Okay, moving this over from another thread, I’d like to pick up where TomS and I left off.
The discussion began because Tom was concerned about my use of 16 bit plugins such as blueline in processing. As I noted to him in a response, even if you have a system capable of processing the information in 64-bit mode (which, as I understand it, isn’t really 64-bit but two 32-bit floating point registers), the very moment you mix the track down to cd audio, lossless, or anything else that uses PCM, you go back to 16-bit, and defeat the whole purpose of using 24 bit, in the first place.
Now, my soundcard is not 24 bit, only 16. Like all 16 bit soundcards, it has a decibel response MAXIMUM of 96db. 24-bit sound cards have a signal to noise ration of 128db. Now this is a monstrous difference in bandwidth, as you might guess. However, aside from RedBook and Super Audio CD format (Audiophile equipment) there is NO WAY to listen to your audio in 24 bit unless you have a card with 24 bit. There are no 24 bit cd players. They are all 16-bit, 44.1 khz PCM.
In this case, you could listen to a song you recorded in 24 bit and get the full response of your audio, but you cannot transfer that to CD and keep the FULL song. It will be converted by the system to 16 bit, and there’s nothing any of us can do about that. I’m not up enough on the technology, but as I understand it, Analog was not subject to the limitations of Digital Audio, so the frequency responses were much better in Analog than in Digital. I imagine there are still recording studios out there that will do completely analog, but the same limitations exist for them that exist for us: the moment the music is converted over to where it can be put on a CD, it is rendered as 16-bit, and you’re back to where you started.
So now, knowing that we cannot get our music out to the public in any format EXCEPT 16-bit, I fail to understand where it is an improvement to record something in AWESOME QUALITY if no one will ever hear it but myself? All that would do is bring me down, I think
At this time, I’m planning my next computer, which will be an AMD 6 core, 64 bit processor, with probably 8 gigabytes of RAM. Why? So I can work on my music without having to mixdown in stages just so I can still get all the tracks onto one song. I dream of being able to run a true 48 track studio out of my home with as many effects as I want on each track. I could probably run 24 tracks now, as long as I used no effects, but the moment I add reverb or eq plugins to a track (let’s not even go INTO what pitch shift does to your cpu load, lol!) then my machine would bog down on me.
Any ideas? Criticisms? Disagreements? Here’s the place for them
Question: Would changing my preferences help me get a better end result? Right now, I’m recording in 44.1 khz. What if my wavs and mixdowns were done in 96 khz format? Would my resultant 320 kbps mp3’s sound better than the ones done from 44.1 khz master files?
Hmmmmm, time to play with my computer, lol!
My understanding is that the advantage of 24 bits is that it gives you the ability to record with more headroom when recording signals with uncontrolled levels like live musicians. If you record with peak levels of say -12dB in 16bit resolution you are actually only using 14 bits to represent your sound and things are getting a bit marginal. With 24 bit you can record and peak at say -18dB and you are still getting 21 bits of available resolution.
Some people think the digital “mix bus” sounds better with signals peaking around -18dB and if you sum a number of tracks your overall level will increase anyway, (there’s no need when working in 24 bit to try and get each individual track peaking near to zero.)
Once you are dealing with a controlled signal level and you can safely peak to near 0dB then 16 bit is probably good enough for most people.
Okay, Rick, that made sense to me, lol!
There is still the issue of the mixdown, however. Whether you make the end product into mp3, flac or wma for the web distribution, or you burn it to CD, it’s still going to be 16-bit, so your headroom is gone. Another thing I just thought of. Probably 80% of all computer soundcards are 16-bit. Only gamers and musicians are buying the 24-bit soundcards, the general public could care less what theirs is. So when you’re playing a DVD on a computer, you’re not getting the 24-audio then, either, the soundcard has to squeeze it into 16-bit to play it, right?
I found this wikipedia article on quantization which makes some points about recording audio in 24-bit. http://en.wikipedia.org/wiki/Quantization_(sound_processing)
Okay, Rick, that made sense to me, lol!
