Android resampling

[bigshot]

There is no audible difference between 16 bit and 24 bit, nor is there an audible difference between 44.1 and 96. The conversion to MP3 is the only thing you should be worried about. If the conversion is done at LAME 320 VBR, you should have an identical sounding file, not matter what the data rate of the original. I've never run across anyone who can discern a difference with that setting.

If you think you can hear the difference between high data rates and 16/44. I suggest you set up a simple line level matched, direct A/B switched blind comparison and find out for yourself. If you're worrying if a file is theoretically different. Yes, any processing or conversion, even upsampling would make the file theoretically different.

 

[castleofargh]

I just want to get my question answered..... with the sources like official android patch note or etc...... and for 16bit and 24bit in now a day many of song gets mastered at 24bit/96khz so when you downscale that into 16bit, aac, mp3 it losses many details that you could be also called dnr. With your observation human shouldn't hear any difference with recorded instrument playing and live instrument playing becuz 16bit/44.1khz is already above the maximum of human hearing but it's not like that. Btw 24bit gives more than 1.6m steps but 16bit gives only 65k steps that shows 24bit audio is much more closer to pure analog sounds. No difference and not audible is totally different topic we can find out that there is no difference with many measurements like thd snr dnr sinad etc but only way to prove "not audible" is doing those blind test and that might change through out times. At 90s people believe that maximum human eyes fps was 30 so they made tv that has upto 30fps but at 2019 we know that is false. It might be true that human can't hear difference with pure 16bit and pure 24bit but it's true difference story about getting resampled and going through digital filter (not for dac digital filter that's goes through during resampling) that can effect the sound somehow. You're focusing at the point of "audible" but I'm focusing on the fact that it's different or not. Android mixer is fixed to 48khz that if I'm listening to 44.1 or other songs that's not 2 multiple it's true that it's lossing some tiny part of it. Try this blind test you'll be able to tell the difference that has been resampled lossy and original https://hydrogenaud.io/index.php/topic,6651.0.html

the DAC used, maybe the app, even perhaps the specific phone or DAP model if they happen to put some fancy audio stuff in it, are going to play a role in the choice for the final resolution output. so I'm not sure you're going to get more specific answers to a global question.

about audibility, well it doesn't really matter. if you want something for some reasons, of course you should try to get it. nobody here is going to try hinder your free will. what causes some reactions is your argumentation about why we should strive for the higher resolution, because it's failing to look at the big picture and is getting worst with each new post IMO. I would guess from what you say, that it is in part due to not having a proper understanding of audio sampling, reconstruction, and waves in general. that stuff is not intuitive, trying to figure it out with common sense and lose analog concepts only lead to false conclusions. said conclusions will be very popular on audiophile forums, but only because the same false reasoning lead them to the same false conclusions. those stuff aren't fact based knowledge.

leaving that aside, there is no 24bit fidelity. only 24bit containers. the distinction is significant. studio noises, mic/preamp self noise and distos, singers and instruments playing at their actual levels even if it's not convenient for SNR and what not, ADC limits. then mixing and mastering. what we end up with on hires albums is often a different master, and some very accurate background noises that you probably can't hear. the effective difference with a 16/44 conversion is much less than most people imagine.
then comes the same routine about playback resolution, listening levels, room noises, and of course human hearing. it would be really great if we could consistently achieve actual 16/44 resolution from start to finish.

the eye argument is wrong, what is tested and confirmed for around 30 frames per second is not what leads to detecting blinking at 60HZ on an old CRT, or tearing effect in a game when turning rapidly even on high refresh rate modern monitors, or whatever other situations you had in mind. taking something out of context doesn't prove it wrong. also as far as I know, people were already pushing for higher sample rates in the 20's. cinemas have been displaying movies using multiple of 24 as shutter speed for a long time to give a smoother experience than just displaying a movie at actual 24fps. and those are just counter examples on the top of my head.

the hydrogen link as far as I can tell isn't what you think it is.

it might be a good idea to stick to your original question and ignore people trying to drag you somewhere else(including me ).

