Topic: Dithering and mixdown to 44.1 from 48kHz - why not just record in 44.1?

[acidjack]

I am assuming I already know the answer to this, but I'd like to hear some scientific explanation if possible. 

A lot of people record in 24/48 or 24/96 and then seem to dither to 44.1 when they release to the public.  Obviously to do that, your software has to engage in dithering, which causes at least some theoretical quality loss from both the process and, of course, the loss of bits.

It seems to me, especially if you aren't one of the people who mixes down to multiple formats, aren't you better off just recording in 44.1 to start with?  Or do you really plan to go back one day and re-mix it down to get back those extra 3.9bits? 

Has anyone ever tried to do some kind of a controlled test to see whether a 48kHz recording sounded audibly different than a dithered 44.1? 

[matahaka]

I believe.. you can refer to this thread..

http://taperssection.com/index.php?topic=125051.msg1662541#msg1662541

[Shadow_7]

I tend to record at or above 24/96 because I generally have two outputs.  Audio only CD(44.1kHz) and Video with audio (48kHz).  Having the higher rate allows better odds of getting a good end result in BOTH.

I'm not sure on the science of it, or even the technicals(yet), but the kHz is the resolution(samples per second).  You can't take an SD video and make a great HD video from it.  With extra bits in audio, you're at worst throwing out half the samples.  At best perhaps time shifting some samples that byte or two to have a more detailed result at the lower kHz.  Which if captured digitally at the lower kHz would only give you one or the other or let them do battle without any hope of arbitration.

I mainly do the higher kHz, so I can EQ, and other harsh edits to the source BEFORE down sampling.  Which, if only in my mind, yields superior results.  In theory mics lose some of the high end over time.  And other less than perfect mics to start with.  So IMO, having the higher starting point allows you to do more processing in post without the ill effects.  Or at least less of them. 

If I had to choose between just 44.1kHz and 48kHz, and only those two, it would probably depend on the project at hand.  Whatever end result paid the bills would be the one to win.  Or depending on the content being recorded.  No qualms from me about recording lectures and other non-musical things at 44.1kHz for video.  Even with DSD capability at my fingertips.

[page]

I'm not sure on the science of it, or even the technicals(yet), but the kHz is the resolution(samples per second).  You can't take an SD video and make a great HD video from it.  With extra bits in audio, you're at worst throwing out half the samples.  At best perhaps time shifting some samples that byte or two to have a more detailed result at the lower kHz.  Which if captured digitally at the lower kHz would only give you one or the other or let them do battle without any hope of arbitration.

I think you might be confusing video sampling/display and nyquist sampling. In video, you have X amount of visual space, and HD video gets crammed into that space where SD video gets spread out so more details renders a sharper picture when viewed on the same tv. In nyquist, you get 2 samples for each frequency per second regardless of whether you sample at 44, 48, or 96. You still only get 2 samples at 600hz. What you get with 96 vs 48 is the ability to (presuming your speakers can render that information) torment your dog with information in the 30khz band where as thats gone at 48 sampling. What's particularly amusing or infuriating (depending on your objective is if you have equipment or an environment that has a hum/buzz in those upper echelons (I once had a mic that did regardless of what where I was recording).

To me, the last technical discussion of personal interest is whether your initial ADC chipset is better at the end result (e.g. 44.1 or 48) than your software resampling algorithm that you're using. As the folks at Sound Devices once said: "recording at 192khz is great for sound effects generation, not effective for recording a PA..." If your algorithm introduces noise elsewhere in the audio band, than it's sort of self-defeating to sample higher then that for a non-time-shifted function (e.g. slowing down information for special effects creation), and I find it somewhat surprising how many algorithms do. Someone did a bunch of graphs on the subject and you get a visual representation of low level noise "smear" depending on what you chose to compare. Neat stuff.

[notlance]

A lot of people record in 24/48 or 24/96 and then seem to dither to 44.1 when they release to the public.

