Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 19:08:09 GMT
After having read and seen several people saying that some older, and praised, DAC chip designs seem to show their best at 44.1Hz (Red Book) quality, i decided to try this myself. I also saw someone claiming this for my specific DAC, that it was better with 16/44, but i was sceptic.
I have a boutique built DAC, built on the Burr Brown 1795 (if that matters), that i have been running with Qobuz for a while, using the native Qobuz resolution.
My streamer has the ability to down sample to a fixed resolution, like 44.1Hz, so i decided to try this earlier this weekend.
The result for me is emotions, involvement, fluidity, love! I don´t think i have used the vague term of "musicality" before but that is what comes to mind.
I put some thought into this and here is what i think, and would like to discuss:
First of all, I want to think that i have a relatively low ethernet noise floor by now.
These older DAC designs, that can handle High-Res, are not originally developed for High-Res and gets sort of "stressed" with handling the data. This is not really noticed as an drawback when you have a normal noise floor but when you start to remove noise the DAC´s own noise floor/limitation eventually becomes the bottleneck. But when downsample to their preferred resolution these old DAC chips works at its best and excels on what they do best, and what people love about them; timbre and tone ("musicality")
After having lowered the digital noise floor in many different steps, prior to this, i felt that the impact was similar here. Except for better timbre and tone i also got less sibilance and more dynamics (partly because i can crank the volume more with less sibilance).
Honestly, this was one of my bigger improvements. Not as obvious, since it is actually hard to detect when going back and forth between the resolutions, but the body and brain reacts to the music (instead of my ears).
I never before felt the same need to just listen, and listen, and listen (just one more album) as i now do.
Has anyone else felt something similar when downsampling?
|
|
|
|
Post by MikeMusic on Sept 15, 2024 19:15:05 GMT
I need to pay more attention to Hz. Currently on 44.1, Underworld 1992-2012
Think my limitation on getting better resolution will either be the file or my less than perfect broadband or both
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 19:38:06 GMT
It is interesting that someone like Peter Qvortrup of Audio Note is claiming that Sony decided on the Red Book 16/44 quality because that is the mathematical perfect balance between resolution, noise and dynamics (i believe it was).
When you go up in resolution you will sacrifice one of these aspects, he claims.
Many that has heard Audio Note gear seem to agree that it is very musical but not very technically impressive. I think i look for musicality (emotions and engagement) rather than technicality (resolution and razor sharp imaging), if having to select one over the other.
I am not saying that you can get both! But maybe you need to invest serious money to get everything?
Or, you actually will sacrifice one aspect?
|
|
bencat
Rank: Quartet
Posts: 353
|
Post by bencat on Sept 15, 2024 19:39:08 GMT
The strong view a few years ago was that the best sound is always at the recorded level used . So for things at 44.1 then replay at 44.1 is the best sounding way to listen . There is also the view that any oversampling should be done in multiples of the original recording so 88.2 , 176.4 . What sounds worse is to mix the sample rate so 48 at 88.2 or 44.1 at 96 , 192 . Using my dCS Delius and Purcell I can try this and confirm that it is better sounding when in sync . For down sampling I would imagine the same is true but have no way to test this .
|
|
|
|
Post by MartinT on Sept 15, 2024 19:49:57 GMT
What I have noticed is that even 16/44 (CD Red Book) material sounds better than ever from Qobuz streaming.
This is not the same as you are claiming, but then I don't downsample anything. I play the highest possible resolution as a matter of course and use an up to date DAC chiset (ESS9038 Pro). Generally, the higher the resolution the better it sounds. Of course, there are exceptions where the engineer has gone too far with remastering and, even at hi-res, it sounds poor.
What also skews the statistics is that there is far more 16/44 material around than hi-res.
One example of terrific hi-res is Santana's Abraxas, which is best replayed at 24/176 in native format.
