[SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
In a thread on the ripping forum recently, mswlogo has sussed out how to 'up-bit-depth' his flacs from 16 to 24 bits. Filesize is similar, as it just compresses out, so no downside. The file is padded out with zeroes, so there is no change Potential upsides he gives are reduced jitter and (if applicable) better DRC from higher bit depth of processing. The latter I don't care about, but the former is interesting. Comments? I would have thought that higher bit depth would change the data timing, but would it always improve it? Analysis from techies welcome! Adam -- adamslim Those are my principles, and if you don't like them... well, I have others http://www.last.fm/user/AdamSlim/ adamslim's Profile: http://forums.slimdevices.com/member.php?userid=7355 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
I don't see how simply having more bits (with no information in them) makes it any more likely that any of them will arrive at the right time...I would have thought that it's even harder to control jitter at higher bit rates, since everything is happen faster if you see what I mean. What am I missing? -- Phil Leigh Phil Leigh's Profile: http://forums.slimdevices.com/member.php?userid=85 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
Jitter will be exactly the same. The SPDIF spec transmits 32 bits of data per sample anyway; up to 24 of these are available for the sample word (with the bottom bits explicitly set to zero if unused!), and the rest are overheads such as sync pulses, checksums and other stuff. Apart from possibly changing a flag bit or two in the general-purpose data area, there is ABSOLUTELY NO DIFFERENCE between transmitting 16 bit samples, and 24 bit samples where the bottom bits are all zero. Since there is no difference at all in the bit stream, there will be no difference in downstream processing either. If a DAC is capable of oversampling / filtering to 24-bit accuracy, it'll be doing that anyway even with a 16 bit input, probably without even knowing the difference. -- AndyC_772 AndyC_772's Profile: http://forums.slimdevices.com/member.php?userid=10472 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
Cheers Andy, good reply. Saves me from agonising over whether to do lots of converting and testing :) Adam -- adamslim Those are my principles, and if you don't like them... well, I have others http://www.last.fm/user/AdamSlim/ adamslim's Profile: http://forums.slimdevices.com/member.php?userid=7355 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
Thanks Andy - that makes sense! -- Phil Leigh Phil Leigh's Profile: http://forums.slimdevices.com/member.php?userid=85 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
If your going to upsample as well doing it in larger word length is more accurate (some would say a better guessitmate). Upsampling doesn't make a lot of sense with Slimdevices because SB3 only does 48khz and Transporter doesn't do 88.2Khz. BUt my library is for other devices beside the Slimdevices. I don't know why but in this stereophile review of the transporter 16 vs 24 changed the jitter measurement. http://www.stereophile.com/mediaservers/207slim/index4.html Snippet: I tested the Transporter's rejection of word-clock jitter using the Miller Jitter Analyzer, which examines a narrowband, FFT-derived spectrum of the analog output of the device under test (DUT) for pairs of sidebands around a high-level tone at one quarter the sample rate, while the LSB is toggled on and off at 1/128 the sample rate. (Both signals are exact integer fractions of the sample rate, meaning that any spuriae that appear in the spectrum are due to the behavior of the DUT, not to quantizing distortion.) Fed 24-bit data via the WiFi network, the Transporter developed just 235 picoseconds peakpeak of jitter with no data-related components (not shown). Decreasing the word length to 16 bits gave the spectrum shown in fig.10. Here the jitter level has increased slightly, to 268ps pp; though there are data-related sidebands (red numeric markers) at the test signal's residual level. The primary jitter components lie at ±15.6Hz (purple 1) and ±1435Hz (purple 10), but this is still excellent performance. -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
But didn't you just say 16 vs 24 wouldn't make any difference in the actual jitter of the 32bit word. Why would it matter which bits are being used in the test? So your saying each bit has it's own jitter measurement? Or if he toggled Bit 8 of the 24bit word he'd get the same measurement as bit 0 of 16bit? That makes no sense to me. -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
