Kronos Piano 44100Hz - SoundCloud's distortions issue

[mihmix]

I only hear very tiny distortion when you push the dynamics after 2 minutes.

I suspect you have the recording level a little hot. Or you can cut the mid to high FX a tad.

I clearly hear distortions on the 6th second with very low volume. Its like tiny tremblings... You can download the source wav-file. It has no any tremblings.

[mihmix]

there's nothing to do !
the problem is that it's converting from
high quality uncompressed audio to
128 Kbps Mp3, medium-low quality
LOSSY compressed audio;
lossy means that something get lost,
and that is the distortion you hear.
In solo piano playing, the distorsion is
udible, with more instruments distorsion
become less evident.
All you can do is upload audio
in best quality possibile, avoid other
conversions. At least soundcloud preserve
original file for download.

Is it really nothing to do? My audio file was already in the best quality:(
The first idea was to decrease the quality in the better manner than the soundcloud does. Are there any thoughts about that?

[mihmix]

Beautiful!! Who is the author of this piece? I'd love to try playing it as well.

Thanx!!! Actually this piece was just my one shot improvisation. I was inspired by the sunny autumn day walking in forest. I promise to share some photos later

[synthguy]

This is just a problem with Soundcloud compressing the audio with artifacts. Unless someone says any different, I'd say to upload stuff at the highest quality you can, converted to 44.1khz yourself if you can do that. And don't expect perfection from Soundcloud, that will help a bit.

As others have said, that's an excellent performance.

[SanderXpander]

The soundcloud website says they transcode all files to 128kbps mp3, but it doesn't say if it also does that if you supply the file in that format already. Possibly your conversion could be better than Soundcloud's. Although that is a pretty crappy format to begin with so you'd also lose the high quality download you at least have now.

[synthjoe]

Possibly your conversion could be better than Soundcloud's.

Unlikely, just as you say. MP3 (and pretty much all other lossy formats) suck, when it comes to quality rendition. It is convenient due to the small file size - but if you want quality, even higher bitrate MP3 will fail to deliver that. It is good for what it was invented for, which is not quality music rendition. CD resolution is bad enough already, better to not to top that with MP3 or other lossy compression.

Nice impro, good playing. Thanks for sharing.

[mihmix]

As I promised here some photos for this audio:
http://www.youtube.com/watch?v=y_5nban29Ig

Youtube also uses the audio compression and makes distorions as SoundCloud...

[SeedyLee]

You're confusing two things. 44.1k and 48k are not bitrates, they're samplerates - the amount of times per second the converter measures or outputs a value. There is no logical reason for samplerate conversion to cause distortion. ... The bitrate is usually 16 or 24 bit and it defines the precision with which that value is measured.

My understanding, and apologies if I'm incorrect, is that bit rate referes to the number of bits in a stream over time, whereas word size affects the resolution of a sample. The word size in isolation cannot, by definition, be the bitrate because it is not a rate. The bitrate would be the word size (16/24 bits) multiplied by the number of samples over time. E.G. for 16 Bit 44.1 kHz audio, the bit rate is 705.6kbit/s.

Not that the nomenclature makes any difference whatsoever to the original poster's question!

[cwall1108]

This first bit might not help, but I'm making sure I'm not confused. A few people have referred to 16/24 bit-ness as "bit rate". I don't believe that's correct. That is bit depth. Bit rate is the number of bits transmitted over time. The bit rate for a 44.1k sample @ 16 bit would be lower than a bit rate at 48k/16bit. Similarly 44.1k/24 would be of higher bit rate than 48/16. If the sample set is mapped to a Cartesian plane, sample rate is the number of demarcations on Y for time, and bitness is the number of demarcations on X.

SO, to your problem, distortions are normal sometimes in a high to low conversion when there is no "preparation". As others have mentioned the piano sample is very robust dynamically and that translates over to complexity in the X and Y as it were. There's a few stages where things can go pear shaped

1) Dithering - without knowing soundcloud/youtube's encoders/converters, sometimes straight conversion causes artifacts, distortion and glitches. 48k has a greater dynamic range than 44.1k.
2) Lossy encoding - in psychoacoustic modeling various compromises have to be made in sound and sometimes a "fast" encode of something into VBR (variable bit rate) leaves more artifacts because the complexity changes faster than a one sweep encode can comfortably react.

My recommendation would be to use your daw, compress and normalize within reason (don't kill your performance's dynamics), and run a dither plugin to take it down from high to low rate (targeting 44.1k/16b). Then fire that sucker over and see if you still have your oddities.

Hope that helps in some way.

[SanderXpander]

Sorry, you are of course correct concerning naming. Got the "rate" stuck in my head for some reason.

I don't see how 48k has greater dynamic range than 44.1k though, as I understand it, bit depth deals with dynamics and samplerate deals with, eh, "audible waveform accuracy" if you will, especially in higher pitches. In other words, if have a 10kHz pure sine, a 44.1k sampling will only represent that in 4.41 steps, making it pretty blocky.

Actually, as I'm writing this, it's pretty clear there's a big interaction between bit depth and samplerate. I've just always been able to hear the difference between 16 and 24 bit, but not much between 44.1k and, say, 48k. Though I've never actually tried 192k and I mostly work with digital gear or in the box.

