Inside High-res audio: PCM vs MQA vs CD: 2L Sampler Comparison

[AudioSceptic]

Of course they are.

Yes, it was rhetorical. ;-)

 

[The Capstan]

Hi Amir.

great video as usual. It would be great to also assess MQA after decoding (let’s say unfolding) and compare with similar bandwidth PCM. Is it technically feasible?

best
Alessandro.

 

[fricc]

I've been listening to this interesting comparison, and I have to say that Amir's voice has a ton of clipping distortion... Maybe use a dynamic mic and more headroom? And/or a compressor?

 

[Hayabusa]

Make sure you read Sony's side of it too on their corporate site.

Personally, based on all the books and research I have done over the years, I don't trust a thing Philips say. Philips were lucky to have Sony, not the other way around.

Bang! That is a dent in my pride that I had for Philips.
I started to work for Philips Research in 1984 and just missed the CD introduction but have seen a lot of the people involved face to face.
I wonder why you have this opinion on Philips?

 

[Herbert]

Did you get the Sony demonstration CD in the box? The first run came with a CD because there were none (not easily) to buy. What is the production date on your machine? (it's on the rear, bottom of the external transformer housing)

Off topic, but the CDP-101 started the digital era in HiFi, so here it is -
No date given on the bottom of the transformer housing but I guess from
the high serial number a later production run. Could be a date is written on the pickup
No CD in the box.
I already had to replace the amps for the tracking motor and tray after one week.
Also the graphite based keyboards are dead whereas the tactile switches work. I had no time yet to build a board with
sub-miniature switches to replace the original. I was able to choose from 3 CDP-101 and allowed to check them,
as as one was on display for years and the shop owner did not remember which one.
Very interestingly, all three Sony CDP-101 had one or two almost unnoticeable
and identical scratches on the front panel.
Later, in a German forum someone confirmed his 101 had the same scratch.
- so something repeatable must have happened on the assembly line for years.
I would like to know what kind of varnish Sony used in order to cover them, a clear variant would be sufficient.
And where to tap 4.2336Mhz from the circuitry to add a digital output - looks like Sony used a PLL
for generating this standard frequency inside a CD-player that was later directly taken from an xtal,
and multiplied as systems started to use oversampling: 4.2336 in the beginning, then 8.4672 / 16.9344 / 33.8688 / 67.7376...

 

[Herbert]

Bang! That is a dent in my pride that I had for Philips.
I started to work for Philips Research in 1984 and just missed the CD introduction but have seen a lot of the people involved face to face.
I wonder why you have this opinion on Philips?

Well I was in short email-contact with Lou Ottens maybe four to five years ago,
he passed away this spring and sadly it took the press a week
to realise that the creator of the compact cassette (mixtapes anyone?),
Image Disc and Compact Disk left this world.
Anyway, he had no explanation why the famous swing arm got
scrapped by Philips. I assume money and outsourcing.
To add two other points:
Who was responsible for the fact that some connectors in Philips players
between p.c.bs could be mixed up as they had the same footprint?
Not possible with Sony, they would simply not fit.
Or for drawing schematics where voltages are referenced by a number,
so you need to look the voltage up on another page?
Just to make things clear, enough space to print a three-digit number
And I even think eyepattern voltage for the CDM-1 swingarm is stated
too high in the service manuals, but I am not sure on that any more.
So I fear for my life now as they have left home entertainment and concentrated
exclusively on the medical device industry...

 

[RichB]

In other hopeful news, there is a Roon thread where subscribers for QOBUZ and TIDAL are asking for a feature to select the higher quality (QOBUZ) over the lower quality (MQA 16 bit) edition.

Primary version choice between MQA and true resolution… - Roon Software / Feature Requests - Roon Labs Community


- Rich

 

[chris719]

Bang! That is a dent in my pride that I had for Philips.
I started to work for Philips Research in 1984 and just missed the CD introduction but have seen a lot of the people involved face to face.
I wonder why you have this opinion on Philips?