There is still the issue of the mixdown, however. Whether you make the end product into mp3, flac or wma for the web distribution, or you burn it to CD, it's still going to be 16-bit, so your headroom is gone. Another thing I just thought of. Probably 80% of all computer soundcards are 16-bit. Only gamers and musicians are buying the 24-bit soundcards, the general public could care less what theirs is. So when you're playing a DVD on a computer, you're not getting the 24-audio then, either, the soundcard has to squeeze it into 16-bit to play it, right?
oops, Sorry Nick, I wrote your name as rick, lol
I’d like to add my .02 cents here… for what it’s worth.
I’d like to think that the resolution of a .wav and their specs. can be measured in two ways…
Vertical amplitude is referred to as bit resolution…
Horizontal amplitude is referred to as Sample resolution…
e.g. 44.1 kHz.
and so-on up to 96 kHz.
for .wav files…
the horizontal resolution…
at variable bit rate or 320 kHz.
My easy understanding of these specifications goes something like this…
Take a photo using an old analogue camera with film…
place the negative in an enlarger to enlarge the image from say 5" x 7", to say, 8" x 10"…
A good camera with a good lense cam produce an image on the film that shows as well at a small enlargement as the same negative when enlarged to any size…
without appearing as a grainy reproduction…
A digital .wav file with poor A/D conversion specs, e.g. poor or low bit-and-sample specs will be grainy when plugs are added to the file when editing and mastering…
The more editing-and-mastering the poorer the file will sound as the number of times the file is manipulated… till it finally gets to the CD for distribution…
Then the buyer says…
that sounds good…
the buyer says what happened to the sound on that CD ?
On the other hand…
Each time the .wav file is edited-and-rendered, it looses it’s original pristine reproduction…
It’s best for the .wav file to start off with the best analogue electronics e.g.
Good Dynamics, as it (the signal) enters
the A/D converters, then return to analogue via the D/A Converter with the best analogue reproduction possible…
Even the quality of the digital converters is important, to the process…
That makes for good audio hearing… In My Opinion…
The key word here is “dither” a complex process of converting 24bit to 16 or even 8bit audio.
For example if you record at 16bit 41,000khz, you can mix down your song within n-track without the need to “dither” and take it straight to CD. However if you record at a higher bit rate such as 24bit@96,000khz you’ll need to enable dithering from the mixdown window options, provided you have done no dithering at this point. Although this is just a very basic method, it will process the audio.
From what I understand and this is just a basic understanding, the wave form of a 24bit audio, say a line that goes up and then down will be made up of series of many points tightly packed together (24bit tight), when you covert to 16bit without dithering, it removes every other point from the 24bit wave form leaving a sixteen bit wave form with gaps left behind. These gaps have no audio, resulting in the sample being lower in volume and quality.
With dithering, those empty gaps/points are replaced with noise, thus allowing the audio to maintain it’s volume with very little loss of audio or quality. There are many mastering/dithering programs in which different dither methods for different situations but the whole idea as I understand it is fill the gaps with the right kind of noise or delete them.
For example taken from within OZONE 3 various methods using the OZONE 3 app, (I don’t understand a good portion of this stuff but it does give me an the idea behind it)
MBIT+: This is a proprietary iZotope word length reduction technology that reduces quantization distortion with minimal perceived noise. While this might sound like a paradox, MBIT+ is a very smooth, quiet and almost “analog sounding” technology.
Type 1: Dither is applied using a “rectangular” distribution function. While this provides a dither noise source with a low amplitude, the dither noise can become modulated by the audio signal and vary in level, which is undesirable in many situations. Also, the non-linear quantization distortion is not completely suppressed in some situations with this low dither amplitude.
Type 2: Dither is applied using a “triangular” distribution function. This dither is larger in amplitude and completely suppresses the non-linear quantization distortion.
By shaping the dither noise, it is possible to provide more effective and transparent dithering by shaping the dithered noise spectrum. There are several different methods for shaping noise so that it is less audible yet still effective. Please refer to our online mastering guide for more information as to the technology behind these methods and how to apply them effectively.
