I have been reading a lot recently about MQA coding for digital streaming.
I recently got a NAD BlueSound unit to play with. It seems to work well. Missing some features IMO, but the iOS app is nice.
My question relates to some discussion on the BlueSound forum where folks have been talking about "MQA capable DACs". This doesn't make any sense to me, so I thought I'd raise it as a topic for discussion here.
As described here: MQA Questions and Answers is really a coding scheme wherein they exploit the noise floor of 24 bit in-band audio to expand the sample rate up to 96 or 192 KHz.
From what I can tell they basically rely on the fact that, in most situations the lower few bits of the 24 bit field in the normal 20 KHz listening band is unused (which is to say from one sample to another, those bits don't change "much"). so rather than repeat them, they apparently repurpose them to carry additional sample info from higher up in the spectrum.
I must say, after reading the "tutorial" there is a lot unsaid, and a fair bit of "audiospeak" (aka "handwaving, using fancy phrases").
While this seems like a clever idea, it also seem like it is solving the wrong problem, and seems to ignore some important issues.
First, the human ear can't really hear much past 20 KHz. Most folks are hard pressed to hear 15 KHz. so reproducing signals up in the 50 KHz -100 KHz range isn't really what high sample rates are seeking to do. Instead it seems to me that the key issue is that placing the anti-aliasing filter at 22 (or 24) KHz is cutting this pretty close. No filter is ideal, so there will be some finite roll-off, and this will mean some higher frequency signal components may get aliased in at lower frequencies. Making the filter tighter can avoid this, but this generally introduces group delay variations (i.e dispersion), which can be easily heard as a smearing or muddling of the sound, specially the high frequency components of percussion instruments. So, while a 192 sample rate is nice, the goal is not to reproduce super high frequencies, it is to avoid superhuman filters (for antialiasing in the recording phase, and sample noise reduction in the playback phase). So, on its face the MQA idea seems to me to be barking up the wrong tree. I'd love to hear other's thoughts on this.
Second, part of the value of 24 bit HD audio files is that there are plenty of times when you ARE using that entire range. Live recordings are a good example, where audience noises are far below the music level, but a good system can bring them out, and make the entire recording feel amazingly real. Use those lower bits for "folding" or packing" higher frequency samples, and you are trading sonic detail in the listening band for sampling detail outside the listening band. Thoughts on this?
Lastly, the MQA process seems sort of mysterious. They claim that the "folding" is done in "two parts" and that the first "fold" can be unfolded in the software decoder. I believe that...since it is really just a matter of somehow tagging the samples to you know which ones have folded content, and then separating out those bits as higher frequency samples, while replacing them with interpolated samples from nearby non-folded samples (here they rely on the fact that supposedly these low level bits don't change much, so you can effectively under sample them, use the unused bits for something else, and then oversample them at the decode phase to recover them). Seems realistic, ASSUMING there really isn't mush variation in those bits. To the extent that there is, then you have lost resolution, and that will show up as lost detail.
And last-lastly, they then claim that the second unfolding step cannot be done in hardware, thereby necessitating, I presume specialized DACS that can perform this magic (while presumably returning license fees to the MQA inventors). One of the inventors (Stuart) claims that this step is an "analog to analog" process, which seems silly if it is a sample based thing. Does anyone here have any idea what goes on in the purported MQA capable DACs?
Looking forward to a lively discussion..
Scott
I recently got a NAD BlueSound unit to play with. It seems to work well. Missing some features IMO, but the iOS app is nice.
My question relates to some discussion on the BlueSound forum where folks have been talking about "MQA capable DACs". This doesn't make any sense to me, so I thought I'd raise it as a topic for discussion here.
As described here: MQA Questions and Answers is really a coding scheme wherein they exploit the noise floor of 24 bit in-band audio to expand the sample rate up to 96 or 192 KHz.
From what I can tell they basically rely on the fact that, in most situations the lower few bits of the 24 bit field in the normal 20 KHz listening band is unused (which is to say from one sample to another, those bits don't change "much"). so rather than repeat them, they apparently repurpose them to carry additional sample info from higher up in the spectrum.
I must say, after reading the "tutorial" there is a lot unsaid, and a fair bit of "audiospeak" (aka "handwaving, using fancy phrases").
