and knew which was which and could definitely hear, for example, that where reverb tails or sustained notes faded /to silence/, the un-dithered version starts to “crackle” just before becoming inaudible, whereas the dithered version fades completely smoothly into silence. This is a normal listening levels - turn it up loud and it’s even more noticeable.
@paul - yes, I’ve had a look at that. As I pointed out, the listeners didn’t know which version was which, but preferred the dithered version. Because I knew which was which, I didn’t express an opinion until after they had. As I said before, why bother dithering if it doesn’t have a perceptible effect?
Weiner’s point is that dithering is not perceptible. That means that you can not hear it, and you cannot prefer it. Did you try his listening test?
Like Weiner, I’m not arguing that dithering is wrong. What is being argued is whether anyone can hear the result.
I’m not disputing that (in 16-bit mixes), except by extension. I’m saying that you can definitely hear the /lack/ of it when you don’t do it. I mean you can’t hear it, but you can sometimes hear when it hasn’t been done. This is real hair-splitting, though. I’m going to shut up about it now.
Here is a good place for anyone who wants to understand what dither is and what it does. And, it agrees with Weiner that is is really inaudible at 16 bit. But they have 8-bit examples where you can hear it. Its is the Xiph.org “Digital Show & Tell”.
http://xiph.org/video/vid2.shtml
http://wiki.xiph.org/Videos/Digital_Show_and_Tell
That’s what I posted about earlier in this thread.
@linuxdsp: Ah! you are right! Sorry.
But then perhaps I can point to another xiph.org link where Monty explains how dither extends the effective dynamic range. The famous “24/192 Music Downloads…and why they make no sense” article:
http://people.xiph.org/~xiphmont/demo/neil-young.html
I was going to shut up about this, but Weiner isn’t really making a point (or certainly not the point that in all cases the effect of dither is imperceptible) - he’s taking the piss out of people saying that they can hear the difference in loud modern pop music (i.e. with negligible dynamic range). I’ll upload a couple of snippets some time this evening to show that in music with a significant dynamic range, and the occasional bit of near- or actual silence, at 16 bits, dither makes an appreciable difference. I admit that you can’t hear the dither - what you can hear is truncation noise on the un-dithered version, but the effect is the same… Anyone who listens to the two in a reasonably quiet environment (i.e. not in a car), even if they don’t know what they’re listening for, will be able to tell that one sounds better, and that turns out to be the dithered version. It’s not a question of “golden” or superhuman hearing, just that you can’t hear a quiet noise in the presence of a really loud one.
my recollection is that he demonstrates dither on some reasonably dynamic-y “classical” music too, and that there, as with “modern pop”, you can’t hear it.
i didn’t go and download his original files, but i know that when he did his “add godawful screeching noise” test with classical music, he only had to take it down to -70dB for me to be unable to hear it. dither/truncate noise is massively below that. and yes, truncation noise is basically just quite noise. but the point is that it is VERY VERY quiet, so quiet that its not audible not only in the presence of “a really loud” sound, but more or less any sound at all.
i could be wrong 
Sure - I know all that, but have a listen to my files when I get to uploading them. I don’t want to prejudge it, and the difference is not glaring, but if you listen carefully in a quiet environment, the truncated one sounds a little nasty on /very/ quiet bits, and the dithered one just sounds smooth all the way down to silence. As I mentioned before, two listeners (one aged 60 and one aged 19 - so it’s not a frequency-response thing), both found that the dithered version was better (in the sense that there was something wrong with the truncated one) without being told anything beyond the fact that they were different.
First: the white noise resulting from dither when the noise is amplified enough so you can hear it, is far better sounding that the truncation noise resulting when no dither is used, so I can’t see any argument for not using it: - it does no harm at all and in a very quiet listening environment the difference is likely to show up.
Second: some people (Bob Katz springs to mind) say they can not only hear the difference between dithered and undithered conversion, they can even hear the difference between different kinds of dither e.g. flat vs. noise shaped. So dithered audio is always going to be acceptable to more people than undithered, and undithered is NEVER going to be preferred.
Anyway, this is all beside the point - if an Ardour system is making strange noises, it’s NOT dither that’s responsible.
Well, this is embarrassing! I’ve uploaded a couple of snippets, as promised https://soundcloud.com/matt-francomb/a and https://soundcloud.com/matt-francomb/b - I won’t saw which is dithered and which is truncated, but I will say that the difference now seems far subtler to me when played back on a computer than it did when I burned the full tracks onto a CD and listened on a stereo system. I don’t know how to explain that - could be fan noise from the computer, I guess. I still think I can hear a difference, but wouldn’t be confident that I could identify which file is which on a completely blind listening. Maybe the snippets are too short, but I don’t own the copyright and the EP isn’t released yet, so I don’t want to put up anything longer. It will be interesting to see what anyone else thinks, anyway.