[Ghoostknight]

2. I presume you mean the infinitewave (http://src.infinitewave.ca/) site. That’s been going for many years. Yes, you can “clearly SEE” differences between resampling algorithms because they have deliberately emphasised the visual differences! The FAQ states:
“Are most SRCs really that bad?
No. If you look at the decibel scale to the right from the graphs, you can see that the range of these graphs is very wide: down to -180 dB. The distortions generated by most properly designed SRCs are below -100 dB and can hardly create audible artifacts. The bottom line is that most tested SRCs range from fairly good to excellent, but the graphs are very sensitive to emphasize the differences.” - Emphasis mine. Incidentally, the Windows 10 DS resampler (the only listing for Windows) is very good, artefacts are roughly -120dB to -140dB and way below audibility or even reproducibility!

yes thats the site
instead of arguing the audibility please everyone should try this yourself, imo its kinda clearly audible, and i dont think this is placebo.. for most of the time i actually checked this site after trying a specific resampling,
tho im unsure if "stuff under -100db" is the only thing the resampling "tackles", its not like i hear a noisefloor at -100db but it loses fidelity in the "normal audible range" like even on 0db to -30db, this is whats so crap about resampling

i never investigated the reason tho... does resampling alter impulse response? (since there are impulse response graphs on the site)

and this windows 10 DS actually shows noise up to -70db... in the case of pipewire its -70 vs -170 (!!!)
https://src.infinitewave.ca/?Top=Pipewire_Q14&Bot=Windows10&Spec=0111

i also think this graph : https://src.infinitewave.ca/?Top=Pipewire_Q14&Bot=Windows10&Spec=0111 shows how stuff gets reflected back into the audible range (realistically it shows stuff around -90 to -120db

but overall... why bother with bad implemented resampling if there are much better solutions... well beside windows sucks which is one of the biggest bummers imo

There are not a lot of different reconstruction filters as far as I’m aware and reconstruction filters have been trivial to implement for 30+ years. What has changed is the number of available anti-imaging filters. This is a relatively new development and is entirely driven by marketing! Before this there were no switchable anti-image filters, DAC chips all had a single linear or minimum phase filter, with typically a 2-3kHz transition band and NO audible artefacts. Then a few years ago audiophile manufacturers applied their typical scam of taking a non-issue and falsely claiming it was an audible issue, this time to anti-image filters. In response to manufacturers demands to provide a solution to this non-issue, the DAC chip manufacturers implemented switchable filters. The vast majority of which are still inaudible but there are some with transition bands starting far lower, which can be audible.

i actually found a really "funny" fact
if you look into the manual of the ess9038q2m chip... you find this sentence: Custom sound signature is supported via a fully programmable FIR filter with 7 presets.
are you as shocked as i was first reading this? the manfacture of the chip thats used in 50 to 70% of all current dacs says reconsutrction filter "can" matter, how is this possible?
i mean they could have said " our chip is so awesome we support 7 ("inaudible") reconsutruction filters " but they went with the sentence above... well "marketing" i guess..

my dac actually can switch between those 7 filters and there are indeed audible differences, so much (tho its subtle if you dont know what to listen for) that i actually believe a huge part of differences heared between dacs actually come from this, also here... impulse response has a effect on the whole audible range!!

this makes it also nearly impossible to really compare one dac to another if they dont use the same filter, there could be differences in the analog section that are just overlayed by the a audible differences between these filter

Objectively and if high fidelity is what you’re after, the answer is: Do nothing! Typically the default filter is the same sort of filter used in DACs a decade or more ago with completely inaudible artefacts. If you’re not sure, don’t choose the option usually called something like “slow”.

ah i somewhat agree ... but minimum phase slow rolloff sounds "the best" imo and with resampling to 192khz you avoid the fact that the roll off isnt that great, imo "currently" i would say this is the best solution... IF you use high quality resampling, tho i might change my opinion on this at some point with a different dac

 

[Ghoostknight]

