Ben Stanley wrote:
So, could we then change the number of channels available depending upon which card is detected? It doesn't seem to make sense to make things available to the user via the driver that they can't plug something into on their particular model card. I guess that to do this I would have to know how many digital output channels are supported by each card. I downloaded the user manuals for a few of the cards; I only got the impression that each card supported one digital output. Do you know which card you tested for multiple outputs?
Unfortunately the datasheets do not give me card specific details. Only how to program the ca0106 chip. The GPIO etc. for each card is determined by trial and error.
Now, the Creative Labs user documentation gives me the impression that I can use the analog output channels at the same time as the digital output channel. That also doesn't appear to agree with how the driver is written. I am currently under the impression that the driver is written to allow 4 stereo analog channels or 4 digital channels. At least that is how things look when configuring things using alsamixer. Perhaps I am confused and should go and look at the GPIO that you mentioned above.
This particular sound card can output to both analog and digital channels. The sound is simply duplicated to both. I disabled it happening at the same time mainly due to what happens when AC3 or DTS is output to the digital. One would not want that send to the analog outputs!
Details:
The card I have is characterised by the following: http://www.soundblaster.com/products/product.asp?category=1&subcategory=... Model: SB0570 serial: 100a1102 Also known as: Sound Blaster Audigy SE
During initial testing, I noticed that 44.1kHz playback was not implemented. Subsequent testing shows that speaker-test works fine with 48kHz, 96kHz and 192kHz. (I only tested 16bit output so far.)
I noticed in the source code that 44.1k was explicitly disabled. I added code in snd_ca0106_pcm_prepare_playback to set up this rate for S/PDIF output as per the comments in ca0106.h . Initial tests using hw:0,0 at 44.1kHz produced recognisable signals with some noisy corruption. Later I accidentally discovered that serially opening hw:0,2 , hw:0,1 and hw:0,0 at 44.1kHz then produces perfectly good 44.1kHz sampled digital audio output. Removing hw:0,1 or hw:0,2 from this sequence causes noisy corruption. It seems that channels 0-2 in reg40 must all be set to the same sampling frequency for S/PDIF to work where 44.1kHz is concerned. Conversely, to sucessfully output 48kHz again, I have to open hw:0,2 , hw:0,1 and hw:0,0 at 48kHz to restore proper output. I do not have such troubles with 96kHz and 192kHz, for which it suffices to just open hw:0,0 at the relevant sampling rate.
The ca0106 can do 44.1kHz for digital output ONLY. The ca0106 cannot output 44.1kHz to the DACs so it will only work in Digital mode. It is a hardware restriction. You are correct, all the inputs and outputs have to be at the same rate.
So far I haven't considered 'inputs', although I do now have the digital I/O module so that I can get digital signals into the card. I suspect I can only generate 44.1kHz and 48kHz sampled input signals to test with.
So for my card where only one output channel appears to be available (can you tell from the doc if this is true?), then I would just slave the other hardware channels to hw:0,0 settings. However, on other cards where the other channels are in fact available, how do you enforce the restriction that all the channels must have the same sampling frequency within the ALSA model?
I'm just trying to plan how I would fix this properly within the driver.
You could set the driver to fix to the rate of the first opened device, and only when all devices have been closed could the rate change again. This is unfortunately not at all intuitive for the user. I would therefore add a global mixer control to set 44.1 or 48kHz. The user could then decide which to use. 44.1 would mean 44.1, 88.2 etc. 48 would mean 48, 96, 192.
By the way, the spdif output on the ca0106 works well at 96kHz also.
Anyway, this is speculation. I'd love to have the docs. I haven't tried to get them. Is it likely/unlikely that I would get them?
You can sign an open-source NDA and get the datasheets. It lets me write drivers like the current ca0106 and E-Mu drivers. If you are interested in a NDA, priv-email me.
I have applied to Creative's 'Partnership' program and faxed off the forms last week. I've written user-space drivers for custom wire-wrapped logic + PAL glued hardware before, but not for modern chips, or coded for kernel space. Anyway, there's a first time for everything :-)
Good luck. Who are you talking to?
James