He seems to have a particular fondness for 80s Japanese gear.

 

[Grooved]

Hi Amir.

great video as usual. It would be great to also assess MQA after decoding (let’s say unfolding) and compare with similar bandwidth PCM. Is it technically feasible?

best
Alessandro.

Amir will certainly, but yes you can do it, especially if you compare 24/96 masters, you can have Qobuz 24/96 FLAC, and Tidal 24/48MQA which is processed only once to match the 24/96.
As the 2L files are MQA 354.8, I tried a different way and compared files not available to purchase by recording them and limiting to 24/96, you can do it while staying in the digital domain and it keep the exact content, even the MQA tag if you don't processed it by turning off the Core decoder.
Higher SR MQA files are a different case as they will need to be processed again so it can only be recorded from the analog outputs of a DAC and therefor brings a DAC-ADC conversion step.
For lower numbers, it will not really match as a 24/44.1 or 48 master will give a 24/44.1 or 48 FLAC file, and a 24/44.1 MQA but being MQA, it will be processed to 24/88.2, so not the same SR than the PCM one.
Maybe it would be interesting to check this ones too, as it should not add something above the original limit.

The test I've done was on Thomas Strønen - Bayou available as 16/44.1 and 24/96 from both Qobuz and Tidal

Qobuz 24/96 :

Tidal MQA 24/48 unfolded to 24/96 :

I also checked the bit depth in another tool (a bit like Musicscope) and added the MQA folded version, the red part are pure noise :
(I also tested Tidal MQA 24/48 folded and bit depth stays mostly at 16 with some peak at 17 or 18 during playback)

Qobuz 24/96 :

Tidal MQA 24/48 unfolded to 24/96 :

 

[amirm]

I've been listening to this interesting comparison, and I have to say that Amir's voice has a ton of clipping distortion... Maybe use a dynamic mic and more headroom? And/or a compressor?

These were produced back in 2017 with a cheap mic, noise gating, compression, etc. At that time I had 10 people who watched these videos so didn't matter. Newer videos have much better production qualities.

 

[frangle]

Someone asked why the high frequency spectrum in DSD is not filtered out. Reason is that it can't! If you digitize analog using a DSD encoder, that is what it is going to do. You cannot edit that stream as is. To take out the ultrasonics, you would have to convert it to PCM, then filter it. Problem is, if you are going to deliver it as DSD, then you have to add that noise back in! It simply is the nature of the beast.

So to answer @Hippocamp 's earlier question, it seems that if you have DSD files and you want to save storage space and get rid of undesirable ultrasonic effects on your system without noticeably impacting quality, you could convert to 192kHz 24bit FLAC to eliminate the DSD noise shaping headroom need without impacting dynamic range, then down-sample maybe to 48kHz 24bit FLAC to be left with something without audible compromise. Does this make sense?

 

[Herbert]

These were produced back in 2017 with a cheap mic, noise gating, compression, etc. At that time I had 10 people who watched these videos so didn't matter. Newer videos have much better production qualities.

I propose Surround. In some years algorythms
will certainly mimic voices, choose Alec Guinness‘ to make you sound the wise man you already are: „Luke, the Force is physics!“

 

[Frank Dernie]

Just watched the first part of the video and got to the part where @amirm explains 88kHz would be required to record the music.
Personally I found the microphones with the lowest noise (and best sensitivity) rolled off from 16kHz and I don't think I have ever owned speakers which went much above 20kHz and my ears went up to 16kHz last time I checked, a few years ago.
Since I couldn't hear it and my speakers couldn't reproduce it anyway is it still worthwhile to consider 88kHz as a worthwhile recording sampling rate?
I am aware a lot of instruments radiate pressure pulses at higher than audible frequencies but what is the down side of not recording them?
I write as somebody who has never tried to do so BTW.

