Apodizing filter

[71 dB]

My maths knowledge is not that great, so maybe someone like @71 dB can explain it better or correct my understanding...

G

I don't think I could explain it much better than you. Apodizing filters are windowed alterations from default filters to make the filtered signal "cleaner". It has relevant use in optics (sharper better focused images), but in digital audio it is more of a "marketing name" for filters with reduced ringing.

--------------

The pre- and post ringing of digital filters is same kind of myth as the infamous staircases. These things do exist when digital signals are analysed from certain engineering perspectives, but in practise these things don't really matter (or matter so little it is crazy to make a fuzz about them). Ringing of digital filters is quite misunderstood. I have seen people with PHD in acoustics and signal processing struggle to understand the topic fully. Nothing is "ringing" when you listen to your music digitally (well, maybe your ears will be ringing afterwards if the volume is too high ). This is because theoretically each sample points has its ringing, but when you sum all the sample points in time (sinc functions shifted in time for each sample point), all the ringing sums to zero. It is when we analyse the signal/filter when the ringing appears for mathematical reasons. Digital audio is like quantum physics (both are quantized systems*): There are limits to how much you can measure and analyse it. Quantum physics is notoriously unintuitive and difficult to understand. Digital audio is much much simpler, but some of the unintuitive nature is still there. Marketing takes advantage of that whenever possible. That's why we have audiophiles worrying about staircases and ringing when they should be worrying about things like room acoustics, speaker placement in room and how music is being mixed. The stuff that really matters.

* I even believe the reality is "dithered" on quantum level explaining the randomness and statistical nature of the physics on atomic scale, but then again my knowledge and understanding of quantum physics is quite limited (one course in university barely passed).

 

[donjoe0]

Nnnno, I don't see any reason to assume that. Dithering is intentional statistical manipulation to remove some of the effects of "natural randomness" that we don't like to see, like you wouldn't show students a long string of coin tosses all coming up tails when first introducing the concept of randomness, even though a long string of tails absolutely can be one manifestation of natural or "true" randomness. You'd have to assume something very strange like a god in order to say QM is dithered, i.e. some intelligent intervention into the fundamentals of reality. It's more likely that it isn't dithered and that's why it produces so many unintuitive results and throws us off with "cloud bunnies" etc.

 

[Ghoostknight]

Can i throw in this one?

atleast in the case of this picture (of Aune X8)

it seems like the apodizing one has less pre and post ringing than the linear fast roll off but overall it seems like an alteration of the linear phase fast roll off


what i found (and many seem to think that way) go with a slow roll off filter (i mainly used the linear phase slow roll off and minimum phase slow rolloff)
now the thing is... you want the gentleness of the slow roll off filter (because of the sound quality) but without its cons (potential nyqist/antialising artifacts)
what you can do is this: upsample 44,1khz to 192khz on your player, and let your dac run at 192khz with material that will never reach the nyqist frequency....

3 benefits:
1. if you use good upsampling and your dac "likes" the samplerate upsampling will improve sound quality
2. the gentle slow roll filter will improve sound quality, without potential artifacts
3. when upsampling to the highest samplerate your dac supports you sometimes circumvent resampling on the chip itself depending on the dac chip, thats why upsampling can sound so good imo

atleast this is what i found works best for me so far

 

[GoldenSound]

what you can do is this: upsample 44,1khz to 192khz on your player, and let your dac run at 192khz with material that will never reach the nyqist frequency....

3 benefits:
1. if you use good upsampling and your dac "likes" the samplerate upsampling will improve sound quality
2. the gentle slow roll filter will improve sound quality, without potential artifacts
3. when upsampling to the highest samplerate your dac supports you sometimes circumvent resampling on the chip itself depending on the dac chip, thats why upsampling can sound so good imo

atleast this is what i found works best for me so far

using external oversampling doesn't overcome the inherent tradeoffs. Mathematically it's exactly the same thing, and so the same questions about filter choice/design are there whether you're doing it internally or externally.
The only benefit to doing it externally is if you're using higher performance tools like HQP or PGGB you can use extremely long filters with extremely good noise shapers.

But in terms of external vs internal oversampling inherently, there isn't a difference

 

[bigshot]

Can i throw in this one?

