Go Timothy! This could wind up being big stuff!
Gary
On 4/4/06, Tom Cook <[EMAIL PROTECTED]> wrote:
Bah, I'll learn how to use a mail client properly some day...On 4/5/06, Timothy Miller <[EMAIL PROTECTED] > wrote:Unless you wanted this to remain private, feel free to forward my
response to you to the list.
On 4/4/06, Tom Cook <[EMAIL PROTECTED]> wrote:
> Here's a list of things I'd want in this sort of hardware:
>
> 1. Balanced inputs. A typical sound card interface (unbalanced 1/4" jack)
> is no where near sufficient for good audio input. XLR connectors.
What are balanced inputs?
The typical balanced system has three conductors. One is an earth and is not used for carrying signal. The signal is the voltage difference between the other two wires. Neither of these is held at any particular reference voltage at either end (ie. neither is connected to ground). By taking the difference between the two wires, you cancel out any noise induced on both of the wires. The idea is that any electromagnetic field in the area will induce (at least approximately) the same noise on both wires, because they are very close together. Then when you take the difference between the two wires you cancel out the induced noise. This is not a very good explanation; see eg http://www.dself.dsl.pipex.com/ampins/balanced/balanced.htm for a fuller one.
Phantom power is achieved by putting 48VDC on BOTH signal conductors, relative to the third conductor. The thing at the other end can then take 48 volts between a signal line and earth, and send a signal back in the difference between the signal lines. This difference induced between the lines is small enough (typically 1mV) that it doesn't much affect the 48V supply. Needless to say, there is lots of coupling/decoupling involved.>
> 2. (VERY) High quality op-amps at the inputs, with matched resistor sets at
> the inputs to give as high as possible common mode rejection.
I actually have some vague clue about what this means, and I have to agree. :)
>
> 3. 48V "phantom" power provided on all inputs. There is lots of kit out
> there (DI boxes, condensor mics and others) that expects to power itself
> from this.
>
> 4. 192kHz 24-bit.
>
> 5. At least 30 channels, preferrably 50-60.
>
> 6. Small frame size.
>
> 7. The ability to switch any channel from input to output in software.
> Needn't be dynamic (in fact probably shouldn't be) but you do need to be
> able to reconfigure it easily.
Well, a channel is either going to be input or output. ADC and DAC
are totally different things. It might be possible to turn off a DAC
well enough that you could have an ADC and DAC on every connector.
But that might cause reflections and introduce noise. As for software
configuration, that's up to the software developer how they want to
rename channels; that's not a hardware issue.
I might be wrong, but I don't think reflections are a big deal on a PCB that is only handling 100kHz tops. Also the ADC/DAC don't need to be directly connected, since you are almost certainly going to want a buffer op-amp going either way. Still need some way to switch between them.>
> Points 1, 3 and 5 (and probably 2) all demand an external box to go with
> this. You can't get 60 XLRs onto a PCI card. This probably means moving
> the ADCs into the external box, to avoid noise on the connectors. At this
> point you might as well put the whole thing into the external box and put a
> 1394 or USB2 interface on it. Point 3 will also require that it have its
> own power supply.
USB wouldn't have enough bandwidth for so many channels at once.
Sure, there's lots of kinds of processing where the data would never
leave the card, but in many cases, you're going to be shipping data
back and forth with the application. You need high bandwidth for
that, and even PCI might not be quite fast enough if all channels are
uncompressed.
?
I make 192kHz x 24bits to be 4608000 bits per second. Sixty channels of that is 276.48 Mb/s. All of USB2, 1394 and gigabit ethernet are capable of that. The trick is probably getting the latency down.But I do like the idea of having an external box with all of the
connectors on it. We do everything digitally inside of the PC, bring
it out, isolate it, and then do the conversion to/from analog.
Two small FPGAs are cheaper than one big one with the same total capacity.
> The small frame size is so that you can use the thing in real time as a
> digital mixing desk for live work. If the frame size is small enough then
> you can digitize, process, mix, process and output quick enough to do
> reasonable live work (the ability to reconfigure channels is also useful
> here).
Definitely a desirable feature.
> Another thing would probably be connectors for inserts on each channel. But
> pretty soon this becomes an analogue mixing desk, not a digital audio
> processor.
Are you talking about putting analog attenuation inline with the
inputs? I think this is something we'd want to control in the
software. If what you want is an analog mixing desk, you can use the
mouse for that.
No. Most analogue desks have an insert jack on each channel. This is a stereo tip-ring-sleeve type connector. If you plug something into it then it breaks the normal signal path and sends it through whatever is connected between the tip and the ring. It's a way of inserting a chain of external processing units into the signal path.
Now, I can already hear you typing, "But you can do all your processing in software! Even better, in the FPGA!" And do you know what? You are RIGHT. But sound engineers don't think like that. Each one has his own effects box that is THE best effects box IN THE WORLD and the only way to stop him using it is to prise it from his cold, lifeless grip.Mind you, I am not against the idea of someone hooking up desk that
looks like an analog mixing desk (but which is digital and on USB or
something) and interfacing it with the software.
Another cool idea, but I know what my wife would say...>
> The software stack to go with this already looks pretty good. Ardour is a
> very nice multi-track recorder and mixer, and can use LADSPA plugins for
> processing. Rosegarden is a pretty good sequencer, lilypond produces very
> nice printed music. JACK is a good transport layer. The list goes on.
Excellent. Plus, for anything missing, we can work together to add it.
It's also worth noting that there are a number of CODEC chips out there that do a lot of this for you.
Tom
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC ( www.duskglow.com)
_______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
