Computer Audio Asylum

Music servers and other computer based digital audio technologies.

Return to Computer Audio Asylum

Message Sort: Post Order or Asylum Reverse Threaded

2L MQA Files

213.7.50.30
Posted on February 7, 2016 at 05:58:24
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 I compared the 352k hires and 44.1k MQA files and it was NOT a jaw dropping experience. If anything, MQA sounds worse in that tonal colours were lost in translation. In passages MQA was clearer but timing does not sound 'right'.

Downsampling and dither must play a part in the SQ of the MQA files

 

Hide full thread outline!
    ...
RE: 2L MQA Files, posted on February 7, 2016 at 08:00:01
ahendler
Audiophile

Posts: 5151
Location: San Antonio, Texas
Joined: January 24, 2003
 Were you decoding the MQA files?
Alan

 

RE: 2L MQA Files, posted on February 7, 2016 at 08:17:01
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 No, but I thought that this was not necessary. Otherwise, why would they put them up?

 

RE: 2L MQA Files, posted on February 7, 2016 at 08:19:04
PAR
 I believe that you were listening to the MQA file undecoded. If so then your findings effectively agree in ranking with 2L's own listening notes (from Morten Lindberg).

The 2L ranking for undecoded MQA files is that the DXD source is first, MQA second and CD resolution third in preference.

So it would seem that the logical next step would be for you to compare the undecoded MQA file with the CD resolution version that 2L have posted.

I look forward with interest to your further posting.

 

RE: 2L MQA Files, posted on February 7, 2016 at 08:25:40
PAR
 "why would they put them up?".

Presumably because MQA decoders became available to the public from last week.

Of course the comparison of undecoded MQA files with other formats is also of interest as presumably single inventory titles will be issued and compatibility with existing formats , especially 16/44.1, is of significance. I don't expect a rush to buy new DACs.

 

RE: 2L MQA Files, posted on February 7, 2016 at 09:12:53
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 With hdds cheap and hires available, there seems little reason to buy a decoder.

Meridian's own 582 offers a significant improvement in CD resolution files.

I wonder if they are trying to recycle some of the technology which was very good but didn't take off.

 

Streaming, posted on February 7, 2016 at 09:33:28
bullethead
 I think it is for streaming.

We'll see what they come up with. If I can get a Sonos solution and perhaps an external DAC and Tidal or Deezer or what have you for $20 a month then it might be worth it, of course the software meaning content should also be available in mass quantities.

Give me Rick Ross "black dollar" in that format, give me remastered tape transfers from esoteric labels , we will see. Things just move so damn slow, you think this is NASA making this crap. Where is all this stuff? You move to South Korea they wouldn't put up with downloads and multiple year waiting sessions to have this happen.

 

RE: Streaming, posted on February 7, 2016 at 10:22:18
PAR
 " You move to South Korea they wouldn't put up with downloads and multiple year waiting sessions to have this happen."

Oh, they do. I was aware of the Koreans working on 24/192 audio for commercial exploitation in, what, 1998 or 99? Have a look at the period of gestation and ongoing journey to market for OLED TVs (all OLED panels being Korean). Hardly overnight.

Actually I think Meridian/MQA have done pretty well in moving from announcement to actual product on the market (albeit very limited indeed) in one year. Especially as they have to persuade other companies to make the software.

 

RE: Streaming-until, posted on February 7, 2016 at 23:01:20
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 I can hear it properly, I shall regard it as another attempt towards monopoly. This is so wrt Meridian's history of leading the pack to kill sacd and failing to successfully promote DVD audio formats.

Why the heck do we need another propriety format which restores resolution lost in 'air'? If the format is open source and sounds better than CD, let's have it. If not, it's another attempt to make more money, needed or not.

Meridians 582 software actually works on CDs like a treat, but they failed to exploit it.

 

Why not, posted on February 8, 2016 at 16:18:40
E-Stat
Audiophile

Posts: 37650
Joined: May 12, 2000
Contributor
  Since: April 5, 2002
 and CD resolution third in preference.

just use higher open system resolutions than 44/16? 96/24, 88/24, 192/24, DXD all exist today.