Just to keep this discussion a little less confusing, let's define the words.

A simple but by no means complete definition of dither is the process of adding noise to a signal before A/D conversion in order to reduce distortion due to quantization error.

Converting a signal sampled at 48 kHz to a 44.1 kHz signal is called resampling or sample rate conversion, and that is the process you are describing.

[notlance]

Has anyone ever tried to do some kind of a controlled test to see whether a 48kHz recording sounded audibly different than a dithered 44.1?

Here is one controlled test:

http://drewdaniels.com/audible.pdf

[Lil Kim Jong-Il]

If you plan to circulate your recordings on CD or DVDA, then you you can record at 44.1 and put the tracks directly to disk.  If you want to use your recording as an LPCM track on a video DVD, the spec allows only a 48k or 96k sample rate.  There was a time, and it may or may not still be the case, that a lot of consumer DVD players don't handle DVDA discs.  So when I put my 24-bit stuff to optical media to play at someones house, I author as a video DVD with 24-bit LPCM track.  Sampling rates other than 48K and 96K won't work for that and I don't see the point of converting from 44.1k (or 88k) to 48k.  There are no issues on any file based players with 48k so it's only for CDDA disks that 44.1k is of any interest. 

[Shadow_7]

A lot of people record in 24/48 or 24/96 and then seem to dither to 44.1 when they release to the public.

Just to keep this discussion a little less confusing, let's define the words.

A simple but by no means complete definition of dither is the process of adding noise to a signal before A/D conversion in order to reduce distortion due to quantization error.

Converting a signal sampled at 48 kHz to a 44.1 kHz signal is called resampling or sample rate conversion, and that is the process you are describing.

But with a lot of software you get to chose the type of dither when you resample.  So it's not as simple as just saying dither or resample, you can't have one without the other.  Whether it's 96kHz to 44.1kHz or 48kHz to 44.1kHz OR just adding EQ.

And then there are those frequencies beyond the sampling resolution that will get sampled at half their frequency, if you don't use a higher sampling rate.  i.e. 30kHz (beyond human hearing) recorded at 15kHz (within the range of human hearing).  And some mics are very sensitive well into and past that frequency.  And many things will produce those frequencies.  Maybe not your speakers, but they'll try.  At the higher sampling rates, you have a better chance of filtering those off.  i.e. less overall noise.

[goodcooker]

I record at 44.1 kHz sampling rate since I can't hear a difference at higher rates and scientifically the sample rate divided by two is the top end of the frequency range of the recording. Half of 44.1 is 22.5 kHz and I can't hear over 17.5 kHz so I don't see the point.

I think artifacts introduced by the resampling process (depending on your software/process, I use Wavelab) are more likely to do damage to your recording than just recording at the lower sampling rate.

[page]

But with a lot of software you get to chose the type of dither when you resample.  So it's not as simple as just saying dither or resample, you can't have one without the other.  Whether it's 96kHz to 44.1kHz or 48kHz to 44.1kHz OR just adding EQ.

Sure you can, I can dither something and not resample it, likewise I can resample something and truncate it without dither. Second, I can think of a couple of software packages, both free and expensive, that have the two segregated so you control when they occur independantly.

And then there are those frequencies beyond the sampling resolution that will get sampled at half their frequency, if you don't use a higher sampling rate.  i.e. 30kHz (beyond human hearing) recorded at 15kHz (within the range of human hearing).  And some mics are very sensitive well into and past that frequency.  And many things will produce those frequencies.  Maybe not your speakers, but they'll try.  At the higher sampling rates, you have a better chance of filtering those off.  i.e. less overall noise.

I'm assuming you are referring to harmonics, because I can't think of anything else at 30khz that I'd want to reproduce verbatim in a music recording without altering the speed, and whatever is coming across to be sampled at 15khz is still there regardless of whether I chose 48k or 96k. Actually, I can't think of anything I'd want to reproduce verbatim at 30khz period.