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 19:57:11 GMT
What I have noticed is that even 16/44 (CD Red Book) material sounds better than ever from Qobuz streaming. This is not the same as you are claiming, but then I don't downsample anything. I play the highest possible resolution as a matter of course and use an up to date DAC chiset (ESS9038 Pro). Generally, the higher the resolution the better it sounds. Of course, there are exceptions where the engineer has gone too far with remastering and, even at hi-res, it sounds poor. What also skews the statistics is that there is far more 16/44 material around than hi-res. One example of terrific hi-res is Santana's Abraxas, which is best replayed at 24/176 in native format. What I am now thinking is that the newer chipsets (ESS and AKM, for example) has more computing power and therefore can better handle High-Res without the same obvious drawbacks. I will dive into the Abraxas album tomorrow, and try some more! It´s an amazing album so it will be fun. |
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 20:19:30 GMT
When having an extremely low digital noise floor, as you have Martin, then it makes sense that a lower DAC activity (due to less resolution) can now also be noticed better and therefore make the 16/44 sound closer to high-res, due to the noise factor.
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 20:54:07 GMT
The strong view a few years ago was that the best sound is always at the recorded level used . So for things at 44.1 then replay at 44.1 is the best sounding way to listen . There is also the view that any oversampling should be done in multiples of the original recording so 88.2 , 176.4 . What sounds worse is to mix the sample rate so 48 at 88.2 or 44.1 at 96 , 192 . Using my dCS Delius and Purcell I can try this and confirm that it is better sounding when in sync . For down sampling I would imagine the same is true but have no way to test this . I can´t get that logic to make sense, that the recorded quality should matter in a digital context. Digital is always perfect and it is a perfect copy of whatever was recorded. In my mind, it can´t matter what the original was recorded at if you have a perfect copy of the original. I now think, even more, that it is the digital noise factor that is crazy extreme. When you have a really low noise floor then even the processing differences in the DAC chip starts to matter. Depending on what DAC chip you have those differences will be more or less noticeable depending on how your chip is impacted by the resolution increase/decrease, from a noise perspective, caused by the chip itself. |
|
|
|
Post by MartinT on Sept 15, 2024 20:59:58 GMT
I think what bencat is trying to say is that replay of any format which is a multiple of 44.1kHz is best played back at one of those multiples, i.e. 44.1, 88.2 or 176.4. Similarly, any recording made at a multiple of 48kHz is best played back at one of those multiples, i.e. 48, 96 or 192. There is a maths and calculation overhead in crossing between the two, e.g. a recording originated at 96 is not going to sound at its best downsampled to 44.1. |
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 15, 2024 21:08:14 GMT
I think what bencat is trying to say is that replay of any format which is a multiple of 44.1kHz is best played back at one of those multiples, i.e. 44.1, 88.2 or 176.4. Similarly, any recording made at a multiple of 48kHz is best played back at one of those multiples, i.e. 48, 96 or 192. There is a maths and calculation overhead in crossing between the two, e.g. a recording originated at 96 is not going to sound at its best downsampled to 44.1. Right, but does it make any sense? It is a digital file, which is perfect, and the analogue content of something that is perfect can´t sound better/worse (in my mind) depending on how that file was created, in a playback scenario. It will be a difference in resolution, that´s it. I mean, our DAC has no clue how it was recorded. It just plays whatever it is handed, so it can´t matter what the original resolution was? I am thinking out loud here, so feel free to correct me! |
|
|
|
Post by MartinT on Sept 15, 2024 21:34:19 GMT
The issue would be in the downsample conversion, way before it's streamed and reaches the DAC.