The simple answer is: No, absolutely none whatsoever. A more complex answer is: There could be rare consumer scenarios where greater bit depths might be advantageous. However, this doesn’t apply to the bit depth of the audio file itself (16bit) but to the environment in which it’s processed, which is invisible and unalterable by the consumer. An example of such processing where a higher bit depth *could* be beneficial would be a digital volume control.

one important fact, just objectively speaking is that 24bit processing is very well appreciated if you use software volume
1. we know that our dacs support around 120db of dynamic range
2. a 16bit file supports 96db dynamic range and actually "injects" noise -atleast- at 96db instead of 120db, so you basicly limit your dac artificially with 16bit processing, sure the whole audible debate stands but why bother if its so easy to avoid?

if you use 16 bit processing with software volume you actually lose 24db of dynamic range... which might end up being 72db (or even less!!) of dynamic range with 16 bit processing and software volume turned down

so a important fact with pc`s imo is please use 24bit processing just for this software volume reason, just objectively speaking

im not sure how cd players preserve the dynamic range of atleast 96db but i guess some do, with pc and software volume its a different story tho

 

[Ghoostknight]

someone said that 24 bit high res recordings actually have a different mastering which might be well true for some of the recordings and would explain some easier audible differences beside the fact that high sample rates could lessen the effects of reconstruction filters for example

but if its a different master (which might be as well "tuned" for "audiophiles" (i guess)) isnt it worth checking out?
i think this is one of the "easier" reasons why someone actually wants to go with high res, even if the format doesnt do crap purely audible speaking
so the whole 16bit vs 24bit debate just because of this fact is.... questionable

and well i somewhat agree.. if it comes down to the mastering its purely marketing going with high res but what alternative there is? either we go for it because we like the master or we "dont wanna support such behaivor"

imo it comes easly down to trying 24bit stuff yourself and decide if you actually prefer it... sure we can massage our brains all day long saying it doesnt matter but... i dont know

 

[gregorio]

one important fact, just objectively speaking is that 24bit processing is very well appreciated if you use software volume

Can you name any devices that use 24bit processing? If there aren’t any, then surely it’s not a “one important fact” it’s a totally irrelevant fact?

if you use 16 bit processing with software volume you actually lose 24db of dynamic range...

Again, can you name any software that uses 16bit processing? And even if there were some, why would you loose 24dB of dynamic range?

so a important fact with pc`s imo is please use 24bit processing just for this software volume reason, just objectively speaking

What PCs use 24bit processing? Old PCs used 32bit processing, modern ones use 64bit. So how could you use 24bit processing and why would you even if it were possible? And again, so how is what you state “an important fact”?

G

 

[Ghoostknight]

Can you name any devices that use 24bit processing? If there aren’t any, then surely it’s not a “one important fact” it’s a totally irrelevant fact?

are you talking about most dacs processing in 32 bit? ... well this doesnt really matter if you set for example windows to use 16bit "processing", then there will be a new noisefloor at 16bit, then turning down software volume goes beyond that... 15bit.. 13 bit .. etc
if you use 24bit processing inside windows for example you have more "headroom" (feetroom?!) to actually lower the volume and preserving the 16 bit dynamic range

Again, can you name any software that uses 16bit processing? And even if there were some, why would you loose 24dB of dynamic range?

with windows you are able to set 16bit / 24bit / 32bit processing, also on linux this is easly possible but most stuff defaults to 32bit, but some people probably run windows in 16bit processing mode

even if for example your player software supports 32bit and you have set windows to 16bit ... windows processing will just cut off the bottom 16 bits ... and if you lowered volume in the player software you are now below 96db dynamic range

the 24db loss was just an example...
lets put it like this
with no volume reduction you get:
1. 16 bit processing 96db dynamic range
2. 24 bit processing 140db (i think it was 140) dynamic range

now we use software volume to reduce the volume by -24db
what happens now is this :
1. 16 bit processing = 72db dynamic range
2. 24 bit processing = 116db dynamic range

as you see 24 bit processing is still well above 96db of the 16bit file we are playing, so we actually preserved the 16 bit file 100% even if we use software volume to reduce the volume

What PCs use 24bit processing? Old PCs used 32bit processing, modern ones use 64bit. So how could you use 24bit processing and why would you even if it were possible? And again, so how is what you state “an important fact”?