 

[bigshot]

I think he's asking if the file is changed by being bumped up to a higher data rate. And the answer is yes, it has been changed. The file size is a lot larger for the exact same sound.

 

[dazzerfong]

I just want to get my question answered..... with the sources like official android patch note or etc...... and for 16bit and 24bit in now a day many of song gets mastered at 24bit/96khz so when you downscale that into 16bit, aac, mp3 it losses many details that you could be also called dnr. With your observation human shouldn't hear any difference with recorded instrument playing and live instrument playing becuz 16bit/44.1khz is already above the maximum of human hearing but it's not like that. Btw 24bit gives more than 1.6m steps but 16bit gives only 65k steps that shows 24bit audio is much more closer to pure analog sounds. No difference and not audible is totally different topic we can find out that there is no difference with many measurements like thd snr dnr sinad etc but only way to prove "not audible" is doing those blind test and that might change through out times. At 90s people believe that maximum human eyes fps was 30 so they made tv that has upto 30fps but at 2019 we know that is false. It might be true that human can't hear difference with pure 16bit and pure 24bit but it's true difference story about getting resampled and going through digital filter (not for dac digital filter that's goes through during resampling) that can effect the sound somehow. You're focusing at the point of "audible" but I'm focusing on the fact that it's different or not. Android mixer is fixed to 48khz that if I'm listening to 44.1 or other songs that's not 2 multiple it's true that it's lossing some tiny part of it. Try this blind test you'll be able to tell the difference that has been resampled lossy and original https://hydrogenaud.io/index.php/topic,6651.0.html

That just demonstrated a fundamental misunderstanding of how digital audio works.

24 bit audio doesn't mean it's more granular, it means it can go louder (or quieter). See, you're thinking that it operates on an absolute scale. Audio doesn't work like that - it operates on a decibel scale (i.e. it's relative).

And that's complete bollocks that the human eyes can detect 30 FPS. Know how I know that's bollocks? Coz there's 60 frames interlaced back in the 90s too. That 30fps thing is a misattribution to the colour changing 30 times a second without brightness changing. Also, your nerves can fire up to 1000 times a second.

That being said, the question asked was the resampling behaviour in Android, which is fair enough. But lying to prove a false point doesn't make it any better.

 

[djyang0530]

So no one able to answer my question with reliable sources...... at the point that many songs get mastered at 24bit going down to 16bit is just making that file lossy a little part of songs will get chopped off. It also effects volume of sound but that's not it. If higher than 16bit audio file is and snake oil why don't you guys just sell all your dap, dac and etc apple lighting gender and Samsung c type dongle has arround 100db snr dnr that's already higher than 16bit dnr
And especially for resampling into totaly different facter losses lots of their details. The main point of hifi system is to reappearence original sound that has recorded and having higher audio file makes the sound be near to original sound since digital signal isn't curve. If you look at the files that has higher than 20khz bitrate you'll be able to see spectrum that is higher than 20k and sometimes goes up to 50khz that shows there is also some sound in that range that can effect other sounds. Measurement and spectrum shows is that there is difference on resampled audio and higher bitrate audio and most simple thing to find out is simply blind testing to your self and especially for resampled audio because of resampling filter it's very easy to be audible whether that's positive way or not.

 

[Killcomic]

 

[djyang0530]

And GIVE ME scientific reasons that why 24bit is not audible and resampling 44.1to 48 isn't audible now a days with sources

 

[old tech]

The main point of hifi system is to reappearence original sound that has recorded and having higher audio file makes the sound be near to original sound since digital signal isn't curve

With that comment, I think you just misunderstand digital audio and therefore cannot get the answer you are looking for. For example, using that plot to demonstrate that a digital signal isn't curved is no different to looking at a scatterplot of magnetic particles of analog tape and then saying that an analog signal is not curved. That would be absurd as it ignores the role of the tape head in converting the magnetic properties into a "curve" signal, just as you are ignoring the role of a DAC in converting the digital signal to a "curve' signal.