While this seems like a clever idea, it also seem like it is solving the wrong problem, and seems to ignore some important issues.
First, the human ear can't really hear much past 20 KHz. Most folks are hard pressed to hear 15 KHz. so reproducing signals up in the 50 KHz -100 KHz range isn't really what high sample rates are seeking to do. Instead it seems to me that the key issue is that placing the anti-aliasing filter at 22 (or 24) KHz is cutting this pretty close. No filter is ideal, so there will be some finite roll-off, and this will mean some higher frequency signal components may get aliased in at lower frequencies. Making the filter tighter can avoid this, but this generally introduces group delay variations (i.e dispersion), which can be easily heard as a smearing or muddling of the sound, specially the high frequency components of percussion instruments. So, while a 192 sample rate is nice, the goal is not to reproduce super high frequencies, it is to avoid superhuman filters (for antialiasing in the recording phase, and sample noise reduction in the playback phase). So, on its face the MQA idea seems to me to be barking up the wrong tree. I'd love to hear other's thoughts on this.
Second, part of the value of 24 bit HD audio files is that there are plenty of times when you ARE using that entire range. Live recordings are a good example, where audience noises are far below the music level, but a good system can bring them out, and make the entire recording feel amazingly real. Use those lower bits for "folding" or packing" higher frequency samples, and you are trading sonic detail in the listening band for sampling detail outside the listening band. Thoughts on this?
Lastly, the MQA process seems sort of mysterious. They claim that the "folding" is done in "two parts" and that the first "fold" can be unfolded in the software decoder. I believe that...since it is really just a matter of somehow tagging the samples to you know which ones have folded content, and then separating out those bits as higher frequency samples, while replacing them with interpolated samples from nearby non-folded samples (here they rely on the fact that supposedly these low level bits don't change much, so you can effectively under sample them, use the unused bits for something else, and then oversample them at the decode phase to recover them). Seems realistic, ASSUMING there really isn't mush variation in those bits. To the extent that there is, then you have lost resolution, and that will show up as lost detail.
And last-lastly, they then claim that the second unfolding step cannot be done in hardware, thereby necessitating, I presume specialized DACS that can perform this magic (while presumably returning license fees to the MQA inventors). One of the inventors (Stuart) claims that this step is an "analog to analog" process, which seems silly if it is a sample based thing. Does anyone here have any idea what goes on in the purported MQA capable DACs?
Looking forward to a lively discussion..
Scott
Last edited:
It's a money grab and a lossy manner that allows for embedding watermarks/etc. I understand why media companies want to move to it, but it's a coding scheme not for the consumer's benefit, no matter how much handwaving.
Lossless FLAC is smaller for the equivalent bit and sample rate and is free/open source. In all ways MQA is technically inferior.
Lossless FLAC is smaller for the equivalent bit and sample rate and is free/open source. In all ways MQA is technically inferior.
Yeah, I thought the whole thing seemed flaky.
There is something cleverly "folded" into your statement! "Lossless FLAC is smaller for the equivalent bit and sample rate"...
Meaning that the resulting "equivalent bit and sample rate" of MQA is less than FLAC...or the FLAC bit depth and sample rate that is equivalent to MQA is lower than the claimed 24/192, or 24/96.
In other words, Shannon doesn't lie, and you can't get something for nothing..you can tart it up and make it look like you are getting 10 Lb of music in a 1 Lb bag, but in the end, it's still just a 1Lb bag...
Cheers,
Scott
Last edited:
Nyquist too...
Just check what an audio recording industry does last decade...
Most so named “remastered” is just a more and more dynamical range compressed so each can easily hear difference between live reproduction and recorded.
Not matter the file coding scheme, lossless or lossy, not matter the codec/bit depth/sampling freq, not matter the DAC type or reconstruction filter envelope, not matter the amplifier class/power, not matter the speakers size/enclosure type/band count, etc...
Each must clearly understand what is payed for and no one label allow you to hear sounding like expensive show in your comparatively cheap headphones.
Carefully fullfill theorem conditions to pick benefit from its statements.
Other is leisure fuction and cut the cash.
- Status
- Not open for further replies.