1) Just because there is information there, and you're speakers try to resolve it and fail, doesn't make it positive. Gallant attempting and failing is still failure.
2) As DSatz, my audiologist, and others have pointed out; humans exposed to concert environments can't really hear much past around 17khz anyway. So recording anything up there is sort of moot for our purposes. Which brings me to #3:
3) Whether you filter something when you have 48khz of information or 22khz of information at the time of recording is moot if at the end of the day your target audience can't hear it, and you're chucking it anyway.

So I'm with cooker et al; I think you're more likely to do damage through the adding of artifacts and distortion to your recording via resampling if you have a bad algorithm rather than recording at a lower sampling rate.

Has anyone ever tried to do some kind of a controlled test to see whether a 48kHz recording sounded audibly different than a dithered 44.1?

Here is one controlled test:

http://drewdaniels.com/audible.pdf

neat read.

[Lil Kim Jong-Il]

But with a lot of software you get to chose the type of dither when you resample.  So it's not as simple as just saying dither or resample, you can't have one without the other.  Whether it's 96kHz to 44.1kHz or 48kHz to 44.1kHz OR just adding EQ.

EQ is also independent of dither.  Digital EQ may add some noise as a result of it's calculations but that is not the same as noise added by the dithering algorithm. 

[F.O.Bean]

I have read all there is to read about this situation, and I record at 24-Bit/44.1kHz!!! It makes putting it to disc easier and is still leagues better than recording in 16-Bit

[stevetoney]

Speaking for myself, it's been at least two years since I've used CDRs, so I stick with 24/48. 

FWIW, I just read somewhere that Sony is producing 20 million fewer CDRs daily than they were at their peak.

[ghellquist]

This is a matter that has been discussed and discussed on just about every forum talking about recording.

When thinking back, I believe most discussions ended up with the following arguments that sort of tend to boil up as concensus.

1) different equipment handles different max sample rates
2) some equipment does handle some of the possible sample rates rather badly ( one example might some of the Sound Blaster generations that only handled 48kHz natively, everything else was software resampled behind the scenes ). Even rather high-end equipment has been shown to support some of the sample rates less well the others to choose between.
4) higher sample rates means larger files and higher load on your computer ( which might be moot if you computer is fast enough and your hard drives large enough )
5) different end media may set restriction on what sample rates are allowed for your end result
6) best bet is for you to try your own equipment in your own application, from recording to end result at different sample rates. You might preferr one of them.

Nothing else seems to be a common consensus. Some like a high sample rate and resampling, others hate resampling, some say that resampling has to be at even divisors, others show scientifically that it is not so. And then there are two camps of people where one camp says that hearing ends at about 20kHz and the other camp says recordings has to go well above that to capture the full story and that we can hear that.  Add to that a lot of peole that, frankly speaking, really does not seem to have a clue either way, but still airs their opinions.

Now, for my own, rather humbling recording needs ( acoustic classical music ) I tend to record at 44.1kHz.
I did follow my own advice 6 above, and did some tests.

// Gunnar

[twatts (pants are so over-rated...)]

I record and listen with the best quality I can get.

I do the 16/44 thing for the Masses, so they can burn their CDRs. 

Terry

[Belexes]

I do the 16/44 thing for the Masses, so they can burn their CDRs. 

With the CD format going away, I am thinking of not dithering anymore when I share to the masses.  Storage is cheap.

[darby]

I do the 16/44 thing for the Masses, so they can burn their CDRs. 

With the CD format going away, I am thinking of not dithering anymore when I share to the masses.  Storage is cheap.

I haven't burned CDs for myself in years
as far as the masses... unfortunately it seems most folks would rather have MP3s than CDs 
I still do 16/44s for some people, but most get the FLACs off my 24/48s

[twatts (pants are so over-rated...)]

I do the 16/44 thing for the Masses, so they can burn their CDRs. 