I'm with others in not understanding why we need yet another (proprietary) format.

 

RE: Why not, posted on February 8, 2016 at 19:57:44
PAR
 In terms of MQA being just another file format you are right. However MQA is not offered as such but as away to stream or to otherwise deliver the equivalent of the hi-rez file resolutions you have listed but within a low-rez wrapper. Its other purpose is to offer a way of improving the end to end accuracy of the recording process including legacy material (they say).

As I believe that Joe Public doesn't really give a hoot one way or the other I think that its commercial success in the consumer market will mirror that of better than MP3 streaming. I remain doubtful of the latter long term if it involves a significant premium over "standard" subscription costs or an additional investment in hardware.

 

deliver the equivalent of the hi-rez file resolutions , posted on February 8, 2016 at 21:56:12
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 we can all dream but this seems to be about monopoly.

There is o way that 44.1k sample rate files (many downsampled) can be equivalent to hires. How many times have folks tried before?

 

RE: deliver the equivalent of the hi-rez file resolutions , posted on February 9, 2016 at 01:23:22
PAR
 "
There is o way that 44.1k sample rate files (many downsampled) can be equivalent to hires".

For accuracies sake that is not what MQA claims. MQA is a kind of compansion system where the components of files larger than 44.1 are buried beneath the noise floor of a 44.1 file to be "exhaumed" via use of the decoder. Please see their simple explanation in the URL below. So the 44.1 MQA file is not equivalent to a hi-rez file , it actually "contains" the entire hi-rez file.

I do not know what you mean by "monopoly". Any record producer or streaming service is free to choose to use it or not (my expectation is "not" for the majority). If it is used then it is not necessary to decode the MQA file in order to listen to the music at redbook resolution (which most of the world seems to be happy with).

 

Sounds like black magic !, posted on February 9, 2016 at 04:01:44
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 ''MQA is a kind of compansion system where the components of files larger than 44.1 are buried beneath the noise floor of a 44.1 file to be "exhaumed" via use of the decoder.''

So the entire hires file is buried below -94dB!

This is not what the papers posted on the subject say.

 

RE: Sounds like black magic !, posted on February 9, 2016 at 04:10:52
PAR
 Not quite. Only the parts of the hi-rez file that exceed the part which is already contained within the 44.1 bandwidth are contained beneath the nouse floor. See the elegant explanation given by Bob Stuart in the MQA " How it Works" video as per the link in my last posting.

 

RE: Streaming-until, posted on February 9, 2016 at 04:19:13
bullethead
 we already have very high-res FLAC, as well as all recording tools to record in high-res PCM and export to PCM/WAV/FLAC.

We have the bandwidth too. If can stream Netflix, I don't see this streaming high res 32bit FLAC/etc... as any different.

I agree with you 100%.

I am a big "open source" fan. It is the reason the internet took off as it did. Could you imagine everyone running Windows web servers? LOL!

 

RE: Sounds like black magic !The link, posted on February 9, 2016 at 05:48:21
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 tells but explains nothing in concrete terms.



 

RE: Sounds like black magic !, posted on February 9, 2016 at 07:55:28
Tony Lauck
Audiophile

Posts: 13629
Location: Vermont
Joined: November 12, 2007
 Bits that "you don't need" are stolen from the original 44/24 file and replaced with the bits "you need" that were in an 88/24 source file. The presumption is that the bits "you don't need" are unnecessary (e.g. the low order 8 bits) and that all the high frequency bits "you need" will fit into the space previously occupied by the stolen bits. This might work better than a 44/24 downsample of the original if the stolen bits are actually less valuable than the benefits of the high frequency bits that replaced them. (Scrambling techniques are used to ensure that changing the stolen bits sounds like added noise when playing back without a decoder.) This might be a net gain in sound quality if one is limited to 44/24 bandwidth, however it can only be a net loss in sound quality compared to the 88/24 original without MQA. In other words, this portion of the MQA story amounts to data compression, not sound quality improvement. Meridian decides which bits "you need". They are no help to audiophiles who want all the bits that were recorded.