mswlogo;198595 Wrote: But didn't you just say 16 vs 24 wouldn't make any difference in the actual jitter of the 32bit word. Why would it matter which bits are being used in the test? So your saying each bit has it's own jitter measurement? Or if he toggled Bit 8 of the 24bit word he'd get the same measurement as bit 0 of 16bit? That makes no sense to me. I must admit, I don't know for sure why the test results are different - but I suspect it's much more to do with flaws in the test itself than the behaviour of the Transporter. For example, the AK4396 DAC is a 128x oversampling delta-sigma device, so there will always be data-dependent noise at frequencies outside the audio band. More than likely there is an analogue filter between the DAC and the output socket to attenuate this noise. Now suppose that, in order to allow excellent flatness up t0 22kHz, this filter has a cutoff around 100kHz. The test equipment is now trying to pick out jitter in the order of 100ps (which is the period of a clock running at 10GHz) in a signal which has been filtered to remove everything above 100kHz. Is that a sensible thing to try and do with any degree of accuracy? Or is the measurement likely to be influenced by the audio signal? I strongly suspect the latter. I'd bet that if the jitter measurement were made at the MCLK pin of the AK4396 - the correct place to measure it - that it would be completely independent of the data. -- AndyC_772 AndyC_772's Profile: http://forums.slimdevices.com/member.php?userid=10472 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
AndyC_772;198601 Wrote: I must admit, I don't know for sure why the test results are different - but I suspect it's much more to do with flaws in the test itself than the behaviour of the Transporter. For example, the AK4396 DAC is a 128x oversampling delta-sigma device, so there will always be data-dependent noise at frequencies outside the audio band. More than likely there is an analogue filter between the DAC and the output socket to attenuate this noise. Now suppose that, in order to allow excellent flatness up t0 22kHz, this filter has a cutoff around 100kHz. The test equipment is now trying to pick out jitter in the order of 100ps (which is the period of a clock running at 10GHz) in a signal which has been filtered to remove everything above 100kHz. Is that a sensible thing to try and do with any degree of accuracy? Or is the measurement likely to be influenced by the audio signal? I strongly suspect the latter. I'd bet that if the jitter measurement were made at the MCLK pin of the AK4396 - the correct place to measure it - that it would be completely independent of the data. Ok, thanks Andy, I think we are closer to the same page now. I figured it wouldn't hurt to play with. I'm also not sure how my processor handles 16 vs 24 bit and may do some odd things to be backward compatible with different devices. So I figured it wouldn't hurt to play with. The first thing my processor does is shift the data down to make headroom for some processing and calibration. I don't know if it does that in 16bit on 16bit data or converts first. It is capabable of outputing 16bit data to older active speakers and that data is just lost. With newer active speakers I can't tell what it does because SPDIF data is encrypted. So far I hear no difference. But I need to try it on a wider range of music. Also just a thought wouldn't replay gain (done digitally) or any normalization now have lot headroom in 24bit? -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
What I was getting at, is that your DAC, or your processor, or whatever else is connected over SPDIF, simply won't know whether it's getting 16 bit or 24 bit data. The SPDIF frame includes 24 bit positions for use with audio data, and a source that only has 16 bits available simply pads the missing bits with zeros. So, there cannot be any difference in how your processor handles the data. That's why you can't hear any difference. SPDIF (at least, the 2 channel uncompressed version we're using) isn't encrypted, BTW - just encoded to ensure plenty of edges for clock recovery. The AES/EBU spec is freely available and explains exactly how to decode it - you could do it with a storage scope and a pencil paper if you wanted. -- AndyC_772 AndyC_772's Profile: http://forums.slimdevices.com/member.php?userid=10472 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
So you're saying I've effectively prepadded the low 8bits before it would have already? -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
