16 vs 24 Bit

[Mephesto]

You show your ignorance with that 32 kHz statement, making everything else you said pointless since you missed that one so badly. The Nyquist-Shannon sampling theorem requires 44.1 kHz or above to fully reproduce audible sounds. The fact that you or others may not be able to hear above 16 kHz does not mean that all people are in the same boat.

LOL, whatever underfunded school you went to you dropped out of m'friend. The Nyquist-Shannon sampling theorem merely states that you must sample at twice the amount of hertz to faithfully reproduce the full frequency response. The 44.1 khz number came from the early video cassettes that they used to record sound on which had 245 lines, 3 samples per line and 60 frames per second which totals to 44,100 samples per second. This was ideal because it was above a doubled 20khz which is the maximum human ears can ear and it was the medium with the closest match at the time and thus became a standard.

Since anyone older than 14 can't hear above 20khz and anyone older than 25 can't hear above 16khz, 32khz is perfectly suitable for casual listening.

I have golden ears, I can ABX a 320 kb/s MP3 and I can personally hear up to 19.2 kHz if it's a tone at maximum volume and I concentrate, but most content in the upper shelf of a typical song is scarce, intermittent and a low dB which is easily masked by the louder, lower shelf. Around 14khz is where I can reliably start to notice dulling if it goes any lower.

Seriously, double-check your shit before correcting me next time.

[hello_hello]

You'll never be able to. AC3 is lossy. Therefore it doesn't have a fixed bitdepth. It can never be 24 or 16 bit etc.

I always use MediaInfo to check the data. It says 16 bit. So that this is a wrong information?

I'm not 100% sure. Maybe the Dolby spec is for it to be decoded to 16bit so that's what MediaInfo reports, but I tried converting a 24bit wave file to AC3 and MediaInfo still reports it as being 16bit. It seems to always report lossy DTS as 24bit so maybe that's inline with the specification for decoding it too.

If in doubt, I go by what foobar2000 tells me as it doesn't seem to lie. Here's what the codec information looks like for a 24 bit wave file.



Same again for AC3 (or any other lossy codec). Note there's no longer any "bits per sample" information.



And when configuring the encoders, there's no option to set a bitdepth for any of the lossy formats as there is for the lossless ones.

[hello_hello]

A quick and simple "casual listening" test for anyone who wants to try it.

I'm no audiophile, but the difference between 32k and 24k was immediately obvious on my headphones.

I had a fresh listen today with rested ears, once again just through PC speakers. Now I can make some noise without disturbing anyone I "think" I can hear a very small difference between 44.1k and 32k through my PC speakers, but only if I give the 14k fader on the EQ a serious push. With the EQ flat they sound the same. With 14k boosted it's fairly obvious a lot of high stuff is missing when sampling at 24k, so using decent headphones I have no doubt you'd hear a difference.

[Brad]

Why do people on the internet find the need to lie about the most miniscule crap imaginable?
It's obvious there's a difference between a samplerate that low and 32kHz, nobody was arguing for 24kHz.
Trolls these days...

What is wrong with you? I was responding to a post which stated he couldn't tell the difference and invited people to comment:

Admittedly my ears took a battering at work tonight so I might try again tomorrow, but I wasn't able to convince myself I could hear a difference until I dropped the sample rate down to 22050Hz and even then I'd probably need to ABX it to be certain.

[pandy]

A quick and simple "casual listening" test for anyone who wants to try it.

I'm no audiophile, but the difference between 32k and 24k was immediately obvious on my headphones.

Are you sure that you can distinguish between two exactly same audio where difference is only sample rate and your equipment perform correct reproduction?

[pandy]

A quick and simple "casual listening" test for anyone who wants to try it.

Any advice how to perform test and compare vegetables with fruits?

[hello_hello]

A quick and simple "casual listening" test for anyone who wants to try it.

Any advice how to perform test and compare vegetables with fruits?

I'm not sure what you're asking, but if you're wanting a method to do ABX testing:

http://web.archive.org/web/20070817111509/http://ff123.net/abchr/abchr.html

http://www.foobar2000.org/components/view/foo_abx

[pandy]

A quick and simple "casual listening" test for anyone who wants to try it.

Any advice how to perform test and compare vegetables with fruits?

I'm not sure what you're asking

Once again - how compare incomparable?

[hello_hello]

Once again - I don't understand the premise on which the question is based, but posting in circles is kind of fun anyway.

