Head-Fi.org › Forums › Equipment Forums › Portable Source Gear › The Fiio X3 Thread.
New Posts  All Forums:Forum Nav:

The Fiio X3 Thread. - Page 415

post #6211 of 17219
Quote:
Originally Posted by Marlene View Post
 

To JamesFiiO & Joe Bloggs:

 

I want to ask again for 32 bit floating point decoding of mp3 & aac.

 

Activating it will be as easy as flipping a switch since this option is already included in both formats.

 

Remember: both formats store with 32 bit floating point internally.

 

FiiO wouldn´t be alone should they activate it... here´s a link to a post in the jRiver forum:

http://yabb.jriver.com/interact/index.php?topic=63691.0

 

and here´s a link where Grammy winning mastering engineer Bob Katz says that it´s vital for perfect sound:

http://yabb.jriver.com/interact/index.php?topic=76912.msg521299#msg521299

 

and here´s the link to Apples 'Mastered for iTunes' initiative:

http://www.apple.com/itunes/mastered-for-itunes/docs/mastered_for_itunes.pdf

 

As you can read for yourself, Apple recommends sending them high definition material; they know very well that aac is perfectly able to encode material with high bit-depth. Right now however, the FiiO X3 is unable to play back those 'Mastered for iTunes' releases in high quality.

Those posts don't talk about decoding but how the 32 bit float output of the decoder is handled. I suspect the decoding engines are 32 bit float. I suspect the V control is as well but don't really know. You may also notice the Apple SCR in your link doesn't use dither which is an interesting decision, especially when changing from a non multiple sample rate. Theirs seems to be a sort of 'significant bits' type of SCR but that doesn't address time as much as frequency and amplitude which will still be reduced in the conversion. Something washed over a bit in their description.

 

On an earlier post you talk about downsampling 24 bit to 16 bit but keeping the sampling frequency high as a personally preferred method. That would be counter intuitive to your last post.

 

Personally, I find the bit depth more important than the sampling frequency. All modern dacs upsample/oversample at their input so the sample rate is already increased before decoding. Filter design can be but isn't always optimized for it so results can vary. We can continue on PM if you like. 


Edited by goodvibes - 11/7/13 at 2:25am
post #6212 of 17219

I'm sold on bigger traffic to help reduce jitter, but the rest is always trading one thing for another and doesn't seem to make that much of a difference when compared to the range of headphones we can pick to really improve or experience.

 

that said I've had a feeling that the X3 was sounding better with better/bigger files. something I never really felt before on my other sources where I've always chosen mp3 for space. on the X3 I've been tempted more and more to use 24bit flac for some obscured reason. and as I don't believe much in a better sound with 24/96, I decided that the X3 wasn't handling mp3 very well.

 

now if you call me before a court of justice, I said notin I saw notin!!!

 

 

 

;)

post #6213 of 17219

I am generally a FLAC listener on the X3, but I think it plays high bitrate MP3's well too. It is less successful at low bitrate MP3's, where it exposes the flaws more readily than some other players. Both my Cowon C2 and my iPod make a better job of lower bitrate MP3's (256k or <) than the X3 does.

 

This is a good thing imo, as it has encouraged me to only listen in FLAC these days, which is good for my ears :D

post #6214 of 17219
Quote:
Originally Posted by goodvibes View Post
 

Those posts don't talk about decoding but how the 32 bit float output of the decoder is handled. I suspect the decoding engines are 32 bit float. I suspect the V control is as well but don't really know. You may also notice the Apple SCR in your link doesn't use dither which is an interesting decision, especially when changing from a non multiple sample rate. Theirs seems to be a sort of 'significant bits' type of SCR but that doesn't address time as much as frequency and amplitude which will still be reduced in the conversion. Something washed over a bit in their description.

 

On an earlier post you talk about downsampling 24 bit to 16 bit but keeping the sampling frequency high as a personally preferred method. That would be counter intuitive to your last post.