[jeremykeys]

How wxactly are you getting you music onto soundcloud? Myself, I have to record audio onto my stand alone audio recorder, mix it and burn a CD withiti. Then I physically have to load the CD into the computer, save it into my library and upload the song to Soundcloud.

Her's an example:
http://soundcloud.com/jeremykeys/01-slow-trip

My speakers on my computer bare work so I can't tell if there is distortion or not but if somebody else can here distortion it would be good for me to know. I also only have one side working. Have to get that fixed one day!

[cwall1108]

I don't see how 48k has greater dynamic range than 44.1k though, as I understand it, bit depth deals with dynamics and samplerate deals with, eh, "audible waveform accuracy" if you will, especially in higher pitches. In other words, if have a 10kHz pure sine, a 44.1k sampling will only represent that in 4.41 steps, making it pretty blocky.

Sometimes dynamic range is used confusingly as well. In audio and digital photography sometimes the term applies to number of steps between the largest number the bit depth can store and zero. This thread on gearslutz.com is pretty interesting, and makes my head hurt at the same time http://www.gearslutz.com/board/masterin ... nship.html

I suspect a dithering problem in this case and, since I didn't mention it, I really enjoyed the piece and the youtube vid that went with it

[synthjoe]

I suspect a dithering problem in this case...

Nope. It is an artifact due to the frequency bands used in the lossy compression. Dithering is something else. Just like 48 kHz has nothing to do with dynamic range (or very little) and bitrate being different both from sample rate and from bit depth.

Gents, I'm sure we all can do better - so a little attention when typing and some reading to be done, if needed. Not applying these terms consistently is one of the prime reasons why jokers can get along in the business and newcomers get confused, so please enforce some discipline!

Thank you for your attention. [/Service announcement]

[cwall1108]

Gents, I'm sure we all can do better - so a little attention when typing and some reading to be done, if needed. Not applying these terms consistently is one of the prime reasons why jokers can get along in the business and newcomers get confused, so please enforce some discipline!

Thank you for your attention. [/Service announcement]

Well, I'll take that in as nice a way as possible, because I could be wrong but not for lack of thought or reading comprehension.

I've run into dither problems where the conversion to a lossy format exacerbated the problem. My suggestion (I wasn't explicit - my mistake) was trying a dither to 16/44.1k, submitting, and seeing if the issue persisted. This was also because earlier in thread some folks mentioned being unsure if the submission of an already compressed file to soundcloud would result in a decode/recode. He's got control of what's submitted, but not the encoder on the other end right?

So, between being a simple test, manipulating as little as possible and fitting a scenario I've run into a couple of times it didn't seem like a horrible theory. I may be new to this board, but this does not automatically consign me to the uninformed idiot bucket.

And yes, dynamic range is used in multiple contexts. In pure audio it can mean one thing, and in the digital domain it can mean another - my context was the latter.

At any rate, this seems to have very little to do with the Kronos or Korg itself at this point right?

[synthjoe]

Well, I'll take that in as nice a way as possible

It is also how it was meant, I'm not picking at you at all - not even sure why you think it was addressed at you specifically. More of a generic statement I tried to make.

I've run into dither problems where the conversion to a lossy format exacerbated the problem.

Highly likely. However, it is not the problem of dithering, but mostly a problem originating from the theory (and working) of lossy compression. Noise is a very difficult signal to compress (data reduction-wise, that's why cymbals and snare - two instruments very much based on a noise-like base waveform - are the ones you notice lossy compression on first). Sine, or a combination of a finite number of distinct sines on the other hand are very simple to compress faithfully - that's what most lossy compression shcemes encompass (sort of modified harmonic synthesis - reversed, when encoding). And real music lies in between those two extremes, meaning that some music will be more compliant to compression than others.

My suggestion (I wasn't explicit - my mistake) was trying a dither to 16/44.1k, submitting, and seeing if the issue persisted.

Dithering is the act of adding noise, which reduces the inherent 'ringing' sound (quantisation artifact) of linearly digitised signals by effectively adding a low level digital noise to the data stream. Not sure what you've meant here.

I may be new to this board, but this does not automatically consign me to the uninformed idiot bucket.

If that's what you've read in my post then please read again. Or maybe I have ambiguously expressed myself, in which case I apologise - not a native speaker of English, myself.

And yes, dynamic range is used in multiple contexts. In pure audio it can mean one thing, and in the digital domain it can mean another - my context was the latter.

Sorry for having to correct you, but it (dynamic range) means the exact same thing in any context - it is the ratio between the smallest and the largest representable distinct value, usually expressed in decibels and calculated as 20*log(VM/Vm). Be it picture quality, sound level or electric signal (and VM indicating the largest representable value, whereas Vm indicates the smallest distinctly quantifiable, to be precise).

At any rate, this seems to have very little to do with the Kronos or Korg itself at this point right?

Indeed and hence we seem to have the mutual wish to stop it here, right? (Well, maybe far more to do with those than it would seem at first, but let's not go there, I propose, indeed it was a bad case of hijacking - for which I apologise.)