• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

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

Frank Dernie

Major Contributor
Forum Donor
Joined
Mar 24, 2016
Messages
5,621
Likes
12,615
Location
Oxfordshire
#61
I partially disagree with the first statement. It really all depends on the type of music/content (e.g., opposite of Pop music), how well it was recorded and mastered (e.g., if they record & master with 16/44.1 in mind the whole way through it's likely the hires variations will be hampered), and then listening environment and playback equipment. I argue a great DAC and amp (we know these exist thanks to @amirm), good noise isolating headphones or sensitive IEMs and/or a quiet room in the evening, and the right content is sufficient to hear an advantage over 16bit content. Heck, the noise floor advantage alone with 24bit content played through sensitive IEMs might be enough even if it's just to minimize the hiss. But to your point, the majority won't need more than 16/44.1. But then again, if you're going to start adding in noise and distortion (through your choice of DAC + amp + speakers/headphones + environment) wouldn't you want to start with the best copy available so you get the *most* out of your (whole) system (even if it's just at or near 16bit reproduction quality)?

Analogy attempt: A photocopier can only do so much, but making a copy (playback) of a copy (16bit source) can't be better than the original (playback from 24bit source). Sometimes you can't really tell (printed text), other times you can when compared directly to the original, and others it's obvious from the get go (a photo/image perhaps).

Second part, partially related to the first, but considering how home theater equipment lags significantly far behind good *music*-oriented products this is even more likely.
I have recorded, mainly classical, music for well over 50 years now.
I have never heard of any music with wide enough dynamic range to tax 16-bit. Back in the old reel-to-reel tape days it required quite a bit of skill to get the levels set so that the quiet bits were not in the noise and the loud bits were not overloaded too much on classical music. The first 16-bit recorder I bought setting the levels was easy so there was no clipping and the background noise is, in any case, so low you can't hear it at all at normal listening level. It was also the first time where I couldn't hear a difference between recorder and microphone feed.

The reality is that you almost certainly haven't ever had a recording with 16-bits of dynamic range, so I believe your premise is invalid.
I do have a 24-bit recorder now. 24-bit makes it totally idiot proof to set levels, but there isn't any music that needs it, it may help at the recording stage if something unexpected happens, that is all. All the music on a recording fits into a 16-bit window, usually very easily, so all the other bits are a pointless waste.

There would be no point in releasing recordings which had so much dynamic range that either the quiet bit was inaudible and the loud bits cause hearing damage! Even if a recording of a full 16-bit dynamic range existed if you set it so you could hear the quiet bits the loud bits would be painful.
 

Frank Dernie

Major Contributor
Forum Donor
Joined
Mar 24, 2016
Messages
5,621
Likes
12,615
Location
Oxfordshire
#63
If I have well understood, a 24 bits records has a better dynamic and a lower noise floor than 16 bits files. And this is probably more 'relevant' than high Khz records because, as you wrote, you probably can't hear past 20 KHz.
Well 16-bit is 96dB of dynamic range. If you have a quiet listening room, 30dB say, and you set the level so the quietest sound a 16-bit recording could contain was audible the peaks would be 126dB, which could cause hearing damage if your HiFi could do it which it almost certainly can't.
Luckily music doesn't often have 16-bits of dynamic range and no commercial recording is going to be released with that much anyway since 99% of the world's population wouldn't be able to play them and would want their money back ;)
 

Frank Dernie

Major Contributor
Forum Donor
Joined
Mar 24, 2016
Messages
5,621
Likes
12,615
Location
Oxfordshire
#64
My pleasure. On your conclusion, these are just a few tracks out of millions available. What I am reviewing here is production quality and standards for the most part, and the formats as a supporting role. In general, it is easy to show that original productions exceed the bandwidth of CD in many high-res recordings.
The fact that it has a wider bandwidth is no surprise, it has been known for decades that a lot of metal percussive instruments emit inaudibly high frequencies.
That doesn't mean many people can hear it.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
33,645
Likes
117,721
Location
Seattle Area
Thread Starter #67
Well 16-bit is 96dB of dynamic range. If you have a quiet listening room, 30dB say, and you set the level so the quietest sound a 16-bit recording could contain was audible the peaks would be 126dB, which could cause hearing damage if your HiFi could do it which it almost certainly can't.
It is only 96 dB if you don't use dither. Depending on the type you use, you will loose 3 to 6 dB. Room noise is also dominated in bass frequencies where our hearing is not very sensitive. Where we are (2 to 5 kHz), there are many rooms that are audibly silent. Room noise is also omni-directional making it less objectionable and audible than what comes out of a speaker.

And of course we also have headphone listening where noise isolation can be provided and getting high SPL is incredibly easy.
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
1,480
Likes
2,101
Location
Netherlands
#68
It is only 96 dB if you don't use dither. Depending on the type you use, you will loose 3 to 6 dB.
Actually, where it counts you’ll actually gain dynamic range when adding dither.

Note: I don’t know if YouTube actually preserves all of the dithered audio when compressing.. I can’t hear anything in the last 1/5th anyway ;)
 
Last edited:
Joined
Jan 28, 2019
Messages
9
Likes
6
#70
Because we encode the signal levels over time, not frequency. There isn’t such a thing as frequency x = null. Well.. there is when doing a Fourier transform, but that means processing the data.



Yes, Shannon proved that a long time ago: you don’t need more that double highest frequency you want to encode (and a bit of headroom for the filters).

People also buy a 8k 40” screen and watch it from 5 meters, and the market also peddles that kind of nonsense.