With the CD format going away, I am thinking of not dithering anymore when I share to the masses.  Storage is cheap.

I would tend to agree, since I haven't burned a real CD in over a year (occasional do one for my wife)...  But I think 16/44 is a good standard that most everyone is familiar with...  And I don't want to mess with the system, since that's what everyone is used to...  If someone wants my 16/48 or higher files, they can ask me...  although they never have before...

Terry

[page]

I do the 16/44 thing for the Masses, so they can burn their CDRs. 

With the CD format going away, I am thinking of not dithering anymore when I share to the masses.  Storage is cheap.

I would tend to agree, since I haven't burned a real CD in over a year (occasional do one for my wife)...  But I think 16/44 is a good standard that most everyone is familiar with...  And I don't want to mess with the system, since that's what everyone is used to...  If someone wants my 16/48 or higher files, they can ask me...  although they never have before...

Terry

and while storage is cheap, it's still a wench to transfer 1.6gb across the wire just to listen once or twice. Not much makes permanent rotation in my book, and the few things that do, I'd go after a better copy if I thought it would improve the experience.

[Lil Kim Jong-Il]

as far as the masses... unfortunately it seems most folks would rather have MP3s than CDs 

I have a long history of being a snob about MP3s but since I take music in the car using the iPhone I have completely changed my attitude.  Over the FM transmitter it doesn't matter.

[rjp]

I have a long history of being a snob about MP3s but since I take music in the car using the iPhone I have completely changed my attitude.  Over the FM transmitter it doesn't matter.

My recently-purchased car came with a stereo that will play MP3s burned to a data disc. Pluses: lots of music can fit on a disc (I usually use 192K CBR for car applications), and it has no problem with 48 kHz material. Minuses: they didn't bother implementing gapless playback, and the display won't scroll for titles/artists that are too long to fit.

[scb]

there are 2 different things going on in this thread. as "notlance" said, going from a higher sampling rate to 44.1 is resampling, while going from a higher bit rate usually involves adding dither (without dither it's just called truncation).

if you want to record in 44.1, go ahead.  But if your recorder can record at 24 bits, I recommend that.  It gives you much higher headroom.  You can run the levels lower and still be using more than 16 bits. If you record at 16 bit, you really need to be cranking your levels to get close to 16 bit resolution. 

[Lil Kim Jong-Il]

(without dither it's just called truncation).

If I may be pedantic, it's linear scaling.  Truncation of the samples would result in a discontinuous waveform.

[scb]

smartass!

[SmokinJoe]

I always think of resampling like this.... you are back in third grade with graph paper, a pencil, and a ruler.  The teacher gives you a bunch of XY coordinates, and then asks you to interpolate other values in between by following the lines in between.  You can do it, but it's not an exact science.  This certainly doesn't improve accuracy, that's for sure.  So don't do it unless there is a good reason to.

While I'm sure some people will cringe at this next statement... I don't figure it's any more of a "sin" to convert 44.1 -> 48 than it is to convert 48 -> 44.1 (assuming there really isn't any data in there above 20k, and I don't think there is on my recordings) because all I'm doing to converting format for compatibility purposes either way.


(this picture serves no purpose other than to visualize how ridiculous it would be to plot points on graph paper and interpolate)

[page]

I don't figure it's any more of a "sin" to convert 44.1 -> 48 than it is to convert 48 -> 44.1 (assuming there really isn't any data in there above 20k, and I don't think there is on my recordings) because all I'm doing to converting format for compatibility purposes either way.

I agree, especially if I'm happy with 44.1 to start with.

[Lil Kim Jong-Il]

I don't figure it's any more of a "sin" to convert 44.1 -> 48 than it is to convert 48 -> 44.1

It's not.  But if the target is a 24-bit DVDV disk, resampling from 44.1 adds an additional step to the authoring process.