 

Personally, I find the bit depth more important than the sampling frequency. All modern dacs upsample/oversample at their input so the sample rate is already increased before decoding. Filter design can be but isn't always optimized for it so results can vary. We can continue on PM if you like. 

 

All those posts generally assume that the output of the decoder is handled using its 'natural' 32 bit floating point. And you´re right: the decoding engines of mp3, aac and ogg are 32 bit float. mp3 works internally with 64 bit float, aac with 128 bit float. They all talk about what they´d do to the signal after it´s decoded, fully assuming that it´s 32 bit floating point and not 16 bit.

 

Apple doesn´t want to employ dithering since they know that decreasing bit depth from 32 bit float to 24 bit integer leads to marginal errors only residing at -140 dB (contrary to what Bob Katz is stating in his post at the jriver forum).

 

My earlier post didn´t talk about a personally preferred method, it mentionend a method to conserve storage space. Using lossy material in HiRes quality would save even more storage space while keeping perfect sound.

 

Please be aware that downsampling hasn´t anything to do with bit depth. Downsampling only refers to samplerate, bitdepth however is increased or decreased.

 

For the FiiO X3 decoding to floating point makes even more sense since its digital volume control is either 24 bit or 32 bit floating point, feeding it with 16 bit is a waste. Why not feeding it the full 32/24 bit signal? And even if it isn´t digital but analogue it still makes sense - well encoded lossy stuff has dynamic capabilities close to that of hires-wave; all of which will show at the analogue output.

 

Someone interested in measurements?


Edited by Marlene - 11/7/13 at 7:13am
post #6215 of 17219

Here are a few measurements:

 

Comparison of the original .wav file to encodes in mp3 & aac.

Both lossy formats yield these results when decoded with 32 bit floating point (as decoded by dbPowerAmp, foobar, jriver).

 

dynamic range, wave 32/48

 

dynamic range, mp3 32/48

 

dynamic range, aac 32/48

 

The FiiO decodes the same mp3 and aac files like this:

 

 

dynamic range, FiiO X3 decoded mp3 - clearly 16 bit decoded (compare it to the pic above)

 

dynamic range, FiiO X3 decoded aac - 16 bit as well (compare it to the pic above)

 

total harmonic distortions, FiiO X3 decoded mp3 - not only 16 bit but horrible distortions

 

By now it should be clear to anyone that 16 bit decoding of mp3 & aac is not only unnecessary but produces quantization noise. Furthermore, the FiiO X3 introduces strong distortions when decoding mp3. Every mp3 you´ll play with the FiiO will have increased distortions, no matter what you do. IMO FiiO should change the decoding - and soon.


Edited by Marlene - 11/7/13 at 8:21am
post #6216 of 17219

quick question, how do you make it "play all"  the SD card rather than onboard?

post #6217 of 17219

This is what I responded to: 'I want to ask again for 32 bit floating point decoding of mp3 & aac.'

and they only reason I'm continuing on this vein. Now it seems to be understood. For me, personally, it's not critical in a HiDef player and are we sure it's post processing is different than other DAPs? 

 

I'm well aware that converting from 24 to 16 bit generally involves removing the 8 least significant bits but if you look at Apples SCR (which I haven't studied and just skimmed) It appears it may take the bit depth into account as part of its resampling algorithm so part of the downsampling but it's pretty much semantics to me. As for the Apple converter, 32 (or even 64) floating point resamplers (if we want to get fussy) are nothing new..The rest of the claims just sounds like marketing to battle true HiDef to me so as to allow patrons to believe that they are getting 'perfect sound forever' all over again. I'm actually a bit:rolleyes: concerned considering how influential they are in music sales. You can't have 24/96 res and dynamic range in a 16/44 file. A claim like that would require a lossy algorithm. I agree with Katz regarding dither but I haven't heard everything so not making claims.