Type 1 or Type 2 Shapes
None: No noise shaping is applied
Simple: High pass filtering is applied to the dithered noise.
Clear: The noise is shifted towards the Nyquist frequency, near the upper limit of our hearing.
Psych 5: A fifth order psychoacoustic shaping is applied to provide dither across the spectrum. The shaping is designed to move the noise away from frequencies that are heard as “louder” at low levels.
Psych 9: A more complex ninth order psychoacoustic shaping is applied.
In general, the “Clear” option is a safe bet for complex program material, although auditioning the dither against the Psych 5 and Psych 9 shapes may be more desirable in some cases.
Please note that Psych 5 and Psych 9 shapes are specifically designed to be used on audio with a 44.1 kHz sample rate. For other samples rates, use None, Simple or Clear shaping, or MBIT+ mode which is designed for effective word length reduction at any sampling rate.
The MBIT+ dither technology also provides options for noise shaping. You can control the aggressiveness of this shaping, ranging from None (no shaping) through Ultra (roughly 14 dB of audible noise suppression).
This is the target bit depth for the audio. For mastering for a CD, for example, you would want this set to 16.
Note that Ozone does not perform the actual conversion of the audio. After processing a mix with Ozone, it is necessary to then actually convert the audio to the desired bit depth in the host application. For example, if you have a 24-bit audio file, you can use Ozone to dither down to 16 bits. The remaining 8 bits are “padded” as zeros. Your file is still a 24-bit audio file, there’s just not anything but zeros in the lowest 8 bits. So when you then convert to a 16-bit file in the host app, the 8 bits (that didn’t have any audio in them) are discarded.
With this process in mind:
1) Do not perform any processing to the audio after it has been dithered with Ozone. You may perform level adjustment with the output gain sliders in Ozone (those come before the dither) but do not change any levels in the host app or with other plug-ins. Almost all host apps have their master faders after the effects slot, so any level adjustment in the host app will destroy the dither.
2) Do not put any plug-ins after Ozone if you are dithering with Ozone. The dither must be the last thing that touches the audio.
3) Turn off dithering in the host app. Basically, you just want to truncate (throw away) the bits, because they’re just zero anyhow.
Num Bits or Dither Amount
This sets the number of bits or amount of dither that will be used for the dither source. For Type 1 and Type 2 dither, in most cases 1 bit will be sufficient, but in some situations the “over-dithering” obtained by setting Num Bits to 2 can be useful. In MBIT+ mode, the dithering amount can be varied from None (noise shaping only) to High. No dithering or Low dither amount can leave some non-linear quantization distortion or dither noise modulation, while higher settings completely eliminate the non-linear distortion at the expense of a slightly increased noise floor. In general, the Normal dither amount is a good choice.
Selecting this option instructs Ozone to completely mute dither output (i.e. dither noise) when the input signal is completely silent (0 bits of audio) for at least 0.7 seconds.
Dither noise is random in nature and has a very low amplitude. However, after noise shaping, especially in aggressive dithering modes, the high-frequency dither noise is significantly amplified, and the overall dither signal can show spurious peaks up to -60 dB FS. If such high peaks are undesirable, you can enable the Limit Peaks option to effectively suppress the spurious peaks in the noise-shaped dither.
If, for some reason, any dithering noise is undesirable, simple truncation remains the only choice. Truncation results in harmonic quantization distortion that adds overtones to the signal and distorts the timbre. In this case you can enable Suppress Harmonics option to slightly alter the truncation rules, moving the harmonic quantization distortion away from overtones of audible frequencies. This option doesn’t create any random dithering noise floor. Instead it works more like truncation, but with better tonal quality in the resulting signal. This option is applicable only in the modes without dithering noise and without aggressive noise shaping.
I hope this gives you an idea of what it takes to enjoy your 24bit audio in a 16 bit format.
Warning: Nurd alert! This can cause your eyes to cross . . .
