LinuxMusicians

Time for me to join in.

I am in the mixing phase of what I hope will become an album, entirely recorded in Ardour 2 on Ubuntu Studio + KXStudio. This is a prog rock type album, of music my buddys and I wrote back in 1986-87, when we were about 18-20 years old. But we have only just now recorded it properly, using the amazing open source tools available with Linux.
Here is the "fun song" of the album. A sort of mixture between Jethro Tull and Focus-being-silly:

http://www.etcetera-music.eu/public/Etc ... _Camel.mp3

The song was recorded and mixed entirely in 48kHz, 32-bit float - which is what I always use - and then mastered down to CD-quality 44.1kHz, 16-bit, using "Jamin". The file in this links was encoded to mp3 from the test mix CD.

/Frank

Sounds good.

I'm interested in why you record at 48kHz if the final version will be 44.1.

Anything higher than 44.1 is an improvement. The highest frequencies that may be represented at 44.1kHz is 22.05kHz. I have personally measured that I can (or at least could some years ago) hear up to 20kHz, which is too close for comfort. Once you start mixing, the high frequency contents of several sources are mixed, and the interference between several high frequencies will produce lower frequencies in the audible range, even though the sources wont be audible. And as the highest frequency 22.05kHz itself will have the same resolution as a square wave - that is high-low-high-low-high-low... - regardless the wave shape of the original source, then I am certain that higher frequency content will produce audible lower frequency artifacts when sound sources are mixed.
With bad resolution in the high frequency area, the audible side bands are bound to be distorted, compared to the interference of natural sounds. But higher sample frequencies improve this as it captures more of the original wave shape of the higher frequencies, and hence also produces more correct interference.
48kHz is a (slight) step forward, and 96kHz, or even 192kHz would be even better.

I find it easier to accept one time dithering in the transfer from 48kHz to 44.1kHz. The fact that you need dithering at all is really proof that the presence of two higher frequencies, in this case 48kHz and 44.1kHz, can produce audible problems.

Now there's a long explanation! A more down to earth explanation is, that 48kHz just appears a bit more natural to my ears.

/Frank

Very cool piece. I enjoyed it a lot.

Frank Carvalho wrote:... A sort of mixture between Jethro Tull and Focus-being-silly...

And a nicely-done mixture it is ! Well played, my compliments to the band and the recordist.

Best,

dp

Nice piece, takes me back to the early 70's.

Those obsessed with high sample rates should read this: http://people.xiph.org/~xiphmont/demo/neil-young.html
It is not true that the highest frequency digitized is output as a square wave - the DAC filters convert the sound to sine waves (in fact to reproduce the sharp edges of a square wave would require a lot more high frequencies).
However, there was a good argument made for using 48kHz in a recent linuxaudio email thread (http://lists.linuxaudio.org/pipermail/l ... 89224.html ) which is that a lot of audio gear uses 48kHz internally and is better optimized for that.

Excellent stuff Frank & band!

I presume you play guitar and/or keyboards on this do you Frank? There's some superb playing by all members whatever. Reminds me of early 70's Crimson and Secret Chiefs 3 - two of my fave bands so I look forward to the finished album which you could (misleadingly) say has been almost 30 years in the making!

Is it a largely instrumental album? When can we expect it?

Thanks Frank (is the messenger of your doom and your destruction!;)

Thanks.

The album will be 50-50 instrumental/song-based. First vocals wont appear until some 16 minutes into the album, but there are several "songs" after that, although all of them have long instrumental parts. And yes, this is going to be really 70'ies kind of music. I play guitars and keyboards, and bass guitar too on half the songs. It is definitely an album that was almost 30 years on its way, but then we made two real albums in between.

I hope I can wrap up the mix soon, and then I will see if I can arrange for distribution. With some luck I can finish it this spring. But now I need a good cover. Hmmm. That requires graphical tools. Let's have a look at what Linux can offer here...

I love nerdy debates on the nature of sound:

varpa wrote:Those obsessed with high sample rates should read this: http://people.xiph.org/~xiphmont/demo/neil-young.html
(Which went on to say:)

article wrote: Neither audio transducers nor power amplifiers are free of distortion
.....
Inaudible ultrasonics contribute to intermodulation distortion in the audible range (light blue area). Systems not designed to reproduce ultrasonics typically have much higher levels of distortion above 20kHz, further contributing to intermodulation. Widening a design's frequency range to account for ultrasonics requires compromises that decrease noise and distortion performance within the audible spectrum. Either way, unneccessary reproduction of ultrasonic content diminishes performance.

Insteresting article. First of all, I notice that this actually confirms what I say, namely that ultrasonic interference produces audible side bands. But the context is also different, as it argues that with the resulting distortion in the DAC this is undesirable. I guess that is true for DAC, and probably a good reason not to use 192kHz directly, unless the DAC is designed to do so.

But I am more concerned with the interference happening in the mixing of several sources to a single mix file, which means the interference happening as a result of the floating point operations in the CPU. At that stage there is no distortion from any DAC involved, and as the floating point operations will be much more accurate, the resulting calculated audible sidebands will also be more accurate, and not distorted, and then I think they become desirable. The area where this is most important is in the highs of cymbals. Real cymbals definitely produce ultrasonics too, and I believe that the liveliness of the cymbals will only survive a mix if the ultrasonics are also captured and maintained in the equation during mixdown. Once you have what you want in the audible range in a mix file, you can destroy anything above audible during resampling.

varpa wrote:It is not true that the highest frequency digitized is output as a square wave

Oh no, that's not what I meant either. I just meant that two different real life sources, a 22.05kHz sine and a 22.05kHz square, would be sampled to the same representation. The low sample frequency does not allow any distinction. But in real life they would not behave the same. The interference between two ultrasonic sines will be an audible sine, while the interference between two ultrasonic squares, will be an audible hard sync-like sound.

/Frank

Frank Carvalho wrote:

varpa wrote:It is not true that the highest frequency digitized is output as a square wave

Oh no, that's not what I meant either. I just meant that two different real life sources, a 22.05kHz sine and a 22.05kHz square, would be sampled to the same representation. The low sample frequency does not allow any distinction. But in real life they would not behave the same. The interference between two ultrasonic sines will be an audible sine, while the interference between two ultrasonic squares, will be an audible hard sync-like sound.

Never mind. Today's converters do some oversampling already, so there are no artifacts and no interferences. It sounds strange but there is no need to use more than 44.1 kHz today ... at least for a human audience