[pandy]

Once again - I don't understand the premise on which the question is based, but posting in circles is kind of fun anyway.

I mean: why you trying to pretend that provided samples are correct for sample rate comparison - this is my question - how compare potato with apple - as a food? as a pseudspehere shape? i asking about methodology behind provided samples.

I can understand when you trying to compare this samples bellow:

[hello_hello]

You didn't ask about the methodology behind providing samples, you asked about comparing vegetables with fruits and about comparing the incomparable, but unfortunately I'd not managed to get my crystal ball working to decipher what you were talking about. Now you've provided samples yourself without explaining the mythology behind those or without attempting to explain why they're more comparable, so I still have no idea what point you're trying to make. Is English your second language because obviously there's a communication problem.

After trying the crystal ball again I'm thinking maybe you took a low sample rate version and resampled it at a higher sample rates in order to compare them but I've no idea what the point of doing so might be. Did you understand why I uploaded those samples in the first place? They were just "quick and simple" examples offered as part of a discussion regarding how low a sample rate you could get away with for "casual listening", and not something from which scientific conclusions were expected to be made.

If you really need to know the methology used, I simply took the beginning of a CD track and re-converted it to wave files with foobar2000 while changing the sample rate each time with the Resampler DSP. I have no idea how clever a resampler it is compared to other methods. They were quick and simple sample rate comparisons as part of a discussion.

[pandy]

You didn't ask about the methodology behind providing samples, you asked about comparing vegetables with fruits and about comparing the incomparable, but unfortunately I'd not managed to get my crystal ball working to decipher what you were talking about. Now you've provided samples yourself without explaining the mythology behind those or without attempting to explain why they're more comparable, so I still have no idea what point you're trying to make. Is English your second language because obviously there's a communication problem.

My point was that generating those samples was unnecessary as every higher sample rate version have more energy up to resampler stopband thus this comparison is not a comparison or judging - to judge how you can hear difference between sample rates you need to have same signal source that is band limited then you can compare different sample rates - in case of your samples each time you comparing different signal (i know this is the same recording but still - signal is completely different). And yes, English is not my native, also it is self learned thus i apologize for any mistakes and communication errors.

After trying the crystal ball again I'm thinking maybe you took a low sample rate version and resampled it at a higher sample rates in order to compare them but I've no idea what the point of doing so might be. Did you understand why I uploaded those samples in the first place? They were just "quick and simple" examples offered as part of a discussion regarding how low a sample rate you could get away with for "casual listening", and not something from which scientific conclusions were expected to be made.

More or less, your source file was filtered in sinc filter (lowpass, 7.75k) then -3.01dB gain was applied, next samplerate conversion to 16000Hz was performed, later this signal was used to create remain samplerates and now they can be compared as they have almost identical energy and they are correlated and comparison can be made.

Casual listening is confusing description - with few techniques (like spectral band replication/extrapolation) probably 16000Hz sample rate will be enough or even less than 16000Hz and so what? - for casual listening comparision one you intended to perform, low pass filter should be used not different sample rates as key is low pass filtering not sample rate.

If you really need to know the methology used, I simply took the beginning of a CD track and re-converted it to wave files with foobar2000 while changing the sample rate each time with the Resampler DSP. I have no idea how clever a resampler it is compared to other methods. They were quick and simple sample rate comparisons as part of a discussion.

Each time you just generated different signal - going on fruit methodology - apple, same type, at various stage of grow - from barely visible through various stages grow up to perfect, ripe fruit and where is point for this - unripe apple vs ripe apple - what is your choice? And this have nothing to do with sample rate but with bandwidth of the signal itself - decent resampler just perform antialiasing lowpass filtration but you can resample signal without lowpass and you can introduce aliasing and i know people that find such audio better than correctly antialias filtered.
This is the point im trying to make...



Additionally - i've performed downsampling by 2 i.e. down to 22050Hz and upsampling back to 44100 but without antialiasing filtering - what is your opinion? About samplerate conversion without anitaliasing? (btw file is less loud, -3.01dB - reason is that in original file clipping occure - visible also on spectrogram)

[hello_hello]

I think we're just looking at the topic from opposite angles. The original discussion was in regard to the minimum sampling rate required for "casual listening". In the context of the discussion, to me we were effectively discussing the minimum frequency response required for casual listening, but of course that would in turn dictate the sampling rate required.... which is why "sampling rate" was being referred to.