im not talking about the pc architecture you seem to talk about
there is a setting in windows where you can set what bitdepth windows is communicating with the dac, even if player software and dac itself has very well 32 bit processing

im pretty sure some people still use this 16 bit setting in windows... maybe even with the mindset of "16 bit is all we need" but if you use software volume its a different story

 

[gregorio]

beside the fact that high sample rates could lessen the effects of reconstruction filters for example

How is that a fact? For quite a few years most DACs oversample 256x (although some use 512x). So with a sample rate of 11.2mHz, how does an input sample rate of 44.1kHz, 192kHz or even 384kHz have any effect at all, let alone “lessen the effects of the reconstruction filter”?

but if its a different master (which might be as well "tuned" for "audiophiles" (i guess)) isnt it worth checking out?
i think this is one of the "easier" reasons why someone actually wants to go with high res, even if the format doesnt do crap purely audible speaking

IF it’s a different master (or different version of the same master) and IF it’s a version without the additional compression then it might indeed be worth checking out. But how do you know it is a different master/version and that the Hi-Rez version is not the one without additional compression?

imo it comes easly down to trying 24bit stuff yourself and decide if you actually prefer it...

Sure but you have to do that every time. Just because the 24bit version of one recording doesn’t have the additional compression that might have been applied to the 16bit version, doesn’t indicate this is the case with another recording.

G

 

[Ghoostknight]

How is that a fact? For quite a few years most DACs oversample 256x (although some use 512x). So with a sample rate of 11.2mHz how does, how does an input sample rate of 44.1kHz, 192kHz or even 384kHz have any effect at all, let alone “lessen the effects of the reconstruction filter”?

honestly i dont get the dac oversampling thing and it which cases it improves things (yet, you can enlighten me tho ) ... but for a 44,1khz samplerate file the reconstruction filter starts around 20khz which indicates that dac oversampling doesnt play a role here, doesnt it?

 

[Ghoostknight]

If it’s a different master (or different version of the same master) and if it’s a version without the additional compression then it might indeed be worth checking out. How do you know it is a different master/version and that the Hi-Rez version is not the one without additional compression?

well then it comes down to preference, i dont think additional compression is nessecarly a bad thing (most of the time it is from a audiophile perspective but i dont think you can generalize this)
additional compression might as well enhance the expierence in some ways (tho i agree, compressed vocals can be kinda harsh 95% of the time)

tho i generally agree... compression is mostly bad and "kinda" used to make the music sound "better" on low end stuff... but it most of the time sounds worse if you got a somewhat proper setup
but there are also i think a lot of artist that use compression as a artistic tool and not push it to the max.. so i wouldnt say compression is always bad

Sure but you have to do that every time. Just because the 24bit version of one recording doesn’t have the additional compression that might have been applied to the 16bit version, doesn’t indicate this is the case with another recording.

imo its not black and white but grey, if you like different masters high res might be worth checking out

if you dont wanna "support" high res you have to accept that you cant try all masters, its that easy

tho generally speaking it would be kinda interesting to know how much masters of high res files actually differentiate from cd quality equavalants

 

[gregorio]

are you talking about most dacs processing in 32 bit? ... well this doesnt really matter if you set for example windows to use 16bit "processing"

You cannot set windows to use 16bit processing!

but some people probably run windows in 16bit processing mode

What people? It’s impossible to set windows to run in 16bit processing mode, unless you run a version of Windows prior to Windows 95!

now we use software volume to reduce the volume by -24db
what happens now is this :
1. 16 bit processing = 72db dynamic range
2. 24 bit processing = 116db dynamic range

1. What 16bit processing?
2. There are no commercial music recordings with 116dB of dynamic range! Around 60dB is the max dynamic range and the vast majority have less than 60dB.

im not talking about the pc architecture you seem to talk about
there is a setting in windows where you can set what bitdepth windows is communicating with the dac

The bit depth that “Windows is communicating with the DAC” is NOT the processing bit depth. The only exception would be if you’re using 32bit processing in windows and communicating with the DAC using a 32bit file format! Still nothing to do with 16 or 24 bit though. Your “important facts” are not even valid facts, let alone important!

honestly i dont get the dac oversampling thing and it which cases it improves things (yet, you can enlighten me tho ) ...