[tapeheadtoo]

If you plan to circulate your recordings on CD or DVDA, then you you can record at 44.1 and put the tracks directly to disk.  If you want to use your recording as an LPCM track on a video DVD, the spec allows only a 48k or 96k sample rate. 

I'm a little confused about this.  I typically record audio at 24/44.1.   I am able to sync the audio file perfectly to my 16-bit video for a video DVD.  The final audio format is LPCM, 16bit, 48khz.  It seems that my master audio being at 44.1 has no detrimental effect on the final output (at least none that I can see or hear).  Or does it? 

[page]

If you plan to circulate your recordings on CD or DVDA, then you you can record at 44.1 and put the tracks directly to disk.  If you want to use your recording as an LPCM track on a video DVD, the spec allows only a 48k or 96k sample rate. 

I'm a little confused about this.  I typically record audio at 24/44.1.   I am able to sync the audio file perfectly to my 16-bit video for a video DVD.  The final audio format is LPCM, 16bit, 48khz.  It seems that my master audio being at 44.1 has no detrimental effect on the final output (at least none that I can see or hear).  Or does it?

his point is that if your end product is going to be in the 48k realm, then recording to 48k accomplishes two things;

1) Saves you work/a step/processing time delay of resampling. Your software may do this for you or you may have an uber fast PC in which case it probably doens't matter.
2) Saves information in the 23khz area that would be discarded had you gone to 44.1 anywhere in the process.

Whether you can hear that information in the bands between 22khz and 24khz is dependant upon each person. If you can't hear a difference (I can't, I've tried), and nobody has said anything, then I wouldn't worry about it.

[taperj]

If I may be pedantic, it's linear scaling.  Truncation of the samples would result in a discontinuous waveform.

Could you elaborate on this? I guess I'm confused as to why does the Sound Devices manual, wikipedia, and all other documentation I can find say it's truncation then? Just wanting clarification since most trusted sources say otherwise.

[morst]

Could you elaborate on this? I guess I'm confused as to why does the Sound Devices manual, wikipedia, and all other documentation I can find say it's truncation then? Just wanting clarification since most trusted sources say otherwise.

I guess they're just not being fully pedantic!

[Lil Kim Jong-Il]

If I may be pedantic, it's linear scaling.  Truncation of the samples would result in a discontinuous waveform.

Could you elaborate on this? I guess I'm confused as to why does the Sound Devices manual, wikipedia, and all other documentation I can find say it's truncation then? Just wanting clarification since most trusted sources say otherwise.

As was pointed out before, I was being a smartass.

Truncation lops off precision.  If you have have a 24-bit ADC and save the output as a 16-bit value by eliminating the lower 8 bits you lose precision but you still get a suitable approximation of the input waveform because all of the samples are shifted by the same proportion to fit within the range of values permitted by the size of the sample word.  Lopping off the lower 8 bits scales the sample's representative value by 1/256.  It's still linear scaling, it's just sort of brute force way of doing it that results in the peak value of the result having the same proportion to the maximum possible value of it's range as the original had to it's range.

In post-production, the conversion 24-bit to 16-bit for CD usually involves normalization prior to truncation so that the final values make almost complete use of the range of values available to represent a sample. Normalization linearly scales the sample value (usually up) and then the truncation scales the value down.  The result is linearly scaled by a factor that is not simply a power of 2 as it would be with byte truncation only.

The ADC in the field simply truncates because it can't anticipate future input values, something that is required to calculate a normalizing factor.  Without bothering to look at the SD manual, I'm pretty sure that is what SD is referring to.  You can choose to simply truncate your 24-bit recording to get 16-bit samples in post.  If your recording has peaks well below the maximum allowed sample value, then you end up with a result that is a less desirable approximation of the input than if normalized first.  Both are linear scaling so I was just being a smart ass.

[taperj]

Awesome, thanks for the detailed explanation Lil' Kim Jong-Il, I was just curious. Always trying to learn everything I can about the in's and out's of audio. Much appreciated. 