The other sound quality claim concerns filters. The claims are sufficiently ambiguous, but the impression is that there is a magic filtering process that can automatically improve older 44/16 recordings and that the MQA encoding process does this. In reality, any old 44/16 recording can be remastered into a higher resolution format and this can improve (some) bad choices of filters. This can even be done by the consumer by selecting appropriate upsampling filters. So the most that can be said about this part of the marketing is that by buying a new DAC you can avoid fiddling around with filters in an existing DAC or computer player. Of course, demos are easily concocted that compare bad older masterings with new remastering in to the MQA format. And one can find record label owners and DAC manufacturers saying how great MQA sounds in the hopes that they will sell more recordings and DACs.

Finally, there is the "authenticated" part of the deal. It seems the main benefit of any authentication is to MQA, in that it creates a captive market (sales, lease or a service business) for MQA encoders and encoding. In addition, because of the use of encoding, laws such as the DMCA in the US, come into effect that make it illegal to build decoders that aren't authorized by the MQA company.

Because the encoding authentication depends on the secrecy of the encoding keys the encoder needs to be specially packaged in a secure box and the encoder comes with a corresponding high price. This means that small record labels will not be able to use MQA economically, instead they will be cut out of the system or have to use a MQA encoding service. If MQA succeeds in the marketplace, the effect will be to make it more difficult for small record labels to compete against larger labels.

It is highly questionable whether users will receive any benefit from MQA's authentication. It will protect the content from the distribution chain, but it will not protect the content from manipulation prior to its being encoded by MQA. In no sense is there "end to end" integrity, since this would have to begin at the recording studio and ends in the listening room.


Tony Lauck

"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar

 

RE: Sounds like black magic ! If this is what MQA is, posted on February 9, 2016 at 08:41:38
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 what was all that hot air about restoring lost information in air in the paper referred earlier, which is supposed to be a secret sauce in MQA.

 

RE: Sounds like black magic ! If this is what MQA is, posted on February 9, 2016 at 08:54:45
Tony Lauck
Audiophile

Posts: 13629
Location: Vermont
Joined: November 12, 2007
 Can't answer. I'm not sure which paper you are talking about.

Tony Lauck

"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar

 

RE: Sounds like black magic !The link, posted on February 9, 2016 at 21:31:27
Jaundiced Ear
 Don't you just love audio marketing?

JE

 

RE: 2L MQA Files, posted on February 10, 2016 at 06:48:12
Thorsten
Manufacturer

Posts: 4209
Location: Somewhere nice on planet dirt
Joined: September 25, 1999
 Hi,

I think to make any call we would need:

1) Original 384k/24(32) bit Master

2) 48k/24Bit MQA file encoded from and then decoded back to 384/24(32); as the way I read it, it is being claimed full MQA decoding can restore the original 384/24 file lossless, if this is incorrect I will happily be corrected and apologise in advance for my misreading.

The intermediate 48/24 MQA file is a bit pointless without decoding, unless it is claimed that the 48k MQA file deliver 384/24 performance directly (which is not my reading of the promotional material).

More crucially, I'd like to see how MQA does with 0dBFS peak noise-loading over a bandwidth of 50kHz.

For anyone who does not know noise-loading, it uses uncorrelated white noise with peaks at 0dB(FS) comb filtered (visually speaking notches cut into the otherwise flat FFT spectrum) to the nominal noisefloor of the format. This can only really be digitally generated if we want that).

If we look at the analogue spectrum after D-A conversion how much the "notches" filled in tells us about noisefloor and noisefloor modulation. Of course the same applies to D-D conversion (interesting to see what "170dB dynamic range" ASRC's do in this) except the job is way easier that way, no DA/AD.

If we take FLAC, we know (well, I know having done so) it can pass the above tests with no alteration to the source file (presuming the sample rate is high enough, read after decoding we have precisely what went into the encoder. Actually it also passes a binary compare meaning the data is bit for bit identical.

Only lossless codecs have so far been able to do that, MP3 even at 320k leaves a gross observable difference (a null test leaves around -66dB residue for MP3 320k and -52dB for MP3 @ 192k VBR).

The price paid for using lossless codec's is that compression rate is not great, a sample 384k/24 FLAC from 2L shows around 2.75 compression factor.