 

[Grooved]

Just watched the first part of the video and got to the part where @amirm explains 88kHz would be required to record the music.
Personally I found the microphones with the lowest noise (and best sensitivity) rolled off from 16kHz and I don't think I have ever owned speakers which went much above 20kHz and my ears went up to 16kHz last time I checked, a few years ago.
Since I couldn't hear it and my speakers couldn't reproduce it anyway is it still worthwhile to consider 88kHz as a worthwhile recording sampling rate?
I am aware a lot of instruments radiate pressure pulses at higher than audible frequencies but what is the down side of not recording them?
I write as somebody who has never tried to do so BTW.

Amir would surely explain it in a better way, but, for one example, using more than 44.1 or 48 can be a way to be sure that any filter won't cut the audible range, and not having the goal to record and then hear frequencies above 20.05 or 24KHz.
Let's use for example a slow rolloff filter that needs 5kHz range for his full effect (from start of decrease to the end), if you use it on a 44.1 sample rate, it can start to cut off frequencies from 15.05kHz. If you use 48kHz SR, it can goes from 19kHz to 24kHz. As some people can hear a bit more than 19kHz, it can still cut off some musical information.
It would mean that if some people can hear higher than 19kHz, let's say 21kHz, you would need (21+5)x2=52kHz, so more than 48kHz.
Push it to a filter that use 10kHz (or the sum of different filters through the audio chain) to cut off and you have (21+10)x2=62kHz As we don't have a 52kHz or 62kHz SR, we need to use the following SR after 48 which is 88.2 (note that some hardware equipment support 64kHz SR)
This way, you are sure that any filter, even one using 10kHz range for his full effect, would be OK and still leave untouched the audible range

 

[Frank Dernie]

Let's use for example a slow rolloff filter

We can stop there.
A DAC with a slow roll off reconstruction filter is not complying with the theorem
In my 10 year old DAC evaluation there was clear audible differences between some reconstruction filters, yes, but only the ones that were "wrong".

 

[voodooless]

We can stop there.

Indeed. I have not seen a single slow roll-off filter that had decent attenuation in the stopband, so they all have significant aliasing. You could of course design a filter that is slow and has some decent stopband (let's say at least 60 dB), but you would need a quite high sampling rate to not touch the audible part.

.. what's the point of this again?

 

[Grooved]

We can stop there.
A DAC with a slow roll off reconstruction filter is not complying with the theorem
In my 10 year old DAC evaluation there was clear audible differences between some reconstruction filters, yes, but only the ones that were "wrong".

It's only one example and I actually did not have the DAC step in mind but the creation of the audio file we will later play through the DAC (and the whole creation from the recording, not just a new transfer from a analog master).
I was not talking about the filter of "listening" DAC, because if a master was created in 88.2, the reason for this choice is not that people have a DAC able to play 88.2 files.
The choice of releasing the 24/88.2 version (along the 16/44.1 version) is because people have DAC able to play this and are asking for that, but the choice to use 88.2 during previous step of the creation are not based DAC and people asking for that.
The question was "explains 88kHz would be required to record the music", and my answer is one example only, there can be other reasons to use 88.2, and sometimes other reasons to not to use 88.2

...You could of course design a filter that is slow and has some decent stopband (let's say at least 60 dB), but you would need a quite high sampling rate to not touch the audible part.
.. what's the point of this again?

If I'm not wrong, and even if you're talking about the "listening DAC" step, you're actually more confirming the answer I gave on "explains 88kHz would be required to record the music" more than confirming a comment stating the opposite of my answer, don't you think so ?