 

[bigshot]

And GIVE ME scientific reasons that why 24bit is not audible and resampling 44.1to 48 isn't audible now a days with sources

Bitrate involves the noise floor. 24 bit has a noise floor of -144 dB. 16 bit has a noise floor of -96 dB. Your living room has a noise floor of about -30dB at best. Simple math. 30 dB plus 96 dB is 126 dB. 120 dB is the threshold of pain. The only way to hear a noise floor below 16 bit is to incur hearing damage. The noise floor of a good recording studio is significantly higher than 16 bit.

Sampling rate involves frequency response. According to the Nyquist Theory, in order to perfectly recreate a frequency, you need a sampling rate of double the frequency. So 44.1 can reproduce up over 20 kHz. 20 kHz is the limit of human hearing. Anything above that is inaudible. If you can't hear it, it doesn't matter. So 48 is overkill. 48 is mainly used for video because it divides neatly with video frame rates.. It doesn't improve quality.

There you go, my man. Look it up for yourself. Google Nyquist Theory and bitrate noise floor. Have fun with it. I'm happy to provide you with breadcrumbs if you are interested in following them. If not, don't let the door hit you in the butt!

 

[Killcomic]

Bitrate involves the noise floor. 24 bit has a noise floor of -144 dB. 16 bit has a noise floor of -96 dB. Your living room has a noise floor of about -30dB at best. Simple math. 30 dB plus 96 dB is 126 dB. 120 dB is the threshold of pain. The only way to hear a noise floor below 16 bit is to incur hearing damage.

Sampling rate involves frequency response. According to the Nyquist Theory, in order to perfectly recreate a frequency, you need a sampling rate of double the frequency. So 44.1 can reproduce up over 20 kHz. 20 kHz is the limit of human hearing. Anything above that is inaudible. So 48 is overkill.

There you go, my man. Look it up for yourself. Google Nyquist Theory and bitrate noise floor. Have fun with it.

Dammit Bigshot! You stole my thunder!
I actually knew this... *Sulks*

 

[bigshot]

don't worry. you're a cool cat.

It's ironic that sound science would be the place where people go to get the most basic of basic information. I guess it shows how devoid of information the rest of Head Fi is.

 

[Killcomic]

don't worry. you're a cool cat.

It's ironic that sound science would be the place where people go to get the most basic of basic information. I guess it shows how devoid of information the rest of Head Fi is.

Well, considering the amount of times DAPs have been described as “musical”, I think we’re looking at a pretty empty barrel.

 

[djyang0530]

Bitrate involves the noise floor. 24 bit has a noise floor of -144 dB. 16 bit has a noise floor of -96 dB. Your living room has a noise floor of about -30dB at best. Simple math. 30 dB plus 96 dB is 126 dB. 120 dB is the threshold of pain. The only way to hear a noise floor below 16 bit is to incur hearing damage. The noise floor of a good recording studio is significantly higher than 16 bit.

Sampling rate involves frequency response. According to the Nyquist Theory, in order to perfectly recreate a frequency, you need a sampling rate of double the frequency. So 44.1 can reproduce up over 20 kHz. 20 kHz is the limit of human hearing. Anything above that is inaudible. If you can't hear it, it doesn't matter. So 48 is overkill. 48 is mainly used for video because it divides neatly with video frame rates.. It doesn't improve quality.

There you go, my man. Look it up for yourself. Google Nyquist Theory and bitrate noise floor. Have fun with it. I'm happy to provide you with breadcrumbs if you are interested in following them. If not, don't let the door hit you in the butt!

So if my room snr is 30db isn't it 96-30 to make the 1db of noise there need to me 66db of signal? Why is it + not -

 

[gregorio]

