High speed USB chip?

Very difficult to get 80 Mb/s through USB.

Ehhh?

"Hi-Speed USB mode is capable of ... 480Mbits/second."

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |

But chopped into 64 byte chunks and host enabled signalings. A missed timing can delay the communications by msecs. You need very good host hardware and firmware control on the PC side.

In theory, you can send 50 mpeg streams over usb (480/9). I have not seen too many device with more than 1 stream.

Well, it's true, you can't use crap USB chipsets and lousy drives, but the GNURadio guys routinely sustain ~256Mbps over USB with decent quality hardware and software; I believe I've heard them say that realistically the USB 2.0 limit is around ~320Mbps (40MBps).

80Mbps should be no problem.

Granted, I wouldn't stick this data acquisition device on the same bus as the one you have a web cam, audio device, mouse, keyboard, etc. on -- I'd give it a USB controller of its very own.

480M(bits)/sec / 8 bits per byte = 60M(bytes)/sec

without any overhead.

It would be easier with a TCP connection.

donald

"Ancient_Hacker" a écrit dans le message de news: snipped-for-privacy@e1g2000hsg.googlegroups.com...

Cypress has some high speed USB enabled microcoontrollers that allows a data stream to be directly routed to the USB engine, see for example the EZUSB FX2LP (CY7C68014A ), data throughput is claimed to be 53Mbyte/s, which should be ok for your application.

Cheers,

--
Robert Lacoste
ALCIOM - The mixed signal experts

We are using the Cypress FX2 series parts, as someone else mentioned, for several projects. The FIFOs are a bit small so we often have to use an external FIFO. For my current project I have one bolted to CPLD and external SRAM for a sorta FIFO.

Note for USB2 the theoretical maximum is 480M BITS/Sec. So that is 60M Bytes per sec. In reality 80% is about all you can do or around

50mBytes per sec (Cypress claims a bit better though). Your project is 8 x (100nS^-1) or 80mBytes per sec. Not doable sustained, Perhaps cached locally and sent later?

Hawker

On 3/20/2007 5:03 PM, The digits of Ancient_Hacker's hands composed the following:

Er. You are out by a factor of eight. He is pulling one byte every

100nSec. Total just 10MB/sec. Perfectly doable.

Best Wishes

Umm... shouldn't that be 80Mbits per second? He wanted to sample one bytes every 100nS = 10 million samples/sec = 10Mbytes/sec = 80Mbps?

The second question though is using 'm', not 'M'. Normally 'b' is bits, 'B' is bytes, 'm' is milli, and 'M' is mega. If he only wanted 80mBits, it could be done using RS232 serial!... The stupid thing is just how easy this type of slip is. I sent a post myself to 'Hawker', and managed to generate the right answer, but type 'mSec', instead of 'nSec'.... It is a big danger of using the 'prefix' letters.

Best Wishes

Yes as Roger and you pointed out I goofed. That is what I get for only half paying attention to what I am doing.

100nS-1 = 10M/bytes *8 = 80mBits per second. So this will fit no problem I blew it.

Hawker

On 3/21/2007 11:46 AM, The digits of Joel Kolstad's hands composed the following:

Only on a dedicated single device USB controller. Protocol overhead and protocol timing limitations matter.

--
 JosephKK
 Gegen dummheit kampfen die Gotter Selbst, vergebens.  

These guys :

do a module based on the Cypress chip which comes with driver software to make USB2 comms fairly painless. However you might need to add a fifo between the ADC and the USB interface to guarantee no data loss.