-24db Precut Sound Quality on Rockboxxed iPod?

[Andhyka]

Of course as I stated on the diagram above, the output will always be some decibels short because Rockbox outputs it on 16-bit audio signal wide to Wolfson DAC.

DIGITAL 16bit -> DSP 16bit on 28bit/32bit -> DAC 14bit on 14bit/16bit (or 12bit, I don't know which one)

As for the signal quality on 32-bit Rockbox internal DSP, there's no depreciation.

If only Rockbox outputs in 24-bit...

Ofcourse, pre-cut on current Rockbox cuts away the 4bit (-24db) of signal data, and I don't intend to use it.

If only Rockbox has normalize/compress algorithm...

[Llorean]

If the only operation you're doing in the DSP is the digital gain adjustment, outputting at 24-bit should be no different than 16-bit. Math again.

And I told you already: COMPRESSING IS NOT LOSSLESS.
Mathematically the least you can possible lose is the least significant bit. Please, do some proper research on the audio side, and learn basic math.

Do you know the mathematical difference between "dynamic range compression" and "Reducing the volume"?

And I remind you again, though I already did in PM, that you need to read the rules. This is your last warning on that one.

[Andhyka]

I understood everything that Llorean said, and is entirely valid.

What I wrote in the few posts beforehand is my wish that Rockbox could develop this feature in future builds so that to get better sound quality out of iPod.

I did decide not to use pre-cut before (as mentioned in the thread).

[Llorean]

I think you don't understand: You cannot magically prevent the loss of data. All you have is a choice of which data will be lost. Why don't you look into what exactly precut does, rather than jumping to a conclusion based on the phrasing a user used, and see and decide for yourself whether that's better or worse than analog volume adjustment.

[Andhyka]

Yes I do know the difference.

Dynamic range compression determines the fidelity/clarity/detail of sound, and is related to 16bit/24bit/32bit.

And also, I can get the same quality with any volume as long as I put my ipod line-out on headphone amp.

Just one last question before I bail out of this unconstructive thread ever:
Does Wolfson DAC accepts 24-bit digital signal?

If not, then I'm throwing away everything I know.

If yes, what I am proposing is "reducing the volume" through the use of wide "dynamic range compression" of 32-bit Rockbox DSP.

Therefore I can listen the cleanest signal reproduction iPod can reproduce from my iPod without requiring headphone amp.

[Febs]

Dynamic range compression determines the fidelity/clarity/detail of sound, and is related to 16bit/24bit/32bit.

You need to go figure out the difference between "dynamic range" and "dynamic range compression."  You're just saying nonsense.

[Andhyka]

What's your point anyway. I knew that pre-cut is bad halfway of the thread, and I don't intend to use it.

I'm giving the feedback to Rockbox developers to have a look at, from the satisfied users of Rockbox.

If it's possible to implement, then go ahead.

if it's possible to implement and there's no plan to do it, then I don't mind.

If it's not possible to implement, then I'm fine with it.

[Andhyka]

Man, of course I know dynamic range.

And everywhere I see it's about how much detail you put in particular frequency range (coarse definition).

Who doesn't know what it is when every computer/audio store is cluttered with HDR SM3 video card, 24-bit audio cards, SACDs, 24-bit audio DVDs.

Edit: Hmm you guys really don't know what I mean, or you are just pretending you don't know about it for whatever purpose. I have had enough discussion and argument about this point anyway.

Edit 2: I just checked the schematics of iPod DAC, and looks like it does not have any amp, outputting line-out all the time, and uses its DAC to control volume digitally. This means the resultant signal will be compressed.

Edit 3: I just re-checked Wolfson DAC specs on iPod. It says it does 24-bit processing inside its chip. Since 16-bit is fed into Wolfson and upscaled to 24-bit inside,  I did quick calculation to check what level of volume control would be optimal without loss of SQ. Results: anything above -48db or >25% volume control.

I rest my case.

Overall, you guys did amazing job to Rockbox, and grats to Apple for making some right decision.

[Febs]

Read this:  http://en.wikipedia.org/wiki/Dynamic_range_compression

[Llorean]

1) Do proper research in the future. You're talking with people who know what they're talking about. You've admitted you aren't one, and you're arguing a point you don't know the inner workings of.

2) You don't know how Rockbox works, and are apparently unwilling to learn how it does things internally before claiming there are better ways (if you don't know factually how Rockbox does it, it is physically impossible for you to claim there is a better way). Saying "precut sounds bad" doesn't mean you have any clue how it works, other than an assumption you made based on a statement by one person. Go and read it, and please don't post again in this thread about it until you know what exactly our algorithm is doing.

3) You apparently refuse to admit the existence of the transitive property of mathematics.

