Open Source/Freeware Budget USB DAC

[keeperx]

great.. (i missed that)

[riyalasers]

I think "Shutter will be TTL output along with 7 other IO lines" - it is bad idea.

will be better to have 8-bit port, and separate "shutter" signal

Some projectors (Neo-Neon's, for example) will not work without high level signal on "shutter" (pin 13 of ILDA connector) line.
So I think high level should be setted on this line within frame translation "as default"

If you need separate controll for this line - you can make separate function for it.

[mixedgas]

OK, we put pads for a open collector 3904 or 2222A (will that %$#@ part# ever die...) with say 1mA of base drive on the board for the shutter. If somebody wants it bad enough, they can drop in a NPN of their choice in a to92 and a 1K resistor and a 4001 diode for the back emf and get it over with. They then jumper a loop of wire from the TTL header to the base and they have a OC for the relay, That way they can have TTL or open collector and just have to find a relay with a 600 or 1K ohm coil, or a opto 22 for the shutter. At worse, this way it adds less then 20 cents to the cost per board.
They provided the 1k, the npn, the diode and the wire and the time, and the hot soldering iron. Its thru hole, so if they pop the magic smoke, they can change it out without sending the board, back to the factory.

we probably should put some 470 ohm resistors in the analog outputs, to prevent shorts from blowing them. Most modern opamps dont mind shorts to ground, but a short to VCC results in sparx.

I'd like a T1 and 3/4s or a TTL or a test point some place that shouts I'm Enumerated, I added said emitter to my soundcard dac, the holes and signals were already on the board, and that does a steady when enumerated and 1 hz flash when pumping audio out, makes for a nice sanity check. They must have made a clear case version at one time and dropped it when they went metal. I mean how hard was it not to notice 4 labeled pads and the silkscreen.

I'm suprised the sound card dac guys missed that one.......... But its there..

Steve

[Paul]

In the past I have made all sorts of stuff using Rabbit core modules. www.rabbitsemiconductor.com
They come loaded with features for not much money. Look at their site for the options. You can get the Dynamic C development environment for free too... If you go for the cheapest option, the RCM5700 (http://www.rabbit.com/products/rcm5700/) then you are looking at $35 for a complete module! The only extra cost would be the DACs with their output buffers.... For the $35 you even get an ethernet interface!

I was looking into building a DAC / player based on the RCM5700 myself, but I decided that for now I am better off buying something good. So I went for the FB3 :-) . But I am still considering building something next to the FB3, even if it is just for fun!

[keeperx]

http://www.rabbit.com/products/rcm4200/
this is analog over ethernet.. neat..

this has flash memory
http://www.rabbit.com/products/rcm3900/
thease things are hella cool..

[Paul]

The good thing is that all of these modules come with diagrams etc. You can create your own custom module and still use the supplied libraries. I was going to use a RCM5700 with a SD flash card like the 4300.... No special coding required, everything is already there!

[mixedgas]

The good thing is that all of these modules come with diagrams etc. You can create your own custom module and still use the supplied libraries. I was going to use a RCM5700 with a SD flash card like the 4300.... No special coding required, everything is already there!

A rabbit is a zilog z80 core, the company bought the z80 rights and have expanded on them, Supposed to be almost perfectly backward compatible with the z80.

Steve

[heroic]

It takes 1.6 us to write to 1 DAC appx (not including the SS pin and misc. overhead - 16 bits per xfer). that is ideal and at 10 MHz SPI clock rate (the max the ATMega family can do). You have to do this x 6 = 9.6 us ... so you now have ~7 us left to transfer data and do other house keeping. Seems doable.

The only other thing would be how to transfer all of the data from the PC to the micro. That is a nice chunk of data. What are the plans?

1. Be efficient. Write two DACs at once if they need the same value (yes, we can do this).

2. Decimating data saves tons. Instead of sending (x, y, r, g, b, v) tuples over the wire, we send variable-length tuples with a type mask at the front. That way we avoid repeating values that don't change.

3. Support palettes, kinda. Most of the data in a full (x, y, r, g, b, v) tuple is colour info, so keep a stash of the last sixteen (r, g, b, v) points to be used and refer back to them.

4. If all else fails, Huffman trees are fast.

[mixedgas]

Or skip the usb and go right to ethernet UDP

Steve

[heroic]

Or skip the usb and go right to ethernet UDP

Steve

That would sadly make the board much more expensive and complicated. The problem is not USB's bandwidth, it is the bandwidth of the USB chipset we are using.