J

[ghellquist]

I have found that there is a much better way to think of sampled signals than how it is generally presented.

The thinking starts with defining that the signal goes between -1 and +1. Not quite up to one  but very close. The signal swings between positive and negative values, up and down with the sound making the music. Not quite reaching 1.0 or -1.0 as that would be where the AD starts clipping.

One sample could then be represented as, exampel,  the value 0.123 . In another system the same value could be represented as 0.1234334, clearly a much more accurate description of the value.

So if we simply chop off the last few decimals we can go from 0.1234334 to 0.123, losing precision in describing the sample value. This is how truncation works on the sample level.

But, our ears do not listen to samples, our ears hears the resulting sound vawes. And then things change a bit. The ear never hears one sample by itself, it can only hear the sound created by several samples, maybe even hundreds.

Our ears hears the truncation as two different things.
One effect our ears hears is an increase in the background noise level. This can be expressed as a diminishing of the SN-ratio ( read as Signal to Noise ) .
One way of looking at it, is that below the third decimal ( below 0.001 ) there is only noise.
Another effect our ears hears is the addition of some very low level sound artifacts. Mostly these are masked by all the other sound but sometimes they can be heard.

Notice that this truncation does not involve any scaling or other stuff, it is simply removing decimals from the representation of each sample.

What we want to do with sound samples instead of simply truncation ( = cutting off ) the decimals is to do a process called dithering. The idea is to add a small, specially formed, noise signal before cutting of the decimals in the representation. This special noise signal is designed to make the sound better for the ear. There are several different kinds of dithering signals, suffice to say that they all work about the same way but sound slightly different.

The effect of the dithering is to make the sound more pleasing to the ear, in two different ways.
First the noise floor the ear perceives is moved down a bit. And the ear can actually start hearing things a small way down into the noise floor. 
The second effect is that some of the artifacts we hear when truncating are not heard anymore.

// Gunnar

[ghellquist]

This is a complicated thing to get full grips on. I know from personal experience ( from a slightly different but related field)  that anything but a fully formal mathematical treatment of sampling will distort aspects of the full process. Not that I know how to it, I have only read papers written by people actually knowing the stuff and half-understood some of it.

I belive however that we should avoid some of the holes in trying to visualize things on a "popular science level".

If you have have a 24-bit ADC and save the output as a 16-bit value by eliminating the lower 8 bits you lose precision but you still get a suitable approximation of the input waveform because all of the samples are shifted by the same proportion to fit within the range of values permitted by the size of the sample word.  Lopping off the lower 8 bits scales the sample's representative value by 1/256.  It's still linear scaling, it's just sort of brute force way of doing it that results in the peak value of the result having the same proportion to the maximum possible value of it's range as the original had to it's range.

There is no need to introduce any aspect of scaling in this popular science level of describing it. It is much better to work with the assumption that a 16 or 24 represention of a sound sample has a value range between -1 and +1. One sample could then have a value of 0.123 in a sixteen bit reprentation and 0.12343342 when represented by more bits. The number of bits then simply describe the degree of details, the precision, each sample can be described with.

This representation is actually how it is programmed inside just about any modern sound handling program. Samples are inside the program handled as "floating point".

// Gunnar

[Lil Kim Jong-Il]

Gunnar, my labeling of the 24-bit to 16-bit conversion by truncation as linear scaling is really very straight forward.  Nothing about it is complicated nor a "popular science" view.  You suggest additional concerns that were not part of my post and those do result in a different situation which is why I excluded them.  The thread has wandered off topic so I'll send my lengthy and good natured follow-up by PM.  If anyone else is interested, let me know and I'll copy you. 

[ghellquist]

Hi.
The forum software cannot return any answer to the adress Lil' Kim Jong-Il.

I agree however that the discussion has gone off topic. Suffice to say that we may both be correct, but not agreeing on how to describe the same thing. Not unheard of in the world.

// Gunnar