Please address all these issues before making statements that relate to them.

[preglow]

I do not know much about Rockbox internals so correct me if I am wrong.
When you set precut to -24 dB you are approximately "shifting" the original sample value by 4 bits to the less significant bits thus the 4 most significant bits are not used by the original signal anymore. At the final stage before sending the signal to the DAC only the most significant 16 bits are taken (optionally with some information from lower bits by dithering) so in our case only 12 bits from the original signal would be used and the 4 most significant bits will be wasted (zero).

This is basically true, but a bit confused. Let's take MP3, for example, which operates at a 29 bit accuracy. A -24 dB cut will approximately throw away the lowest four bits, and this leaves us with a 25 bit accurate signal, and the 16 highest of these bits is what is passed to the DAC (possibly after further processing). This still leaves us with 9 bits to waste before the output should suffer at all. This is the entire reason we opted to go for a 32 bit DSP path in the first place.
This is of course assuming you do some EQing to boost some bands again. If you just cut -24 dB and do nothing, then yes, the resulting signal will basically be 12 bits thanks to the level being so low overall.

[saratoga]

If you just cut -24 dB and do nothing, then yes, the resulting signal will basically be 12 bits thanks to the level being so low overall.

Unfortunately, this is what the original poster wants to do!

Edit 2: I just checked the schematics of iPod DAC, and looks like it does not have any amp, outputting line-out all the time, and uses its DAC to control volume digitally. This means the resultant signal will be compressed.

No.  There has to be an amp.  You can't drive headphones off of a logic gate.  

Dynamic range compression determines the fidelity/clarity/detail of sound, and is related to 16bit/24bit/32bit.

Dynamic range compression is a type of DSP effect that you can apply to audio.  It is not a metric.  It does not determine anything.  Its not relevant here.

And also, I can get the same quality with any volume as long as I put my ipod line-out on headphone amp.

You've said this a few times.  I don't see why it should be true.  

Just one last question before I bail out of this unconstructive thread ever:
Does Wolfson DAC accepts 24-bit digital signal?

If not, then I'm throwing away everything I know.

If yes, what I am proposing is "reducing the volume" through the use of wide "dynamic range compression" of 32-bit Rockbox DSP.

Did you read my previous post on this idea?  As I said above, controlling the volume digitally will always reduce the SNR.  Always.  What you are suggesting doesn't make a lot of sense.

[Andhyka]

Cheers for all the feedback (although a bit harsh), and I think I've found what I'm looking for.

The reason that I bring all these matters up is because I've got highly detailed headphone MDR-SA5000 and was thinking if it sounds better with imod + headphone amp + dock connector cable.

But ofcourse, this is beyond what I can afford at the moment.

Therefore, I started this whole thread to discuss by proposing that line-out output (100% volume) can still be audible through headphone by feeding 24-bit data signal from Rockbox DSP to DAC (which then I found out this is wholly unnecessary, from you guys).  

But as it turns out, the iPod has DAC chip much better than I think, most MP3 players out there. It does 24-bit signal processing in both ADC/DAC filters.

If that is true, digital signal received by DAC will be upscaled from 16-bit to 24-bit, then the volume control will compress/normalise according to the volume level specified.

Thus I can derive that volume level below 100% and above 25% DOES NOT AFFECT the compression of final output.

Don't get me wrong, the Wolfson DAC chip still compresses dynamic range of received signal according to specified volume level but it already upscales the signal to 24-bit internally (16-bit data in 24-bit channel), leaving 8-bit of interpolated fidelity.

Taken from Wolfson WM8393 white paper, 5G iPod Video DAC:
"The digital audio data is converted to oversampled bit streams in the on-chip, true 24-bit digital interpolation filters. The bitstream data enters the multi-bit, sigma-delta DACs, which convert it to high-quality analogue audio signal. The multi-bit DAC architecture reduces high frequency noise and sensitivity to clock jitter. It also uses a Dynamic Element Matching technique for high linearity and low distortion."
http://www.wolfsonmicro.com/uploads/documents/en/WM8983.pdf

I must admit though that my understanding on electrical engineering & audio manipulation are limited, and this matter may seem trivial to you, especially Rockbox developers cos they definitely knew this already years ago.

Reply to saratoga: I think iPod does have an amp but is set at static level (which volume can't be increased or decreased). iPod relies on digital volume on its DAC which results in dynamic range compression (less fidelity/clarity).

[Febs]

One more time, since you seem to have great difficulty understanding this concept:  Changing the volume on an ipod DOES NOT RESULT IN DYNAMIC RANGE COMPRESSION.  You are mis-using that term.

[Andhyka]

Exactly, it doesn't.