A chart is just wiggly lines without a x and y axis scale to put it in context.

 

[donjoe0]

you want the gentleness of the slow roll off filter (because of the sound quality)

I turn on oversampling in Neutron Player when I'm using some older DAC that has an unknown reconstruction filter setting, in case something dumb is selected like a slow rolloff option that cuts early and affects the audible band. So I do it to work against the selection of a slow rolloff filter. When I have control over the filter, I just select linear-fast and listen at whatever the material's rate is, 44 or 48.

I have no idea why you would think slow rolloff improves the sound in any way.

 

[bigshot]

A slow rolloff is easy with an equalizer if that’s what you want. And an EQ has more uses than a bunch of esoteric filters, most of which make no audible difference.

 

[gregorio]

Nnnno, I don't see any reason to assume that.

Nothing is being assumed, except in your case of incorrectly assuming what you think the “foot” of “apodizing” means. We’ve merely explained the basics of what “apodizing” means in the context of digital audio filters.

Dithering is intentional statistical manipulation to remove some of the effects of "natural randomness" that we don't like to see …

Dithering is entirely another subject, not related to digital filters, so you’ve gone off on a bit of a tangent here and unfortunately got that wrong too! Completely contrary to your assertion, dithering is not a “manipulation to remove some effects of natural randomness” it’s the opposite, a process that actually adds randomisation to the data/signal.

It's more likely that it isn't dithered and that's why it produces so many unintuitive results and throws us off with "cloud bunnies" etc.

I’m not sure exactly what you’re trying to say here but the likelihood that some digital audio signal/data is not dithered, is precisely zero! It is an inherent part of the digitisation process and every professional ADC ever made dithers the signal. Therefore your “and that’s why” you get unintuitive results and whatever “cloud bunnies” are is entirely false/non-existent.

what i found (and many seem to think that way) go with a slow roll off filter (i mainly used the linear phase slow roll off and minimum phase slow rolloff)

What “many seem to think”, a relatively few delusional audiophiles? The professionals; audio engineers, digital audio filter designers, the various manufacturers of professional ADCs and DACs all unanimously think that a standard fast, linear phase filter is the “way to go”. In fact there’s no choice, all professional ADCs only implement that standard filter type, there are no filter choices. It’s only in the audiophile world where that unnecessary (marketing) choice exists.

now the thing is... you want the gentleness of the slow roll off filter (because of the sound quality) but without its cons (potential nyqist/antialising artifacts)

What do you mean you want a slow roll off filter “because of the sound quality”, why would you want lower sound quality and why would you recommend that to others? A slow roll off filter “without it’s cons” (the loss of audible freqs or potential anti aliasing/imaging artefacts) is called a fast roll off, linear phase filter, which is why professional ADCs only implement fast roll off, linear phase filters!

what you can do is this: upsample 44,1khz to 192khz on your player, and let your dac run at 192khz with material that will never reach the nyqist frequency....
3 benefits:
1. if you use good upsampling and your dac "likes" the samplerate upsampling will improve sound quality
2. the gentle slow roll filter will improve sound quality, without potential artifacts
3. when upsampling to the highest samplerate your dac supports you sometimes circumvent resampling on the chip itself depending on the dac chip, thats why upsampling can sound so good imo

1. No it won’t, unless something is seriously broken it will have no effect on SQ at all.
2. No, it will do the exact opposite! How does reducing potentially audible treble freqs in the recording somehow improve SQ, how does this deliberate reduction in fidelity “improve sound quality”?
3. Why do you want to upsample, in order to avoid upsampling? How does that make any sense? What imagined deficiency do you think a DAC chip has, that’s specifically designed to upsample and that DAC chips have been achieving perfectly (beyond the limits of audibility) for over 30 years?

As all three of your “3 benefits” are actually not beneficial then your assertion to upsample on your player is false!

G

 

[Ghoostknight]

1. No it won’t, unless something is seriously broken it will have no effect on SQ at all.
2. No, it will do the exact opposite! How does reducing potentially audible treble freqs in the recording somehow improve SQ, how does this deliberate reduction in fidelity “improve sound quality”?
3. Why do you want to upsample, in order to avoid upsampling? How does that make any sense? What imagined deficiency do you think a DAC chip has, that’s specifically designed to upsample and that DAC chips have been achieving perfectly (beyond the limits of audibility) for over 30 years?

