dithering/quantizing

[ScienceOne]

Somebody god help me. waht the hella are these two things? do i need them? i made a song without using either and put it on my bands webpage ( www.oldfashioned.tk ). its just a rough draft of the song, do I need quantization or dithering for the final?????????

 

[ambi]

Quantizing is a midi thing. You can record a midi keyboard into a program that supports it and has quantizing, and you can lock the notes into a certain time signature. Lets say you played in 1/8 time, you tell it to quantize to that speed, and it locks it in perfect time. Now this is only with midi, you can't just record a regular drum kit and do this.

Dithering happens automatically when you bounce, or mixdown, or whatever term you like, into a stereo wave file. Lets say you have a 24 bit song, it has 13 tracks, all recorded at 24/48. Cd's are 16/44, so you need to dither down from 24/48 - 16/44. So you don't "need to use it", its not like its some special feature your missing, it's just something that happens if you do that.

Quantizing is cool if you are using midi intruments, but it can take some of the human feeling away from the playing. I guess they have a humanizing option though.

 

[Boray]

Let's say you have 24 bit sound. If your just choose to read the first 16 of the 24 bits, you will get 16 bit sound. If you just choose to read the first 10 or them, you will get 10 bit sound etc... It converts perfectly that way. A perfectly linear conversion.

Dithering is something else. It adds random values to the whole 24 bit values. Bigger values the less bits you "dither to". So if you "dither to" 16 bits, then values big enough to sometimes carry rests of the additions over to the first 16 bits are added. This way these rests quite magically represent the cut away bits inside of the 16 bits that are left. - A bit more noise to the sound, but also the cut away bits gets represented within the remaining bits.

/Anders

 

[Blue Bear Sound]

Let's say you have 24 bit sound. If your just choose to read the first 16 of the 24 bits, you will get 16 bit sound. If you just choose to read the first 10 or them, you will get 10 bit sound etc... It converts perfectly that way. A perfectly linear conversion.

This is truncation, NOT conversion..... and it's something to be avoided.

Dithering algorithms help smooth out the LSBs of a signal being downsampled to a lower word size.

 

[sonusman]

Let's say you have 24 bit sound. If your just choose to read the first 16 of the 24 bits, you will get 16 bit sound. If you just choose to read the first 10 or them, you will get 10 bit sound etc... It converts perfectly that way. A perfectly linear conversion.

Dithering is something else. It adds random values to the whole 24 bit values. Bigger values the less bits you "dither to". So if you "dither to" 16 bits, then values big enough to sometimes carry rests of the additions over to the first 16 bits are added. This way these rests quite magically represent the cut away bits inside of the 16 bits that are left. - A bit more noise to the sound, but also the cut away bits gets represented within the remaining bits.

/Anders

Boray, a perfect linear conversion? Try truncating a 24 bit file to 10 bits sometime. You will will get anything BUT a perfect linear conversion!!!

Ed

 

[TexRoadkill]

Bitrate conversions are not the same thing as dithering but dithering (or noise shaping) takes place when you do a bitrate conversion. It is not automatic unless that is how your software does it. Usually you need to turn it on or use a plugin.

Why Dither? Because it helps maintain the dynamic range of a 24bit recording even though it has been converted to 16bit. The actual dynamic range of a dithered 16bit recording would be similar to an 18 or 20bit recording.

 

[Boray]

This is truncation, NOT conversion..... and it's something to be avoided.

Dithering algorithms help smooth out the LSBs of a signal being downsampled to a lower word size.

Yes, this conversion is called truncation.

 

[Boray]

Boray, a perfect linear conversion? Try truncating a 24 bit file to 10 bits sometime. You will will get anything BUT a perfect linear conversion!!!

Ed

Try to do some maths and you will see that it's a linear conversion.

When truncating from 24 to 16 bits, the first 256 values (0-255) in the 24 bit sound becomes 0 in the 16 bit sound. 256-511 becomes 1, 512-767 becomes 2, etc, etc... A linear conversion of 256 steps for each 16 bit number.
You can test this by take any number from the 24 bit range (0 to 16.8 million), calculate how many percent of the total it is, convert it to binary, cut the last 8 bits off, convert it back (or simply divide the decimal value by 256), then calculate how many percent this value is of the 16 bit range (0 to 65535). You will get the same percent from both.


