Enabling 24bit-44.1/48/88.2/96khz playback - iAudio and others?

[preglow]

Did a resampling of the Joni Mitchell album Blue, to 16 bit 44.1 khz, in foobar and I agree . It is accualy better on the sansa now.

No surprises involved. If you have music files you know you're only going to be listening to on a current Rockbox, you should resample them to 44.1 kHz on a computer. Even if we do get a good resampler some day, the kind of resampling you can do offline on a computer will always be better, since we need to do our resampling realtime. I believe some of our targets do support 96 kHz playback in hardware though, so when we enable support for that, you might be interested in keeping them at the original sample rate. I kind of cringe every time I have to call our current deal a resampler at all. It's just a linear interpolator and is basically the the second worst resampler there is.

[saratoga]

I think that the quantization noise of the 12 bit recording will be easily noticeable and distinguishable from the ambient noise.

I don't if the ambient is much louder then 30dB, which it will in virtually all cases.

Please do not mix quantization and ambient noise. They are very different things. The ambient noise is a part of the recording (EDIT: which is usually desired).

Of course, however, the ambient noise masks the quantization noise.  In general, to hear quantization noise, it cannot be masked since properly dithered quant noise is not easy to pick out of other sounds (by design).  This is the principle that mp3/ogg/etc are based on.  Thats how you can listen to an mp3/ogg/DTS/AC3/AAC/whatever file with an SNR of just 30 or 40dB and not notice the (quite loud) quantization noise thats constantly blasting at you.

[jaylee]

Many targets of RB are well designed, enable higher sample rate and bit depth may get better sound quality?

[soap]

Many targets of RB are well designed, enable higher sample rate and bit depth may get better sound quality?

As answered above:

Unlikely.  Can you hear frequencies above 22 KHz?  Can you hear noise 96 dB down in a mix?

The only reason you might hear a difference with native 48 playback is where Rockbox's 48->44 resampler is poorly performing.  This could be addressed for the benefit of all.

[jaylee]

Unlikely.  Can you hear frequencies above 22 KHz?  Can you hear noise 96 dB down in a mix?

The only reason you might hear a difference with native 48 playback is where Rockbox's 48->44 resampler is poorly performing.  This could be addressed for the benefit of all.

emmmmm, you're justified in this. it seems the only meaningful thing for RB (and a DAP) is enabling 48khz playback.

[soap]

Unlikely.  Can you hear frequencies above 22 KHz?  Can you hear noise 96 dB down in a mix?

The only reason you might hear a difference with native 48 playback is where Rockbox's 48->44 resampler is poorly performing.  This could be addressed for the benefit of all.

emmmmm, you're justified in this. it seems the only meaningful thing for RB (and a DAP) is enabling 48khz playback.

It IS enabled.  Not hardware 48 playback, but 48->44.1 resampled playback. Do you have a specific audible issue with it as it stands now or a theoretical distrust of it?

[Benway]

[..]
Unlikely.  Can you hear frequencies above 22 KHz?  Can you hear noise 96 dB down in a mix?
[...]

Humm, I think you're confusing sample rate and audible frequency. 
Of course a sample rate of 22 kHz is audibly better than 44 Khz. This has nothing to do with being able to hear frequencies above 22 kHz.

[torne]

No we aren't. By the Nyquist-Shannon sampling theorem, you only need the sample rate to be double the highest audible frequency. So, unless you can hear above 22KHz, there's no need to use a sample rate higher than 44KHz.

[jaylee]

It IS enabled.  Not hardware 48 playback, but 48->44.1 resampled playback. Do you have a specific audible issue with it as it stands now or a theoretical distrust of it?

i don't trust resample like 48<->44.1, that's not integral, deteriorates sound quality.

[Llorean]

Whether or not it deteriorates sound quality, the important question is whether it audibly does so. Rockbox has many optimizations that deteriorate sound quality in one way or another in minimalistic ways allowing files to, for example, be decoded faster. But if the amount it deteriorates is something you can't reasonably distinguish, does it matter?

[soap]

It IS enabled.  Not hardware 48 playback, but 48->44.1 resampled playback. Do you have a specific audible issue with it as it stands now or a theoretical distrust of it?

i don't trust resample like 48<->44.1, that's not integral, deteriorates sound quality.

I'll repeat my question.
Do you have a specific audible issue with it as it stands now or a theoretical distrust of it?

[saratoga]

Regarding other sample sizes:

We already use full 32 bit precision for everything.  If theres ever a target with a 24 bit DAC it will work as is with rockbox.

Regarding other sample rates:

We currently resample everything using linear interpolation which results in a lot of distortion/aliasing at higher frequencies.  We could improve this either by using better interpolation (cubic spline seems like a good choice, but some kind of IIR based approach might work too) or by supporting more sample rates in playback.

If we wanted to support more sample rates in playback a few things would have to change:

• EQ

  This will probably work at 48kHz fine, since thats only 9% different.  We'd just have to scale the band center frequencies by 9%.  It probably won't work at 32 or 96k though, so it would need to be disabled for really wierd sample rates.

• Crossfade

  This needs to be disabled if either of the tracks has a different sample rate

• DSP Engine

  All the NATIVE_FREQUENCY macros need to become variables and we need to hunt through the code and make sure theres no hidden assumptions about sample duration.

• Gapless

  Transitions between different sample rates probably cannot be gapless, so we don't really have to do anything here.  They'll be a click when you reclock the DAC anyway
[/list]

[amishfury]

No we aren't. By the Nyquist-Shannon sampling theorem, you only need the sample rate to be double the highest audible frequency. So, unless you can hear above 22KHz, there's no need to use a sample rate higher than 44KHz.

there is a very good argument for higher samplerates though... the closer you get to the upper end of the frequency range supported by the samplerate the more aliasing... higher quality DACs of course handle that aliasing better than lesser DACs but also higher samplerates eliminate much of the aliasing simply by having the increased resolution

of course these arguments don't generally apply to portable players where the most common preference is to sacrifice some audio fidelity for smaller file size on top of the fact that most portable players are not exactly designed for samplerates higher than 44.1KHz

[torne]

there is a very good argument for higher samplerates though... the closer you get to the upper end of the frequency range supported by the samplerate the more aliasing... higher quality DACs of course handle that aliasing better than lesser DACs but also higher samplerates eliminate much of the aliasing simply by having the increased resolution

See the sampling theorem again: there's no aliasing at all as long as the sample rate is at least double the signal frequency. Aliasing doesn't start until you *exceed* that. Human hearing even in the young stops around 20KHz, so 44.1KHz sampling is sufficient for playback purposes.

[soap]

I believe amishfury is attempting to refer to problems caused by the low pass filters on non-oversampling DACs having the tail of their roll-off in the (upper part) of the audible range.