2. doesnt matter, because the slow roll off is happening at 192khz in my case
3. you should know pretty well how different upsampling algorithms can perform but yeye "its beyond audible"

 

[donjoe0]

I’m not sure exactly what you’re trying to say here

Obviously you completely missed that I was going way off-topic and replying strictly to the section about quantum mechanics. May be worth checking your compulsion to reply to things you're not sure you've even understood in the first place.

 

[71 dB]

2. doesnt matter, because the slow roll off is happening at 192khz in my case

If a digital filter has a roll off starting at say 16 kHz at 44.1 kHz sampling frequency, using the same filter at 192 kHz sampling frequency makes the roll off start at about 70 kHz. Of course, you should have nothing there because the 44.1 kHz version didn't have anything above 22.05 kHz.

At my age it doesn't matter what happens above 16 kHz anyway, but children probably can detect a difference...

3. you should know pretty well how different upsampling algorithms can perform but yeye "its beyond audible"

Is it so difficult to believe some things in audio really are inaudible? Is it really so difficult to admit you can't hear some things? Why ignore placebo effect/confirmation bias when those things not only exist, but have a significant influence on how we experience sound?

 

[gregorio]

2. doesnt matter, because the slow roll off is happening at 192khz in my case
3. you should know pretty well how different upsampling algorithms can perform but yeye "its beyond audible"

2. It matters if we’re talking about 44.1kHz but if you’re only playing 192kHz recordings then you’re right, it “doesn’t matter”. So please explain how “it doesn’t matter” is somehow one of your quoted benefits!
3. Yes I do know how they “can perform”, the oversampling algorithms in DACs have performed beyond audibly perfect for more than 30 years, as already stated!

Obviously you completely missed that I was going way off-topic and replying strictly to the section about quantum mechanics. May be worth checking your compulsion to reply to things you're not sure you've even understood in the first place.

Maybe worth you checking and not going so off topic and if you do, then at least make it clear that’s what you’re doing.

G

 

[71 dB]

Obviously you completely missed that I was going way off-topic and replying strictly to the section about quantum mechanics.

Off-topic indeed. Since this is not a forum for QM, I won't debate that topic. I simply mentioned some similarities between QM and digital signals I recognise as a system thinker. My knowledge/understanding of QM is quite limited and in no way do I suggest my thoughts represent scientific facts. I do think about the mysteries of the Universe for intellectual and recreational purposes and that leads to my own ideas, but I know they are just my silly own ideas.

 

[bigshot]

I don’t understand why people expend so much time and energy on things that don’t matter. I’m sure there are audible things in their system that would benefit more from their attention, but they don’t even consider those. I can only conclude that for them, the hobby is more about OCD and obsession with meaningless details, not sound you can actually hear.

 

[Ghoostknight]

I don’t understand why people expend so much time and energy on things that don’t matter. I’m sure there are audible things in their system that would benefit more from their attention, but they don’t even consider those. I can only conclude that for them, the hobby is more about OCD and obsession with meaningless details, not sound you can actually hear.

either you heared the difference yourself or you didnt, not much more to say

there is no point starting arguing with people that also say all op amps sounds the same and many other examples... i guess it goes by "you live what you preach", honestly this counts for both camps but apparently most subjective experiences are honest ones where objective arguing goes by "no, look at this study, cant be!", honestly im sick of it, not so much about this specifically but more of you guys trying to force your opinion on others "because its the only true one"

you are missunderstanding what i (and others i guess) do,
sure you can say its OCD, but if you trust you ears you can take half an hour time to compare things, and just use the one you like best, tho my last take on resampling/filters was more of a conclusion of multiple half hour sessions (and a bit of reading to understand things better, so i could somewhat connect the dots...)

the thing is also, with slow roll off and the dac PLAYING at 44,1khz i definitely heared cons in the sound, tho this becomes unimportant when upsampling to higher samplerates before the dac and in these cases i actually preferred the slow roll off types

there is a point where you start listening for (natural) transients and overall natural-ism and imo thats the point where you start hearing differences with different filters