mswlogo;198587 Wrote: http://www.stereophile.com/mediaservers/207slim/index4.html Stereophile's jitter measurement in this case is just nonsense. They are using the Miller jitter analyzer, which implements the test described in Julian Dunn's Jitter and Digital Audio Performance Measurements. The test measures data induced jitter in -a manchester clock recovery circuit-, i.e. an s/pdif receiver. It can not (and is not intended to) measure anything else! Since they used Transporter's internal clock, not its s/pdif receiver, the test is simply meaningless. -- seanadams seanadams's Profile: http://forums.slimdevices.com/member.php?userid=3 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
mswlogo;198613 Wrote: Also just a thought wouldn't replay gain (done digitally) or any normalization now have lot headroom in 24bit? No, the volume function is always 24 bit. -- seanadams seanadams's Profile: http://forums.slimdevices.com/member.php?userid=3 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
seanadams;198655 Wrote: No, the volume function is always 24 bit. But if someone did normalization on a 16bit file you could lose data. Where is if was done on 24bit data your unlikely to lose anything. Are you saying that even digital volume done on the SPDIF output of a SB3 is effectively done on a 24bit word? -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
mswlogo;198663 Wrote: But if someone did normalization on a 16bit file you could lose data. Where is if was done on 24bit data your unlikely to lose anything. I think I follow what you're trying to say there, but it does not apply. The volume function is always 24 bits wide. It doesn't know or care whether the lower 8 bits are used, and all 24 bits of output are meaningful regardless of the input word length. Are you saying that even digital volume done on the SPDIF output (i.e. replay gain) of a SB3 is effectively done on a 24bit word? That's exactly right, if you drop the word effectively. There is no effectively. It's precisely the same operation, not merely the same result. The _input signal_ in either case is precisely the same. -- seanadams seanadams's Profile: http://forums.slimdevices.com/member.php?userid=3 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
seanadams;198672 Wrote: I think I follow what you're trying to say there, but it does not apply. The volume function is always 24 bits wide. It doesn't know or care whether the lower 8 bits are used, and all 24 bits of output are meaningful regardless of the input word length. That's exactly right, if you drop the word effectively. There is no effectively. It's precisely the same operation, not merely the same result. The _input signal_ in either case is precisely the same. Ok, I understand the SPDIF part. But if I have a 16bit wav file, and applied a normalization (not replay gain, static normalization of the data). which may reduce it's volume and it wrote out a new 16bit file, data is lost due to a Volume change. So if I converted that file to 24bit first then did normalization it would unlikely lose any data. Correct? On a similar topic, do plugins like the room correction also work in 24bit (when fed 16bit wave file)? -- mswlogo mswlogo's Profile: http://forums.slimdevices.com/member.php?userid=9090 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
yes any static normalisation is a bad idea, regardless of how many bits you have to play with. The engineers did their best to master your bits...do not mess with them - you will not make them better - only different :0) -- Phil Leigh Phil Leigh's Profile: http://forums.slimdevices.com/member.php?userid=85 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles
Re: [SlimDevices: Audiophiles] Potential benefits of converting to 24 bit
mswlogo;198673 Wrote: Ok, I understand the SPDIF part. Actually I was talking about the volume function, but yes, this applies to the s/pdif link also. But if I have a 16bit wav file, and applied a normalization (not replay gain, static normalization of the data). which may reduce it's volume and it wrote out a new 16bit file, data is lost due to a Volume change. So if I converted that file to 24bit first then did normalization it would unlikely lose any data. Correct? I can only guess what you mean by unlikely or losing any data, but the only reason you would get more precision in the latter case is because the volume function you are describing happens to be designed to output 16-bits when given a 16-bit input. A volume function could just as easily generate 24 bits of output, or 137 bits for that matter, from a 16 bit input. There is nothing about padding the the input signal to a longer word which causes a more accurate output. Obviously if you truncate your result to 16 bits, it is less precise than if you have not truncated it. -- seanadams seanadams's Profile: http://forums.slimdevices.com/member.php?userid=3 View this thread: http://forums.slimdevices.com/showthread.php?t=34892 ___ audiophiles mailing list audiophiles@lists.slimdevices.com http://lists.slimdevices.com/lists/listinfo/audiophiles