(Sorry, have no high quality mp3 to show that I know what I'm talking about) (Just a remark to BlueBear).

 

[ambi]

As far as he's concerned it turns a 24 bit file into a 16 bit file. Thats all he wanted to know. He now knows it's an integrated function not a special "feature" he was missing.

 

[esactun]

As far as I know....

Quantization refers to more than MIDI time-syncing. In general, it refers to grouping things into discrete units.

Say you want to rate your rack gear on a ten-point scale, using only whole numbers. One reverb may rate a 7 in your book, and another may rate a 7.4. Since you're only using whole numbers, this rounds down to 7. This is a type of quantization. It's also a type of distortion, as your rating scale cannot discern the difference between your two reverb units, and calls them both a 7.

In digital audio, quantization is what happens when you run DSP on your tracks. It's most noticeable on fadeouts and reverb tails in 16-bit (hence the old Sound Forge included a dithering option for certain volume envelopes, to counter this quantization noise). It's the sound of your track's bits being quantized-- the sterile, brittle, harsh, cold "digital sound" is the sound of bad and/or too much DSP. (Trust me, I know it all too well! ) This isn't so noticeable in 24-bit due to the greatly increased resolution. Bit depth sort of means that there's only so many discrete volume levels that can be represented -- sort of a larger version of the rack-gear rating scale. A process that wants a sample to say "22.6" will call it "23" instead.

You can see this happen in a wave editor by taking a nice pure 16-bit sine wave and reducing the volume greatly (keep it out of the noise floor though!!) and then increasing the volume back again. The nice sine wave now has little steps in it, and sounds different.

This is my understanding at least.... I'm sure a Bear will eat me if I'm wrong.

Sonusman had a big thread a while back on this that's great stuff. http://www.homerecording.com/bbs/showthread.php?s=&threadid=48346&highlight=QUAN+errors

 

[Boray]

As far as he's concerned it turns a 24 bit file into a 16 bit file. Thats all he wanted to know. He now knows it's an integrated function not a special "feature" he was missing.

I don't know what system he uses, but on the VS1680 it's very much a "feature" that you have to turn on/set. It's an integrated part of the mastering toolkit effect there. I bet it's the same in most mastering programs.

 

[TexRoadkill]

As far as he's concerned it turns a 24 bit file into a 16 bit file. Thats all he wanted to know. He now knows it's an integrated function not a special "feature" he was missing.

Dithering is not the same thing as simple file conversion and not an integrated function in most software and hardware.

 

[Blue Bear Sound]

Yes, this conversion is called truncation.

You idiot...... it's not a "conversion" at all - you're not converting anything, you're simply chopping-off bits........

 

[sonusman]

Indeed Bruce! If there was an actual "linear conversion", than the 24 bit audio's full dynamic range would enjoy a "linear conversion" to fit a 10bit dynamic range. Adding "perfect" would mean that you could do so without hearing horrible sonic artifacts. This would be known as "compression" in most cases.

But that is sematics! I jest!!!

Ed

 

[ScienceOne]

Ok, I think I got it guys, no need to get nasty now. I haven't used dithering (unless cubase does it automatically), so I hope the dynamic range isnt messed up. If anyone wants to tell me that it is, you can listen to the song at www.oldfashioned.tk . I'm only 16 so its not going to be perfect, but this is the very first song I have ever recorded....... EVER. Its not even the final mix, just a rough draft. Give me your thoughts, if youre bored.

 

[ambi]

well when i bounce in logic it automatically dithers as far as i know. But then thats logic. I guess i don't really know anything else.

 

[Bigus Dickus]

Man, calling someone an idiot for their use of words when English isn't even their primary language?

 

[Bigus Dickus]

doh, scratch that last remark. I just read a few of his "other" posts (I apparently missed all the fun earlier), and I see it's based on a lot more than his poor choice of terminology.

 

[Blue Bear Sound]

Exactly....!

 

[Boray]

You idiot...... it's not a "conversion" at all - you're not converting anything, you're simply chopping-off bits........

By cutting away bits, you are indeed converting to a different bit length. It works the other way around as well. You can CONVERT from 16 to 24 bit format by adding 8 zeros to every 16 bit word. That is hardly called truncation. Idiot!