@amirm maybe it’s time to make a proper video on digital audio sampling, the sampling theorem and the proof of it.

OK, so I guess I'm dense, too. :) It is perfectly possible to filter in the frequency domain. So why not just filter out the high frequency noise from hi-res files and convert back to the time domain? Is production just too lazy to do that, or are there other considerations?
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
1,480
Likes
2,101
Location
Netherlands
#71
OK, so I guess I'm dense, too. :) It is perfectly possible to filter in the frequency domain. So why not just filter out the high frequency noise from hi-res files and convert back to the time domain? Is production just too lazy to do that, or are there other considerations?
That’s what you do if you convert to 44.1 or 48 kHz. You could also do that with your 96 kHz file of course, but then @amirm would be on top of it and complain about all of that empty spectrum and how they once again robbed us :facepalm:;).
 
Joined
Jan 28, 2019
Messages
9
Likes
6
#72
That’s what you do if you convert to 44.1 or 48 kHz. You could also do that with your 96 kHz file of course, but then @amirm would be on top of it and complain about all of that empty spectrum and how they once again robbed us :facepalm:;).
Got it, that's what I thought. Downsampling will "null" all frequencies above half the sampling rate, eliminating the noise. One thing that wasn't clear to me in Amir's video, though, was the idea that very high sample rates allow you to "move" noise from the audible band to inaudible high frequencies (not exactly sure how that would be done). But after doing so, and assuming this improves the signal in the audible band, why not then just downsample to 48K and get rid of the HF noise?
 

voodooless

Major Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
1,480
Likes
2,101
Location
Netherlands
#73
Got it, that's what I thought. Downsampling will "null" all frequencies above half the sampling rate, eliminating the noise.
Well, not exactly. The high frequency content is just not there, so it’s not even “null”.

One thing that wasn't clear to me in Amir's video, though, was the idea that very high sample rates allow you to "move" noise from the audible band to inaudible high frequencies (not exactly sure how that would be done).
This is a bit more complex, but in essence, yes. An ADC can use noise shaping to move noise up in frequency so the audible band is clean. But so does a DAC if you play the file back again. The trouble starts if the ADC cannot push the noise far enough up and it is still present in the samples audio, like one can see in those 384 kHz examples.

But after doing so, and assuming this improves the signal in the audible band, why not then just downsample to 48K and get rid of the HF noise?
Because then you can’t sell it as high-res audio anymore :facepalm:
 

jam

Member
Joined
Feb 28, 2020
Messages
54
Likes
80
Location
Montreal
#74
Another video in this series analyzing objective quality from the label 2L in PCM, MQA and CD.
Thanks for yet another fascinating video with the analysis Amir. My first Hi-Res Blu-ray audio came from the 2L label. It's great that they offer this free service to download sample tracks in different encoding formats and sample rates/bit depths so people can perform tests.
 
Joined
Jan 28, 2019
Messages
9
Likes
6
#76
Well, not exactly. The high frequency content is just not there, so it’s not even “null”.



This is a bit more complex, but in essence, yes. An ADC can use noise shaping to move noise up in frequency so the audible band is clean. But so does a DAC if you play the file back again. The trouble starts if the ADC cannot push the noise far enough up and it is still present in the samples audio, like one can see in those 384 kHz examples.



Because then you can’t sell it as high-res audio anymore :facepalm:

Ahhh... All becomes clear. Thanks! So, the benefit of very high resolution is at the recording/production end and not necessarily at the reproduction end. If a track is well mastered, then the benefit of "noise shaping" in production will be just as present in a traditional 44.1K or 48K file as in a DSD file. And DSD can't improve a poorly mastered track. Conclusion: don't waste money on hi res audio.
 

Herbert

Active Member
Joined
Nov 26, 2018
Messages
274
Likes
239
#77
@Amir, I think the 19kHz Tone is already in the 24/96 recording, watch your video from 3:44.
There it looks correlated but it sticks out in the quieter passages.
First I was a bit shocked to see that 60kHz would be a sensible sample rate for natural instruments and voices.
But on the other hand in the example the overtones were already at-80db around 20kHZ and down between -110 and -130.
So I guess CD ist still sufficient...
 

Don Hills

Addicted to Fun and Learning
Joined
Mar 1, 2016
Messages
679
Likes
407
Location
Wellington, New Zealand
#78
So, Amir described that an advantage of DSD is that it pushes noise from the audible band into higher frequencies. So, why not, as a next step in digital mastering, simply filter out everything above ~40-50Kz and then downsample to a lower rate? Or, perhaps that is just how well-mastered music is produced in the first place...? I guess I'm just not following why you would leave all that inaudible high frequency stuff in the file.
DSD has a SNR (difference between maximum level and noise level) of only 6 dB. This is because it has a bit depth of 1. The greater the bit depth, the greater the SNR, as others have explained. You can improve the SNR by increasing the sample rate or increasing the bit depth. DSD does it by increasing the sample rate. The noise is still there, it just gets moved up into the ultrasonic range. If you filter out the ultrasonic and downsample, you have to increase the bit depth to maintain the SNR. "A short, wide container holds as much as a tall, narrow container."
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
33,645
Likes
117,721
Location
Seattle Area
Thread Starter #80
Actually, where it counts you’ll actually gain dynamic range when adding dither.
That would be getting something for nothing. Dither randomizes the LSB so by definition, you lose bits and dynamic range. You can noise shape that so that in the audible band the effect is not there but then you better have ultrasonic spectrum to park the noise.
 
Top Bottom