Now if Meridian have achieved the same feat (namely lossless compression of a 384k/24 file so it can be completely restored) while getting a compression ratio of 16.7 once FLAC encoded, this is a huge achievement.

I mean we are talking a reduction from 51MB/min to 11MB/min, that is packing 384k/24 into a data-stream that is about twice the classic CD standard 44.1k/16 uncompressed and two thirds the file size for 96k/24 uncompressed.

So I am looking forward to the public release of the encoder/decoder (even as cripple-ware with silences inserted every other minutes or such) in order to verify this extraordinary feat.

Indeed, if Meridian really want to show off what their system can do and achieve the widest possible adoption, releasing such trial-ware encoder/decoder would likely do much to convince both those indifferent and the sceptics.

Ever which way, good on Meridian for pulling it off.

Of course, this says jack about the commercial side.

For now MP3 (Spotify and most other streaming services) and AAC (Apple iTunes/Music) rule streaming delivery and they seem adequate for > 90% of the people who listen on Dr Dre Beats or Apple Earpods via their phones or Laptops.

And even if the rest of the people were to pay triple for getting MQA (which they are unlikely to do) it does not equal anywhere near the revenue generated from the "90%".

And going from 320k MP3 / 256k AAC to MQA will mean a huge infra-structure investment and increase in bandwidth used (around 4 - 5 times the disk-space/cpu-load/bandwidth per MQA track over MP3/AAC) for streaming services. And all of these (disk-space/cpu-load/bandwidth) equal cold hard cash expended.

Without necessarily equal (or greater) increases in revenue potential than what is spend, I doubt services will clamour for MQA, so I think we have to watch this space.

Now as for the "improved filter response" or whatever is being applied in the re-mastering I see no direct requirement to have this tied into the MQA codec. Moreover it seems a pretty simple re-filtering using something related to apodising filter Meridian have been championing, based on my reading. And this is available with a wide range of hardware (both Meridian and not) anyway.

But again, there is way too little information to be conclusive in any way.

Thor

At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to intolerable pain. Really, what more could we need?

 

It is a lossy process. It uses perceptual coding just like MP3., posted on February 11, 2016 at 13:46:00
meisterkleef
Audiophile

Posts: 446
Location: Northest USA
Joined: January 31, 2003
 The point as I see it is not to be able to recreate a bit-perfect copy of the original hi-res file, but rather to create a version that sounds much better than a 44/16 downsample with roughly the same bitrate.

IMHO a very "iffy" proposition.

 

RE: It is a lossy process. It uses perceptual coding just like MP3., posted on February 11, 2016 at 16:48:20
Tony Lauck
Audiophile

Posts: 13629
Location: Vermont
Joined: November 12, 2007
 Not the same 44/16 bit rate. The bit rate would be more like 44/24. And if you compare you will find that a FLAC compression of the 44/24 will compress better than a FLAC compression of the MQA 44/24. So there isn't even 1/3 saving in size. It may sound better than the 44/16 (or FLAC 44/16) version after you have bought a new DAC. But it CAN NOT sound better than an original hi-res file (unless of course it's the new DAC, and not the MQA sauce).

Tony Lauck

"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar

 

RE: It is a lossy process. It uses perceptual coding just like MP3., posted on February 12, 2016 at 01:24:48
Thorsten
Manufacturer

Posts: 4209
Location: Somewhere nice on planet dirt
Joined: September 25, 1999
 Hi,

> The point as I see it is not to be able to recreate a bit-perfect
> copy of the original hi-res file,

Well, that is not my reading of what is being presented here:

http://www.stereophile.com/content/ive-heard-future-streaming-meridians-mqa#hguKuhPFiWHo2Eq2.97

In fact, in this article I see:



To me it looks like a lossless codec is claimed which on top of this uses what has been "origami like folding" of ultrasonic content into the lower bits and if my reading is right has been claimed that a 48k/24Bit MQA file can be unfolded to 192kHz sample rate and I remember reading that hardware MQA decoding is supposed to "unfold" to 384kHz - but I cannot find the reference yet..

https://www.youtube.com/watch?v=BrgjycGhoSM

> but rather to create a version that sounds much better than a
> 44/16 downsample with roughly the same bitrate.