 

 Easy way to say it is that I prefer to shrink file size for portable use by maintaining bit depth and you prefer to maintain sample rate. I honestly prefer 24/48 to 16/96.and the first file is 25% smaller. We can agree to disagree.:bigsmile_face: 


Edited by goodvibes - 11/7/13 at 8:28am
post #6218 of 17219
Quote:
Originally Posted by goodvibes View Post
 

This is what I responded to: 'I want to ask again for 32 bit floating point decoding of mp3 & aac.'

and they only reason I'm continuing on this vein. Now it seems to be understood. For me, personally, it's not critical in a HiDef player and are we sure it's different than other DAPs?

 

That´s exactly the point: the FiiO X3 IS a HiDef player so it should be able to play every capable format in HiDef. And formats like mp3, aac or wma are HiDef capable.

 

Quote:
I'm well aware that converting from 24 to 16 bit generally involves removing the 8 least significant bits but if you look at Apples SCR (which I haven't studied and just skimmed) It appears it may take the bit depth into account as part of its resampling algorithm so part of the downsampling but it's pretty much semantics to me. As for the Apple converter, 32 (or even 64) floating point resamplers (if we want to get fussy) are nothing new..The rest of the claims just sounds like marketing to battle true HiDef to me so as to allow patrons to believe that they are getting 'perfect sound forever' all over again. I'm actually a bit:rolleyes: concerned considering how influential they are in music sales. You can't have 24/96 res and dynamic range in a 16/44 file. A claim like that would require a lossy algorithm. I agree with Katz regarding dither but I haven't heard everything so not making claims

 

Of course it´s marketing on Apple's side. They want to have the customers now at HDTracks, that much is for sure. But it doesn´t matter; people will buy it, believing it to be in better quality and they should subsequently be able to enjoy the sonic advantage.

 

Quote:
Easy way to say it is that I prefer to shrink file size for portable use by maintaining bit depth and you prefer to maintain sample rate. I honestly prefer 24/48 to 16/96.and the first file is 25% smaller. We can agree to disagree.:bigsmile_face: 

 

Look above at the measurements I posted. They were derived from 32/48 files. Technically speaking HiDef... although the FiiO ignores that still (in case of mp3 with horrible results).

post #6219 of 17219

Your last comment and those graphs don't really relate to the sample vs word discussion but that's fine. 

 

 Do you know the actual bit rate of the MP3 file in question? I believe that's a pretty important parameter when looking at measurements. I would think most using this player would use 320 or at least directly available 256.


Edited by goodvibes - 11/7/13 at 9:00am
post #6220 of 17219
Quote:
Originally Posted by goodvibes View Post
 

Your last comment and those graphs don't really relate to the sample vs word discussion but that's fine. 

 

A for those measurements, he fellow that took those mp3 measurements claims analog output is VG with noise + THD 94db down.:bigsmile_face:

 

 

I believe that you may be overstating. Do you know the actual bit rate of the MP3 file in question? I believe that's a pretty important parameter when looking at measurements. I would think most using this player would use 320 or at least directly available 256.

 

What do you mean with the analogue output sentence? The table you quoted above compares encodes directly decoded via software on my PC. The software used was stated below that picture, meaning, 32 bit decoding of mp3 & aac always produces this result and it doesn´t matter if you use dbPowerAmp, foobar or jriver.

 

And I did those measurements myself. No one else here :bigsmile_face: The table comparisons are made with Photoshop (only so that I wouldn´t have to post too many seperate pictures). And the 'Excellent' or 'Very Good' statements inside the table are not my assessments, those are assessments done by the software I used for measuring: RMAA.

 

As for the bitrate: mp3 and aac were encoded with 320 kBit/s CBR (constant bitrate). In case of mp3 I used LAME, aac was encoded using Apple's encoder.

 

You also have to consider that both mp3 and aac don´t work very well with measurement signals; they simply weren´t constructed to encode those. mp3 and aac are perceptual codecs, meaning they work best with real-world examples, aka music. That they measure as well as they did on my examples only proves what I´ve been saying for quite some time now: they´re better than their reputation and they´re certainly able to encode HiDef material.