If you “don’t get the DAC oversampling thing”, then don’t you think it would be wise, especially in a science discussion forum (!), not to make assertions about it? If you don’t know, it would be far wiser and far more polite to this subforum to ask questions rather than just making up false assertions (or repeating false marketing assertions)!

but for a 44,1khz samplerate file the reconstruction filter starts around 20khz which indicates that dac oversampling doesnt play a role here, doesnt it?

Good, a question! Although also a false assertion! The reconstruction filter is an analogue filter that operates to remove all frequencies above the Nyquist point (half the sampling frequency). So in the case of a typical 256x oversampling DAC, the reconstruction filter has to remove freqs above 5.64mHz, which is very easy to accomplish without phase or other issues as it can have a very gentle roll-off. In fact it’s so easy that as far as I’m aware, most DAC reconstruction filters remove everything above a few hundred kHz.

However, before the reconstruction filter, most DAC topologies require an anti-image filter, set at the Nyquist point of the input file format (so 22.05kHz in the case of a 44.1kHz sampling rate). The advantage here is that the anti-image filter is in the digital domain, where it’s far easier and cheaper to implement a brickwall filter with a relatively small bandwidth (say 20kHz to 22.05kHz). This would be difficult/expensive to accomplish with a reconstruction (analogue) filter without fairly serious, potentially audible, artefacts. Hence the main benefit of oversampling, both a higher fidelity output and at considerably lower difficulty and cost. It’s not often in engineering that you get both at the same time (better performance and less cost) which is why the delta/sigma oversampling topology quickly dominated the market.

G

 

[Ghoostknight]

However, before the reconstruction filter, most DAC topologies require an anti-image filter, set at the Nyquist point of the input file format (so 22.05kHz in the case of a 44.1kHz sampling rate). The advantage here is that the anti-image filter is in the digital domain, where it’s far easier and cheaper to implement a brickwall filter with a relatively small bandwidth (say 20kHz to 22.05kHz). This would be difficult/expensive to accomplish with a reconstruction (analogue) filter without fairly serious, potentially audible, artefacts. Hence the main benefit of oversampling, both a higher fidelity output and at considerably lower difficulty and cost. It’s not often in engineering that you get both at the same time (better performance and less cost) which is why the delta/sigma oversampling topology quickly dominated the market.

oh, sorry then i was talking about the anti-image filter, thanks for clearing it up
i will do some reading on the reconstruction filter stuff , but since we cant influence it in any way we have to deal with what we get i guess
tho im not much worried about a analogue filter above some hundred khz (i guess its somewhere in the 800khz range since many modern dacs support 768khz)

the anti-image filter debate stands tho, you can influence the output with a different samplerate and the anti-image filter itself can have a influence on the output

The bit depth that “Windows is communicating with the DAC” is NOT the processing bit depth. The only exception would be if you’re using 32bit processing in windows and communicating with the DAC using a 32bit file format! Still nothing to do with 16 or 24 bit though. Your “important facts” are not even valid facts, let alone important!

well maybe i have used the wrong word but the fact stands, if you let windows communicate with 24bit with the dac then you can preserve the full bitdepth of 16bit files longer with software volume control, no?
its not "really" important for the conversation about 16bit and 24bit FILE formats but i thought its worth a mention that 16 bit is not "all" we need

 

[bigshot]

Boy, there is absolutely no focus here. Stuff is just being thrown out randomly. I can't tell what your point is because it's all disorganized. Did the line by line replies chop up the context, or did it start out that way?