|
|
|
|
Post by John on Sept 16, 2024 4:54:15 GMT
Nothing wrong with 16/44. I tend to not like resampling these days but think it can be system dependent. I remember on my Limetree most streams sounded better at 174, but with my Aurender I do not enjoy 174
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 16, 2024 5:29:56 GMT
Nothing wrong with 16/44. I tend to not like resampling these days but think it can be system dependent. I remember on my Limetree most streams sounded better at 174, but with my Aurender I do not enjoy 174 I agree that it seem to be related to the system (DAC hardware, in combination with current noise level). I know that the very best streamers/dac´s, according to many, actually upsample even more. This is beneficial as long as the DAC conversion can handle this without sacrificing the other factors (noise and dynamics) Another piece of the puzzle is that DSD, at least PCM converted, is often said to sound a bit lean, meaning that it sacrifice dynamics, as Sony was claiming. In my case, running the old Burr Brown technology, i think the sweet spot might actually be 16/44 when having a low digital noise floor to protect, even if it can technically handle 194. Again, this is only what i think after having seen what people seem to say about this topic and trying things seem to confirm this, in my setup. |
|
|
|
Post by MartinT on Sept 16, 2024 6:47:00 GMT
I've never liked upsampling and don't use it.
I thought we were discussing downsampling?
|
|
|
|
Post by John on Sept 16, 2024 7:03:29 GMT
I guess I should of mentioned I prefer to listen to native rather than change sample rates, no matter up or down.
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 16, 2024 7:05:18 GMT
I've never liked upsampling and don't use it. I thought we were discussing downsampling? Yes, but since my theory is that it is the noise caused by sample rate, on the chip itself, that actually impacts the sound, when we have removed a lot of the other noise sources. I wanted to test this theory by applying the same logic when upsampling, even more. Here it is sometimes said that dynamics suffer, and maybe also noise in some scenario, which makes sense from what Sony has been saying about 16/44 being a sweet spot mathematically. The problem is that since this is electronics, there will always be a level of noise. But the more we manage to remove this noise the closer we will get to a point where dynamics also starts to suffer, since noise is no-longer such an overwhelming problem. (If we believe in the mathematical Noise-vs-Dynamics-vs-Resolution idea) For me, this certainly seem to makes sense in practice when i down sample and there is a pattern where it seems to make sense in general, when listening to what people are reporting, and what type of DAC and noise floor they have. I mean, just the fact that you prefer Abraxxas in lower resolution goes somewhat in that direction, that your DAC is having to work a tiny bit less and therefore emits less noise, which impacts the DAC itself positively. |
|
|
|
Post by John on Sept 16, 2024 7:07:56 GMT
I prefer native on my system. So no up or down sampling.(Sorry should have mentioned this)
With my portable headphones system for holidays. I up sample in Bluetooth 96/24. However I don't consider this serious hi-fi.
In the end I trust my ears
|
|
Tobias
Rank: Quartet
Posts: 320
Member is Online
|
Post by Tobias on Sept 16, 2024 7:14:08 GMT
Saying that via Bluetooth I sample to 96/24 if using LDAC. It gives the best sound quality. But here we not talking serious hifi. I guess in the end I trust my ears for the best results. Yes, in a wireless scenario we are going to have to convert the signal in our HiFi equipment which causes noise. This noise will ruin the noise level potential in your DAC and the resolution is therefore no-longer a big problem, from a noise perspective. It is always noise, noise, noise, that is the root issue in the digital domain, i believe. Even when we discuss resolution. |
|
|
|
Post by MartinT on Sept 16, 2024 7:38:20 GMT
I mean, just the fact that you prefer Abraxxas in lower resolution goes somewhat in that direction, that your DAC is having to work a tiny bit less and therefore emits less noise, which impacts the DAC itself positively. That's not what I said. I prefer Abraxas in 24/176.4 because that is its native resolution from mastering the analogue master tape. I'm like John, I always prefer playback in native resolution. Sometimes there is no choice, for instance the output of Dire Straits and Genesis are almost all remastered at hi-res but are not available in this format in Qobuz. |
|
bencat
Rank: Quartet
Posts: 353
|
Post by bencat on Sept 16, 2024 7:39:46 GMT
The main reason for the original sampling being the best sounding is that the DAC chip in that case has to do less work and makes less demands on cpu power . The same also applies if there is upsampling at multiples of the same rate . The straight 2 x , 4 x , etc is much easier on computing power and so current demand than the more complicated calculation of 44.1 to 96 .
|
|