 

BTW, the analogue output of the FiiO is perfectly able to show at least part of this advantage. Using the line-out I´ve measured a dynamic range of 113 dB. For a portable device this is a fantastic result. I used FLAC, using mp3 or aac would have yielded worse results of course. In case of mp3 much worse results.


Edited by Marlene - 11/7/13 at 9:10am
post #6221 of 17219
Does anyone know if the USB dac on the x3 is good enough to replace something like the e17?
post #6222 of 17219
Quote:
Originally Posted by audioKyle View Post

Does anyone know if the USB dac on the x3 is good enough to replace something like the e17?

Although using the same dac chip, x3 is much better on paper. It runs in async mode vs e17's sync mode, which, not only support 24/192 over usb, but also eliminates jitter pollution effectively. However, x3 does not have any digital inputs. It's a DAP after all. In reality, I couldn't hear major differences - maybe that's just my ear...

post #6223 of 17219
Quote:
Originally Posted by Marlene View Post
 

 

What do you mean with the analogue output sentence? The table you quoted above compares encodes directly decoded via software on my PC. The software used was stated below that picture, meaning, 32 bit decoding of mp3 & aac always produces this result and it doesn´t matter if you use dbPowerAmp, foobar or jriver.

 

And I did those measurements myself. No one else here :bigsmile_face: The table comparisons are made with Photoshop (only so that I wouldn´t have to post too many seperate pictures). And the 'Excellent' or 'Very Good' statements inside the table are not my assessments, those are assessments done by the software I used for measuring: RMAA.

 

As for the bitrate: mp3 and aac were encoded with 320 kBit/s CBR (constant bitrate). In case of mp3 I used LAME, aac was encoded using Apple's encoder.

 

You also have to consider that both mp3 and aac don´t work very well with measurement signals; they simply weren´t constructed to encode those. mp3 and aac are perceptual codecs, meaning they work best with real-world examples, aka music. That they measure as well as they did on my examples only proves what I´ve been saying for quite some time now: they´re better than their reputation and they´re certainly able to encode HiDef material.

 

BTW, the analogue output of the FiiO is perfectly able to show at least part of this advantage. Using the line-out I´ve measured a dynamic range of 113 dB. For a portable device this is a fantastic result. I used FLAC, using mp3 or aac would have yielded worse results of course. In case of mp3 much worse results.

Thanks. I'd be interested to see how other players compare even though I'd stick to lossless on products like this.

post #6224 of 17219
Quote:
Originally Posted by goodvibes View Post
 

Thanks. I'd be interested to see how other players compare even though I'd stick to lossless on products like this.

 

Well, I have multiple MD recorders to compare the FiiO to. Though a comparison would be pointless since they use their propriatary compression scheme.

 

When it comes to 'modern' portable, I´ve measured the Sony NW-A1000 and the Sansa Clip+. The first with AAC + MP3, the latter with FLAC.

 

Both don´t have a chance against the FiiO except when the FiiO decodes MP3. In that case both are superior.

post #6225 of 17219
Quote:
Originally Posted by Marlene View Post
 

Here are a few measurements:

 

<snip>

 

By now it should be clear to anyone that 16 bit decoding of mp3 & aac is not only unnecessary but produces quantization noise. Furthermore, the FiiO X3 introduces strong distortions when decoding mp3. Every mp3 you´ll play with the FiiO will have increased distortions, no matter what you do. IMO FiiO should change the decoding - and soon.

Most interesting.

 

It explains why I've been feeling the mp3 decoder needs a complete overhaul. As it is it isn't worthy of inclusion in a serious DAP. The thing I first noticed was the horrible quanitization noise.

 

Could you do the same measurements for the ogg decoder so we can see what is so seriously flawed in that too.

 

And while I'm whining, can FiiO please note that the mp3 decoder is most certainly not capable of gapless playback.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Portable Source Gear
Head-Fi.org › Forums › Equipment Forums › Portable Source Gear › The Fiio X3 Thread.