I just want to get my question answered..... with the sources like official android patch note or etc......
[1] and for 16bit and 24bit in now a day many of song gets mastered at 24bit/96khz so when you downscale that into 16bit, aac, mp3 it losses many details that you could be also called dnr.
[2] With your observation human shouldn't hear any difference with recorded instrument playing and live instrument playing becuz 16bit/44.1khz is already above the maximum of human hearing but it's not like that.
[3] Btw 24bit gives more than 1.6m steps but 16bit gives only 65k steps that shows 24bit audio is much more closer to pure analog sounds.
[4] It might be true that human can't hear difference with pure 16bit and pure 24bit but it's true difference story about getting resampled and going through digital filter (not for dac digital filter that's goes through during resampling) that can effect the sound somehow.
[5] You're focusing at the point of "audible" but I'm focusing on the fact that it's different or not.
[6] Android mixer is fixed to 48khz that if I'm listening to 44.1 or other songs that's not 2 multiple it's true that it's lossing some tiny part of it.
[7] Try this blind test you'll be able to tell the difference that has been resampled lossy and original https://hydrogenaud.io/index.php/topic,6651.0.html

Oh dear. It should be obvious that if you are going to post statements of fact in a Sound Science forum, then you really need to have some understanding of sound science. If you are not sure or don't understand the science then by all means ask, but then don't post incorrect/false statements of fact. So to address your false statements:

1. The amount of "details lost" obviously varies here. As the name indicates, a lossy codec (say aac or mp3) looses a lot more detail than a lossless format, say 16bit. The question is whether these lost details are audible, a question that's been answered countless times over the course of many years. However in your given scenario, an android device resampling to say 16/48, we're not dealing with "lossy" encoding/decoding! And no, you could not only call these lost details "dnr".

2. This statement is completely false and demonstrates a fundamental lack of understanding of both digital audio and sound recording! When we record an instrument playing we use microphones and mic pre-amps (typically several). These are analogue devices (that are far from perfect) and are not positioned where your ears would be when listening to a "live instrument playing". Therefore, you SHOULD hear a difference between a recorded instrument and a live instrument but it has nothing to do 16/44.1 digital audio!

3. This too is unfortunately completely false! 16bit or 24bit doesn't have any "steps", the output is a continuous analogue waveform (without any steps) that is NOT closer (or further) to pure analogue sounds, it IS pure analogue sounds!!

4. There's a couple of serious errors in your statement. Firstly, there is no such thing as "pure" 16bit or 24bit, ALL 16bit and 24bit recordings are resampled through a digital resampling filter, MORE than once, BEFORE you even load it into your android device!

5. Whether it's different or not is irrelevant, whether that difference is audible is what's relevant. The reason that whether it's different or not is irrelevant is because it's ALWAYS different. Even the exact same song file played twice in succession without changing any setting or playback parameters will be different because all the analogue components in your playback chain (amp and headphones for example) produce thermal noise and as thermal noise is random noise, it will be different every time!

6. What "part of it" is it loosing? In this example, no part of it will be lost! Some part of it may have been lost using 25+ year old resampler but I'm assuming your android device is newer than 25 years?!

7. You seem to have linked to the wrong test. The test you've posted demonstrates that one or two test subjects were able to successfully ABX the analogue output of two devices, a 1996 soundblaster computer card and a 1991 Yamaha CD player. I'm assuming your android device does not contain either a 23 year old soundblaster card or a 28 year old CD player?!

G

 

[dazzerfong]

OK, since there's a fundamental misunderstanding of what bit rate is (not in the lossy context, as in between 16 and 24-bit), here's a graphical explanation:

If your music is in 24-bit, and was truncated to 16-bit, notice how the red bar remains the same? That's because exactly zero music will take advantage of something that has the dynamic range of 24-bit audio. Probably because it'll kill you.

24 vs 16-bit audio does not mean that each 'gap' between volume gets 'finer'. Want to know why?

The decibel scale is not an absolute scale, it's relative. It means that, if you have more bits, you can express a higher (or lower) volume absolutely speaking, but it doesn't mean it gets finer relatively.

So if my room snr is 30db isn't it 96-30 to make the 1db of noise there need to me 66db of signal? Why is it + not -

Nope, that's not how it works at all. Your problem is that you're treating it as a summing game, when in reality decibels doesn't work that way.

Your room has no SNR, it can be considered just 'N' (noise) here. In other words, if you want a signal has an SNR of 96 in a room with 30 dB of noise, your signal's volume has to be 126 dB, or a gun shot going off next to you (a whimpy one, but still!).