 

[Frank Dernie]

It's only one example and I actually did not have the DAC step in mind but the creation of the audio file we will later play through the DAC (and the whole creation from the recording, not just a new transfer from a analog master).
I was not talking about the filter of "listening" DAC, because if a master was created in 88.2, the reason for this choice is not that people have a DAC able to play 88.2 files.
The choice of releasing the 24/88.2 version (along the 16/44.1 version) is because people have DAC able to play this and are asking for that, but the choice to use 88.2 during previous step of the creation are not based DAC and people asking for that.
The question was "explains 88kHz would be required to record the music"

Well my first digital recorder was a StellaDAT. It used 16/48. It was the first recorder I had used where the output was indistinguishable from the microphone feed. I have higher sampling rate recorders now but am still not convinced they are capable of making an audibly superior recording to lower sampling rate, though I do use 24/96 as my default.

 

[Grooved]

Well my first digital recorder was a StellaDAT. It used 16/48. It was the first recorder I had used where the output was indistinguishable from the microphone feed. I have higher sampling rate recorders now but am still not convinced they are capable of making an audibly superior recording to lower sampling rate, though I do use 24/96 as my default.

This is one of the good examples that confirm using 88.2 or 96 is not mainly use to get and then hear frequencies above 20kHz, the main thing is how it sounds, and some can sounds more transparent with 16/48, or 24/44.1 than other with higher SR, and sometimes it can be the opposite.
Nothing guarantees you to get a more transparent sound with 88.2/96 than 44.1/48.
If you compare your StellaDAT to the StellaDAT II which can also do 88.2 and 96, it could be yours the more transparent one, but it would be interesting to compare (if there's not too much things changed between both versions)

The thing to keep in mind is that it doesn't only apply in the first "audio recording" step, it can also happen during mixing or mastering, where you can continue 100% digitally, or going through external analog process (which mean DAC then ADC steps), and if for any reason, the digital software/hardware or the DAC-ADC converters you're using works best/have more stability/... at 88.2, you will decide to do the first recording at 88.2, and there are certainly more 88.2 or 96 audio files that were made available for that reason than for the goal of providing higher than 20kHz to the listeners.

It's also only one part of the technical audio chain, and other parts like choosing the right microphones, the right placement, the good mic preamps... will be more important than the difference between your best and second or third converters, but it's a different story.

 

[Frank Dernie]

This is one of the good examples that confirm using 88.2 or 96 is not mainly use to get and then hear frequencies above 20kHz, the main thing is how it sounds, and some can sounds more transparent with 16/48, or 24/44.1 than other with higher SR, and sometimes it can be the opposite.
Nothing guarantees you to get a more transparent sound with 88.2/96 than 44.1/48.
If you compare your StellaDAT to the StellaDAT II which can also do 88.2 and 96, it could be yours the more transparent one, but it would be interesting to compare (if there's not too much things changed between both versions)

The thing to keep in mind is that it doesn't only apply in the first "audio recording" step, it can also happen during mixing or mastering, where you can continue 100% digitally, or going through external analog process (which mean DAC then ADC steps), and if for any reason, the digital software/hardware or the DAC-ADC converters you're using works best/have more stability/... at 88.2, you will decide to do the first recording at 88.2, and there are certainly more 88.2 or 96 audio files that were made available for that reason than for the goal of providing higher than 20kHz to the listeners.

It's also only one part of the technical audio chain, and other parts like choosing the right microphones, the right placement, the good mic preamps... will be more important than the difference between your best and second or third converters, but it's a different story.

Yes, I know.
Microphone position makes far more difference than the recorder sampling rate (or even what sort of recorder it is).
Having a higher bit-rate makes level setting easier (pretty well completeluy fool-proof in fact) and gives more latitude in subsequent manipulations, if you do them, so I do see a point in using 24-bit rather than 16-bit particularly for novices in level setting.
What I have still yet to see any convincing explanation of a way higher sampling rates could make an audible difference to music.
My StellaDAT won't boot any more. I suspect the EPROM was wiped due to the battery discharging whilst in storage (stupid of me to leave it connected, I know). Do you know anybody who could burn me a new EPROM or could tell me if that might be the problem? I did consider a StellaDAT II but went solid state instead, I don't need the StellaDAT but would like very much to get it working again.