 

[gregorio]

well then it comes down to preference, i dont think additional compression is nessecarly a bad thing (most of the time it is from a audiophile perspective but i dont think you can generalize this)

True! It’s partially preference and also listening environment. If you’re in a car, walking, travelling or exercising for example, you’re going to have a high noise floor environment and a quite highly compressed recording is going to be preferable, otherwise you’re just not going to hear the quiet parts of the track above the noise floor without the loudest parts being uncomfortably/painfully loud.

tho i agree, compressed vocals can be kinda harsh 95% of the time

No, quite the opposite. Vocals are virtually always compressed and it’s pretty much a requirement with rock/pop genres. The high quality master will therefore have compressed vocals. The potential problem is adding additional compression to vocals/the master which already has an optimal amount of compression.

high res might be worth checking out

It might be or more likely you’ll be wasting your time because it’s either audibly identical or very occasionally, worse. We have audiophiles/audiophile marketing to thank for this nuisance!

the anti-image filter debate stands tho, you can influence the output with a different samplerate and the anti-image filter itself can have a influence on the output

The debate doesn’t stand though! Yes, the sample rate and anti-image filter do influence the output but not audibly, the “influence” is all above the range of hearing. There’s one exception, it is possible to deliberately design a sub-optimal anti-image filter, for example, a filter that starts (unnecessarily) rolling-off well inside the audible range. That’s still not a “debate” though, because I don’t think anyone here would debate that it’s not possible to have an audible filter if you set out to deliberately screw it up!

well maybe i have used the wrong word but the fact stands, if you let windows communicate with 24bit with the dac then you can preserve the full bitdepth of 16bit files longer with software volume control, no?

Still using the wrong words I think. If you lower the volume of a 16bit file, then all the data is preserved. If you then output the result of that processing to a 16bit file then you will not preserve the “full bit depth”. However, unless you lower the level quite drastically, then the bits you’re loosing are just noise (the noise floor of the recording). And, you’re going to loose those bits anyway, they’re going to be inaudible as a consequence of lowering the volume! The only exception would be user error, incorrect gain-staging, lowering the digital level significantly and then whacking up your (analogue) amp to compensate.

G

 

[gregorio]

Did the line by line replies chop up the context, or did it start out that way?

It happened that way because of cross posting. So some of my replies include quotes from more than one post.

G

 

[bigshot]

Unfortunately, that doesn’t ENSURE the highest possible quality, occasionally the Hi-Res version is poorer quality, although this is quite rare AFAIK.

If we're completely disregarding fidelity because hires and 16/44.1 (or high data rate lossy) all sound the same, then when you say "poorer quality", you're talking about the quality of the engineering of the album itself. That is a subjective judgement. The truth is, that hires music is more likely to sound *different*, but that doesn't necessarily mean *better*.

Hires can also mean more than just remastered. An extreme example are the albums of The Rolling Stones. Several of them were completely remixed. They replaced wire reverbs with digital ones. They took the "phone futz" off of Jaggers' vocals and lowered the tone of his voice. They went through and polished and cleaned up the sound. It definitely sounds cleaner, but it doesn't have the raw edge and energy of the originals. Similar things have happened to Led Zeppelin's catalog. Certain of their albums sound best on LP because the remastering and remixing made the music different, but not better. I'm told that Sinatra is all over the map too.

There is absolutely no relationship between format and sound quality. Fancy expensive hires can sound flat and soft, and an LP with obviously inferior fidelity can sound better. The only way to tell is to talk to a knowledgeable person who has heard and carefully compared the various releases.

Sound quality is a function of engineering more than format.

 

[gregorio]

im not much worried about a analogue filter above some hundred khz (i guess its somewhere in the 800khz range since many modern dacs support 768khz)

Missed this. Most modern DACs don’t support 768kHz sampling rate, maybe it’s becoming more common in some audiophile DACs though. However, the reconstruction filter only needs to pass the frequencies below the Nyquist point, which is half the sampling rate. In the case of 768kHz, the reconstruction filter would need to pass frequencies up to 384kHz. Of course, with the oversampling rate at say 11.2mHz, there’s no reason why the reconstruction filter couldn’t start rolling off at 1mHz or more. All pretty pointless though as there’s nothing in a recording at 384kHz except noise/distortion, plus no studio mics record that high, no HPs or speakers I know of can reproduce anything that high and of course, adult hearing is pretty much gone by 16kHz anyway!

G