What about sound on the Parallella ?

[theover]

My board (7010 + connectors + accessories) isn't there yet, unfortunately, but I was wondering what the easiest way will be to have sound on the ARM Linux, for instance.

A major application of mine would involve massive sound processing, and it would be interesting to simply start with running jack and Ladspa on the ARM cores with some DA convertor supported by alsa in the kernel.

I saw some videos last week, like the pretty cool Ray Tracing example, that had music in them, how was this done ?

Of course it would be a very good idea to connect a quality 16 (20 ?) bits, 2MHz DA converter to the FPGA (with a connector convertor PCB?), program a very long sinc based resampling filter in the Adapteva cores, and start from there, but that's alot pf work to do...

T.V.

[shodruk]

There is a built-in HDMI digital audio output.
The sampling rate is 48kHz by default, the chip supports up to 192kHz audio.

http://www.analog.com/static/imported-f ... DV7513.pdf

You can use an external high-quality DA converter through HDMI.
Also, you can design your original professional quality audio AD/DA daughter card.

[timpart]

Professional audio equipment generally doesn't go above 192 kHz at 24 bits. Human hearing doesn't go much over 20 kHz so a 48 kHz sampling rate can do that frequency without aliasing. (CDs are 44.1 KHz at 16 bit.) So unless you want to talk to dolphins I'm not sure why you need to go over 192 kHz. The human ear will have a hard time noticing any difference at all. Also most recording microphones have a poor response over 20 kHz since people won't hear frequencies higher than that.

Higher sampling rates add more data points, but after they go over twice the highest frequency they don't really capture more information. You can still recreate the original waveform in terms of a series of sine waves of various frequencies. (Or if you prefer wavelets.)

Tim

[theover]

Your undergrad EE may have taught you about reconstruction filtering (with sinc functions, (!= wavelets)).

All DA convertors (I use various ones, pro too, and 192 on daily basis) use a reconstruction filter (and anti-aliasing, etc.), but of a very short interpolation kind.

At some point I'd like to get rid of the always present reconstruction errors, by making sure I use a very long reconstruction filter, with accompanying fast and accurate DA convertor.

Don't bother if my the pro level audio quests don't interest you, it's not my intention to start that discussion here !

T.

[hamster]

[theover]

I want to try 3 seconds (!) of reconstruction. because my math tells me that's where the sin(x)/x becomes diminished enough (compared with 20 bits or so of vertical resolution: 1/x with x=3*44100 ~~> -100dB) in the face of many high frequency components.

But aside from that: interesting I'll check links out, do you have more info ? Is what you did Open Source, or commercial ?

T.

[hochopeper]

I too have been considering high speed i2s 384k + out of my parallella, haven't started any work on it yet but will post back if/when I get something done. Some modern audio DACs will take 768kHz or more input if filtering is done externally.

I also had daydreams about 5MHz 20bit SAR ADC for measurements that might go on the same hardware design just for my own flexibility, but that's another story.

[over9000]

[hochopeper]