We need to separate the application of a different digital filter from the Codec I think.

> IMHO a very "iffy" proposition.

Well, applying a minimum phase apodising digital filter on playback sounds almost as good as using no filter at all and to me much better than typical digital filters.

So this part, if MQA embeds such a filter in the specification (and possibly provides pre-emphasis for the treble loss from such a filter in mastering) could be a a pretty serious proposition.

Would it be better than a copy down-sampled with a standard SRC and then having a hardware apodising minimum phase filter applied? Or by just bypassing any digital filter? As it so happens I have DAC's at hand that offer these options.

Here is the impulse response using filterless operation:



and here it is using what is in effect minimum phase apodising:



We will have to hear and test.

As said, I am especially interested in a wideband noise test of MQA.

So lets hope Meridian release a software encoder/decoder for evaluation purposes soon.

Thor

At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to intolerable pain. Really, what more could we need?

 

Sounds a bit like HDCD, posted on February 12, 2016 at 08:52:54
Feanor
Audiophile

Posts: 9868
Location: London, Ontario
Joined: June 17, 2003
Contributor
  Since: March 12, 2004
 HDCD, as I understand, tried to squeeze 20 bits of dynamic range into 16 bits where the 20 could be extracted by decoding.

I'm thinking MQA is trying to squeeze 88.2+kH resolution into 44.1 (using 24 bits), again, to be extracted by decoding. I have know idea how that could possibly work, though I suppose there's an explanation somewhere.


Dmitri Shostakovich

 

RE: Sounds a bit like HDCD, posted on February 12, 2016 at 11:08:31
Tony Lauck
Audiophile

Posts: 13629
Location: Vermont
Joined: November 12, 2007
 The number of bits required to represent a signal at a certain accuracy depends on the bandwidth and the resolution. With regular PCM all frequency bands are given the same bit depth up to half the sampling rate. However, it is possible to reorganize the original signal into multiple frequency bands and encode these separately, then take the resulting pieces and get back the original signal. (In some cases, the exact same original signal, but I don't know if/when MQA does this.) This basic ability then allows the separate bands to be allocated different resolution by some clever signal processing techniques that have been used since the 1970's. By reducing the bits allocated to one band, bits are available to provide increased allocation to other bands. So, for example, if one sends regular 44/24 PCM one uses all of the available bits equally for frequencies from 0 to 22050 Hz, and there are no bits available to encode frequencies from 22050 to 44100 that would be available in an 88/24 carrier.

MQA robs bits from 0 to 22050 Hz and uses these to provide some resolution in the range 22050 to 44100Hz. This (supposedly) can be done in such a way that no distortion is introduced, just random noise that is not correlated with the music. The noise above 22050 can be fairly loud (supposedly) and still not be perceived.

This process of robbing Peter to pay Paul doesn't do anything to the sound quality that started in 88/24, but it could (possibly) produce better use of a channel capable of running at 44/24 but not 88/24. However, this analysis gives an unfair comparison of the actual bandwidth requirements for the 88/24 and the 44/24 recordings, because it doesn't take into account the possibility of lossless compression to these bit streams. It turns out (and can be seen by looking at example files on the 2L test site) that the amount of FLAC compression to 44/24 PCM is significantly more than the amount of FLAC compression to the 44/24 MQA encoded streams. Comparing a FLAC encoded 88/24 stream with the FLAC encoded 44/24 MQA stream shows that the entire MQA process hasn't really saved that many bits, but it has messed with the sound based on questionable psycho-acoustic research.


Tony Lauck

"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar

 

RE: Why not - because, posted on February 13, 2016 at 01:15:22
fmak
Audiophile

Posts: 13158
Location: Kent
Joined: June 1, 2002
 this doesn't make more money for some companies that try to reintroduce monopolistic technologies for captive users.

If one looks at the history of science, this occurs infrequently and it is difficult to predict the chances of success.

 

Page processed in 0.039 seconds.

[ Top Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2024, Audio Asylum, all rights reserved ]
[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]