SparkFun Forums

[KeyserSoze]

Would it be possible to have something like the DLP-USB245M, but with more bandwidth?

I really like the FIFO interface on the 245M for it's simplicity, but would love something a little faster. I don't need USB 2.0's theoretical max of 480Mbps, but 20Mbps+ would be very much appreciated. A 16-bit wide data path instead of 8-bit wide would be okay, or even 32-bits wide if that would allow more throughput. Since the DLP-USB245M is IIRC under $20, under $50 for the uber-high-speed replacement would be ideal.

Thanks.

[dlharmon]

Would it be possible to have something like the DLP-USB245M, but with more bandwidth?

I really like the FIFO interface on the 245M for it's simplicity, but would love something a little faster. I don't need USB 2.0's theoretical max of 480Mbps, but 20Mbps+ would be very much appreciated. A 16-bit wide data path instead of 8-bit wide would be okay, or even 32-bits wide if that would allow more throughput. Since the DLP-USB245M is IIRC under $20, under $50 for the uber-high-speed replacement would be ideal.

Thanks.

Check out
http://oscar.dcarr.org/ssrp/
David Carr is using a board with the Cypress CY7C68013 USB 2.0 Chip. It has a 16 bit wide FIFO interface. The chip itself costs $13 from digikey. I believe the board used in the project linked costs $90. The only bad thing about the CY7C68013 is the docs are terrible. It has an integrated 8051 so that is something else to program and another cross compiler. That would have put me up to 4 versions of gcc on my box (arm, blackfin, x86, 8051).
It comes in SSOP-56 which should be easy to solder.

I ended up using the NET2272 from PLX, but it doesn't really fit your description. It looks like 16 bit SRAM to the microcontroller which is perfect for my application since I have it wired up to a Blackfin at 390Mhz. It is well documented. They claim I should be able to move 320mbits/sec. I still have not tested it for datarate, but I have got it working at high speed.

Maybe spark fun could do a CY7C68013 board.


Darrell Harmon

[jayjay]

Hi

>A 16-bit wide data path instead of 8-bit wide would be okay, or even 32->bits wide if that would allow more throughput

Taken a look at the Freescale's MC9S12UF32? Not in module form yet, but looks like a neat chip for USB stuff. I've got the sdk and sample chips, yet to play with.

Cheers
Jay