The photo analogy works pretty well, but you can do the comparison in digital as well.
As I understand it:
Think of digital music as a camera that is taking pictures of the sound:
The 44,100 or 96000 is the number of times the computer “takes a picture” of the sound being produced - the number of times the sound is “sampled” in a second. So, if you have a very complicated sound, an orchestra for example, you can raise the sample rate (the number of pictures" taken) to get more detail - a movie at 16 frames per second or at 35 frames per second, the slower frames per second may be a bit “blurry” , but it doesn’t matter so much if the subject is not moving very fast. Neither may truly be fast enough to capture everything that is happening, but the human mind fills in the spaces.
The bit depth is the “resolution” of each sample - the amount of information that can be stored in each “picture.” More information is stored in each picture. This means that the recording can handle a wider range of information in each picture (more head-room is the analog term), the chances of “overloading the sample” is reduced. So, recording at 24 bit can avoid clipping because it has a wide dynamic range before there is more information than can fit on the picture. *1
The question of how much all this matters is a long standing debate. The advantage of 24 bit is pretty easy to see, it makes recording easier and when you dither down to 16 bit you are throwing away sounds that are not in the hearing range of humans. So, does it matter if you record in 24 bit - well, it does have the advantage of avoiding clips and some people believe they can hear the difference - I can’t hear the difference, but I record in 24 bit sometimes to avoid clips. Recording at 96k, well, it make huge files and again I can’t hear the difference, but I mostly record singers and guitar players.
As as been mentioned, CD quality is 44,100 at 16k. That is pretty good for most everything I have ever recorded. If I was going to use DVD that will take much higher rates, and that may be the future.
*1 Digital sound is made of 0s and 1s. Think of the sum as the amount of energy that is being created. Comparing to an electrical circuit: (I have no idea the actual comparisons) think of 16 bit as a 15 amp circuit, 24 bit as a 60 amp circuit. If too much information (sound/ Energy) is sent to the digital math the sum comes to more than the computer is set up to handle, and we get a “clip” - we have “overloaded the circuit” and an unpleasant noise is produced.
I want brain-to-brain Bluetooth.
i want a brain!
The short answer is - unless or until you find that it is not sounding the way you want - 16 bit 44,100k will do the job.
All of the other stuff we are talking about can improve a mix OR really screw it up.
So, to get good everyday recordings standard 16 by 44,100 work great! In general, the less you do to a mix past re-balancing the volumes and panning is nuance or more to the mastering side of recording.
“Bang bang Baxwell’s sivler hammer…”
I would have to agree with bax… everything I have read so far about converting from one bit-depth to another involves some form of dithering (as Paco mentioned) and therefore its possible for the integrity of the original track to be degraded by the conversion process.
Did audio techs and home recording enthusiasts have to worry about all this back in the days of analog???
Aye, Danny. Stick on a bit of Leadbelly or Ottlie Patterson, Motown…
Still sounds magical:-)
I think we worry about all this technical stuff because now we have the capacity to mess with it. You can do some great stuff with digital recording and I love it -HOWEVER,
With analog, you went for the performance, that was the first measure of the recording and should still be! Think about a recording session in about 1930: everyone is in one room and the recording is literally cut into the record as the performance is done. You get it right or you do it again.
I was a performer in the late 50’s and 60’s. We went to studios and paid good money to record on mics that were not as “good” as today’s, with 2 tracks if we were lucky - some of that stuff sounds great! Not pristine like a digital, but the performance is there and if it wasn’t there, then little or nothing we could do about it, except record it again.
All that said - I still love what I can do with NTrack to make a “professional” sounding recording.
i want a brain!
You can have mine, I'm getting a new one next week that can,
A: Drink without getting a hangover.
C: Know how to change the bobbin on the weed whacker.
D: Has interchangeable parts that can be upgraded.
I would have to agree with bax... everything I have read so far about converting from one bit-depth to another involves some form of dithering (as Paco mentioned) and therefore its possible for the integrity of the original track to be degraded by the conversion process.
Only when going from higher to lower.