Jump to content
IGNORED

Upsampling


Recommended Posts

If you take a "lossless" recording and upsample it is it, "still lossless?" or is it now "lossy?"

HQ Player (#1) & Audrivana (#2) (wow! love the Apple w/music!!) .. these two software make my system "Amazing!", Purist USB- Benchmark DAC2 HGC (love it!), Purist Audio XLR , ATC SCM25A's (To Die For!) & Focal sub6 . Triode Power Cables with Uber Buss (Yes!) Also enjoy Audeze LCD3 w/"fat pipe cardas."

Link to comment

I tend to reserve the term "lossy" for describing psychoacoustic compression, so I would pose your question more as "destructive" vs. "nondestructive" conversion.

 

If the upsampling just involves padding 16-bit with zeroes to make it 24-bit, I would call that nondestructive. Anything that involves dithering, filtering, or sample rate conversion I would consider destructive as the process is not reversible, even though there would likely be little or now audible degradation.

Link to comment

Also, if you're doing upsampling on the fly, the file itself is not modified in any way, so the lossless/lossy distinction doesn't really come into play.

 

This begs the question of whether the playback is bit perfect, but I think we can leave that for another day (or just leave it).

 

--David

Listening Room: Mac mini (Roon Core) > iMac (HQP) > exaSound PlayPoint (as NAA) > exaSound e32 > W4S STP-SE > Benchmark AHB2 > Wilson Sophia Series 2 (Details)

Office: Mac Pro >  AudioQuest DragonFly Red > JBL LSR305

Mobile: iPhone 6S > AudioQuest DragonFly Black > JH Audio JH5

Link to comment

The term 'bit perfect' is related only to computer's digital output (USB, S/PDIF) in relation to source music file.

 

Resampling occurs in almost every today's DAC. Upsampling in a PC player can be also considered as digital processing before the D/A conversion itself. There is no need to see it differently as the digital processing in DAC. Therefore I would not call it destructive.

Link to comment
I tend to reserve the term "lossy" for describing psychoacoustic compression, so I would pose your question more as "destructive" vs. "nondestructive" conversion.

 

If the upsampling just involves padding 16-bit with zeroes to make it 24-bit, I would call that nondestructive. Anything that involves dithering, filtering, or sample rate conversion I would consider destructive as the process is not reversible, even though there would likely be little or now audible degradation.

 

 

+1.

 

"Lossy" is a data conversion term that doesn't mean what most audio-types seem to think it means. In answer to the OP's Q; yes up-sampling a "lossless" file, from say, 16-bit to 24-bit and from 44.1 KHz to say 96 KHz sampling rate is "lossy" by the data definition, because the two files would no longer have the same check-sum. And even though I've never tried it, I also strongly suspect that if you re-converted that up-sampled file back to it's original 16/44.1 state, that the newly reconverted file still wouldn't have the same check-sum as the original file had.

George

Link to comment
If you take a "lossless" recording and upsample it is it, "still lossless?" or is it now "lossy?"

 

It is lossless in that it will still have all the audio information in it. That's the really important part.

 

It is not lossless in terms of data compression, which requires that the compressed file be able to be restored to the original with exactly the same information/checksums, etc. If we were doing data compression, which we are not, and which distinction seems to get lost in this somewhat hotly debated issue.

 

-Paul

Anyone who considers protocol unimportant has never dealt with a cat DAC.

Robert A. Heinlein

Link to comment
It is lossless in that it will still have all the audio information in it. That's the really important part.

 

It is not lossless in terms of data compression, which requires that the compressed file be able to be restored to the original with exactly the same information/checksums, etc. If we were doing data compression, which we are not, and which distinction seems to get lost in this somewhat hotly debated issue.

 

-Paul

 

 

Agreed. But to the average audiophile the term "lossy" is a pejorative because they see the word as meaning a loss of (musical) information (as in lossy compression), when all it really means data-wise is that the original data file cannot be perfectly reconstructed. It really says nothing about whether any intelligence or quality is lost.

George

Link to comment

What if you upsample to 128DSD...you can't reconvert it so it is destructive but you don't lose any information (debatable)?

Positive emotions enhance our musical experiences.

 

Synology DS213+ NAS -> Auralic Vega w/Linear Power Supply -> Auralic Vega DAC (Symposium Jr rollerball isolation) -> XLR -> Auralic Taurus Pre -> XLR -> Pass Labs XA-30.5 power amplifier (on 4" maple and 4 Stillpoints) -> Hawthorne Audio Reference K2 Speakers in MTM configuration (Symposium Jr HD rollerball isolation) and Hawthorne Audio Bass Augmentation Baffles (Symposium Jr rollerball isolation) -> Bi-amped w/ two Rythmic OB plate amps) -> Extensive Room Treatments (x2 SRL Acoustics Prime 37 diffusion plus key absorption and extensive bass trapping) and Pi Audio Uberbuss' for the front end and amplification

Link to comment

PCM and DSD are VERY different formats.

 

If you convert good PCM to DSD128 you could get an improved SQ, but this very DAC dependent (when playing).

 

Once on DSD I would stay there, because I don't believe the DSD back converted to PCM could be of any value. Anyway, you already have the original PCM track.

 

DSD to PCM conversion has a value for users don't have a DSD capable DAC and the file is on DSD only.

 

Just my opinion,

 

Roch

Link to comment

It is lossy because it just is in 99.99% of cases.

It is allowed to be named lossy because there's 0.01% of cases where it is not. And I mean in the very realm of the subject.

 

If I play 24/192 through a NOS DAC, nothing in the digital part of the chain is changed nor does it need to (read : no digital filtering needed). Now it remains lossless (digital input into the D/A conversion can be reversed into the original).

 

It is also lossless for those upsampling algorithms that do not alter the original audio samples and only inject new samples and lastly are again played through a NOS D/A converter. This too can be reversed into the 100% original (Redbook becomes Redbook again with no bits changed). Notice that this is *not* referring to the "inject zeroes" method as mentioned by others; This is lossy.

 

So remember, *because* in this context the data can be lossless, the term lossy applies to it just the same. So no need to refer to "bit perfect" which is used in other contexts as pointed out by others. Still my first example of the 24/192 that needs no adjustment and is played through a NOS converter will be just that : bit perfect playback (but I too think that this is wrong usage of an existing term meant for something else).

If this is again compared to upsampling into a lossless stream like in my second example, this is not bit perfect because the data has changed (no matter it is lossless).

 

Btw for a digital volume control all applies the same, except for the 24/192 into NOS because that would not be bit perfect any more, while it still can be lossless (and regard this in 0.0001% of cases).

 

How to make a small subject large ...

Peter

Lush^3-e      Lush^2      Blaxius^2      Ethernet^2     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment
  • 2 weeks later...

I would say there are two aspects:

1) whether something is "lossless" in digital domain

2) amount of precision loss when converted to analog

 

While upsampling (AKA oversampling) may not be always "lossless" for (1), it aims to reduce loss on (2).

 

While NOS DAC may be "lossless" in digital domain, it has horrible losses on (2) without upsampling/oversampling.

 

My personal aim has been to keep errors of (1) below -200 dBFS, so it is safely well below thermal noise of any realizable audio system and better than 32-bit PCM resolution (-193 dBFS). Unless you manage to use superconductors throughout and keep your entire system very close to absolute zero (0K) temperature.

Signalyst - Developer of HQPlayer

Pulse & Fidelity - Software Defined Amplifiers

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×
×
  • Create New...