Support for 24 bit / 192 Khz FLAC?

[Kitlope]

I have been ripping some DVD Audio discs of mine to 24 / 96 and 24 / 192. I can play the 24 / 96 on my Rockboxed iRiver iHP-120 but the 24 / 192 just skips through the songs real quick with no audio.

Just wondering if its supported or if I need to downsample to 24 / 96.

Thanks


Kit

[Llorean]

You should actually downsample all of it to 16/44. Rockbox is going to play them back at that anyway, and your resampler is going to be better than the built in one.

[Kitlope]

ahhh yes I remember that now. Ok thanks.

[Packgrog]

I've been rummaging around in the forums trying to find more concrete info on this, but is there any reason why 16/44 is always set as the hard limit?  I've seen it written that everything is upscaled to 32-bit before being run through all of the various effects, then downscaled (and dithered if enabled) to redbook.  Why is 16/48 not even an option?  Don't most of the targets support that (iPods certainly do)?  I know that's the upper limit for any MP3 files I've encoded in the past (I stick with FLAC nowadays).

Is it simply a matter of too much effort required to determine the maximum possible output of each build target that prevents higher bit depth and sampling rates?  Sure, the DAC in the iRiver H1xx has some lower limits than desired, but if using spdif exclusively (like most people still using the H1xx at this point, I would assume), why not just pass it through?

Also, is there no means for bit-exact throughput?  If you're always upscaling everything that gets played to 32-bit and downscaling it back to 16-bit, even if the source files are already at 16/44, there's not really any benefit in doing the dithered transcode from 24/96 to 16/44 before loading an album onto the player.  No matter what, you're still stuck with whatever algorithms Rockbox uses for dithering and downsampling, so you may as well still use the highest supportable resolution (24/96 at the moment).

Sorry if this is rambling.  Too many interruptions.  One way or another, the player sounds better with my external DAC (iBasso D1 with upgraded opamps to a Denon AH-D2000 with JMoney pads) than any other source I have available at the moment.  I'm not trying to troll or make demands or anything, I just wish I could squeeze that last drop of quality out of the best source I own. 

[saratoga]

  Why is 16/48 not even an option?  Don't most of the targets support that (iPods certainly do)? 

Because the DSP system only works at 44.1khz.  You can playback files at other sample rates on most targets, but not through the playback system (only plugins) since the sample rate would be wrong for all the effects and filters.

Also, is there no means for bit-exact throughput?

I think the one rockbox player with digital out can.  Obviously no other device can do it.

  If you're always upscaling everything that gets played to 32-bit and downscaling it back to 16-bit, even if the source files are already at 16/44, there's not really any benefit in doing the dithered transcode from 24/96 to 16/44 before loading an album onto the player.  No matter what, you're still stuck with whatever algorithms Rockbox uses for dithering and downsampling, so you may as well still use the highest supportable resolution (24/96 at the moment).

I think you're confusing resampling (which means changing the sample rate) with changing the bit depth (which means converting from 32 bit to 16 bit in the case of rockbox's audio drivers.  Rockbox's decoders all operate at 32 bit precision (because we run on 32 bit CPUs), but they do not automatically convert sample rate unless the source is not already 44.1khz. 

[Packgrog]

Heh, yeah I'm confused.  This helps a little, but I'm still unclear on some things.

If, for example, someone is only using the digital output, is the DSP system being bypassed?  Is the resampling done even when using digital out?  Is the resampling bypassed if the file is already at 44.1kHz?  Is the DSP required when using ReplayGain (seems it would be)?  It seems incorrect to say that the output is bit-exact if it has to go through the DSP and get resampled even when going through spdif.

Again, I'm mostly just looking for information, not making demands.  Everything I'm playing through the thing sounds amazing, so I'm definitely not complaining.  I want to make sure I understand exactly what works, and what efforts on my part to improve things by manipulating files before loading them onto the player will provide the best results.

[saratoga]

If, for example, someone is only using the digital output, is the DSP system being bypassed?  Is the resampling done even when using digital out?  Is the resampling bypassed if the file is already at 44.1kHz?  Is the DSP required when using ReplayGain (seems it would be)?  It seems incorrect to say that the output is bit-exact if it has to go through the DSP and get resampled even when going through spdif.

As I said I think bit exact playback is possible, but I'm not sure.  Since theres probably 3 people on earth using spdif in rockbox its not very interesting to me.

[AlexP]

I use s/pdif on my H140 to receiver, but I've never checked what the output is.

[Packgrog]

If, for example, someone is only using the digital output, is the DSP system being bypassed?  Is the resampling done even when using digital out?  Is the resampling bypassed if the file is already at 44.1kHz?  Is the DSP required when using ReplayGain (seems it would be)?  It seems incorrect to say that the output is bit-exact if it has to go through the DSP and get resampled even when going through spdif.

As I said I think bit exact playback is possible, but I'm not sure.  Since theres probably 3 people on earth using spdif in rockbox its not very interesting to me.

3?!  I think you're vastly underestimating the popularity of the H1x0 for use with spdif.  Perhaps it's only 3% of the people using Rockbox, but still probably a lot more people than you think.  It's pretty much the target audience for iBasso's D1 and D10.  Maybe I should do a poll on Head-Fi.org to see just how many people are using this...

EDIT: I grabbed the source from svn.  I'll see if I can find the time to figure out how to circumvent the resampler with spdif, even if just for my own purposes. 

[saratoga]

3?!  I think you're vastly underestimating the popularity of the H1x0 for use with spdif.  Perhaps it's only 3% of the people using Rockbox,

I'd be pretty surprised if even 3% of people have an h100 as their rockbox device, and of those I'd be even more surprised if even 3% use an amp with digital in.  Devices like that are extremely rare outside of a handful of headfi users. 

[Packgrog]

Well, I tinkered with the code a bit and yeah, altering the dependencies on the sample rate would be pretty challenging.  Ick.  Messing with the bit depth seems equally nasty.  On the bright side, at least I'm able to create my own Rockbox build if I really need to! 

But at least I have a definitive answer: Redbook (16-bit 44.1kHz) is bit-exact (I saw the conversions being bypassed in this case), and while Rockbox handles higher-res quite nicely, transcoding to redbook beforehand with better dithering will provide better results.

Sorry for the spam, and thanks for all the hard work guys.  I love the hell out of Rockbox.