The idea of posting samples wasn't to compare sampling rates as such, but to (roughly) determine if a 32k sampling rate could reproduce a frequency response adequate for "casual listening" and whether the difference between 32k and 44.1k would be noticeable.
Yes, I guess I could have used just a low pass filter and left the sampling rate out of the equation, but that's why I called it a quick and simple test.... it was never intended to be anything very "scientific". Plus then there'd no doubt be people questioning the low pass filtering used if the sampling rate remained the same each time..... there's really no way to win (please everyone) when posting samples from a quick experiment like that (do a Google search on "ABX test" and you'll probably even find links where people claim the ABX test itself skews the results of an ABX listening test..... sigh....).

Additionally - i've performed downsampling by 2 i.e. down to 22050Hz and upsampling back to 44100 but without antialiasing filtering - what is your opinion? About samplerate conversion without anitaliasing? (btw file is less loud, -3.01dB - reason is that in original file clipping occure - visible also on spectrogram)

It obviously sounds quite different (at least compared to the original 44.1k source), but unfortunately it's night here, everyone else is asleep and I can't find my headphones, so I'll need to have a proper listen tomorrow.

By the way.... for self-taught English your English is very good. Better than some native English speakers I know. I actually didn't expect you to say it was your second language, I was just having a little dig because I thought you were being purposely vague when it came to explaining the question. Sorry.

[pandy]

I think we're just looking at the topic from opposite angles. The original discussion was in regard to the minimum sampling rate required for "casual listening". In the context of the discussion, to me we were effectively discussing the minimum frequency response required for casual listening, but of course that would in turn dictate the sampling rate required.... which is why "sampling rate" was being referred to.

Once again - key to different perception was signal bandwidth not sample rate and this was my point - to test sample rate, signal must band limited to lowest common factor to provide correct methodology of test.
With help of lowpass filter and without changing sample rate similar effect can be achieved and this is key to problem - everything was related to antialias filter not sample rate per se.

The idea of posting samples wasn't to compare sampling rates as such, but to (roughly) determine if a 32k sampling rate could reproduce a frequency response adequate for "casual listening" and whether the difference between 32k and 44.1k would be noticeable. I just posted the samples using lower sampling rates while I was at it.

Still - samplerate conversion without antialiasing filter is possible and some people claim that this kind of sound is perceivable better than signal with correct antialias filter in processing chain - they don't hear or they ignore (as hearing process is deeply related to brain perception and brain can ignore/mask some distortions) alias. In fact there large group of audiophiles that prefer so called NOS DAC which at some circumstances can produced aliased sound - they perceive this kind of sound as better than sound form normal i.e. Non-NOS DAC's.

Additionally - i've performed downsampling by 2 i.e. down to 22050Hz and upsampling back to 44100 but without antialiasing filtering - what is your opinion? About samplerate conversion without anitaliasing? (btw file is less loud, -3.01dB - reason is that in original file clipping occure - visible also on spectrogram)

It obviously sounds quite different (at least compared to the original 44.1k source), but unfortunately it's night here, everyone else is asleep and I can't find my headphones, so I'll need to have a proper listen tomorrow.

I will try to prepare some audio for 32kHz sampling rate - this is not so easy as sampling rate ratio between 32 and 44.1 is beyond software processing capabilities - with some tricks i think some people may like it even more than original audio...

By the way.... for self-taught English your English is very good. Better than some native English speakers I know. I actually didn't expect you to say it was your second language, I was just having a little dig because I thought you were being purposely vague when it came to explaining the question. Sorry.

There is no need to Sorry - i didn't feel offended or such thing, i know that my English is limited but anyway thanks for positive opinion - it is appreciated.

--

Crude example of resampling without bandlimited signal - alias obviously audible but still some people prefer this kind of sound over "dull" correctly processed audio.

[Mephesto]

Almost 10 long posts about antialiasing? Wow.

I checked the 32k and 44.1k songs and I see no signs of aliasing at all. Whatever filter was used did the job fine. I'd be very surprised if someone correctly ABX'd the 32k and 44.1k flacs.

[hello_hello]

If you're referring to my original samples, I just used foobar2000's Resampler DSP when converting. I've no idea how clever it is or isn't when it comes to filtering.... I've probably only needed to resample something a couple of times so I can't say I've ever really thought about it.... but it was only intended to be a rough example anyway.