Virtual USB-port for AVR

albertsm · Joined: 09 Apr 2004 Posts: 5921 Location: Holland

compared to which previous version?
The only noticeable changes were for xmega port handling, special registers are used which need less code.
Further, some chips that have movw use movw instead of 2 mov instructions.
And the setbit/reset bit handeling uses R23 instead of r24. These are isolated instructions and should not give impact.
If you can not find it, you can always send me the project and let me know the versions so i do a check.

six1 · Joined: 27 Feb 2009 Posts: 553

one hint:
the whole thing didn't llike long USb Cables! with 1,8m cables it's working... most the time, but there are dysfunctions also! longer cables result in a lot of errors

Petr_ · Joined: 24 Mar 2010 Posts: 66

Mrshilov · Joined: 24 Jan 2009 Posts: 314 Location: Russia

Mr. Ollopa! Any changes?

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

Hello,

I've just moved again to a new house (I always have an excuse, don't I?). If I remember correctly, the last time I worked on the library I was having some new issues with the 15MHz crystal (but 12MHz was working OK... It is usually the opposite!), and I had discovered a bug thanks to Petr_'s test program. I found that I could cause the USB code to fail if I sent a particular pattern of bits over the wire. I think I was able to reproduce the problem reliably with only one or two bytes of data, but it proved to be very difficult to isolate the problem to something specific like a particular stuffbit. I now have access to a USB protocol analyzer which I have not yet used to help troubleshoot this problem, but I believe it might help.

Let me use the analyzer this weekend and I'll report back my findings.

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

I found and fixed one very obvious bug and now the random test runs much longer. It has uncovered a few more test patterns that fail, however, so I have some more debugging to do this weekend.

In all cases, the failed packets have a stuff-bit that I am probably not handling correctly. Petr_, thanks again for the test program. It has been very helpful Smile

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

Got another bug today. Looks like 15MHz is working fine. Here are some screenshots to show the progress -- I haven't seen a single failure at 15MHz since fixing the two bugs found this week.

12MHz works for a little while and then quits. Doesn't matter what pattern of data I send, so I think it's more of a timing issue than an algorithm issue. I'll do more troubleshooting at 12MHz and put out an update for testing soon....

Anybody still out there to DO any testing?

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

Yup -- found timing bugs at 12MHz today. I've got several test patterns of data that are designed to expose bugs in the SIE -- I'm now able to run through all the patterns without any trouble. I also haven't had any problems running random data.

I have finally resolved some long-standing bugs I've never had the time to track down before Very Happy

And Mark will be happy to hear that the problems weren't related to the compiler version after-all. I made a change to fix a bug probably a year ago and that change introduced several new timing bugs. Sorry, Mark Embarassed

I'll be making my first official release soon and then I can put my effort into writing examples and documentation. The next step will then be support for clock speeds other than 12MHz and 15MHz.

Mrshilov · Joined: 24 Jan 2009 Posts: 314 Location: Russia

Nice news! We are waiting.

compurap · Joined: 23 Oct 2007 Posts: 43

Thanks, We will be wait. Smile

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

Hi all,

I'm checking-in again.

A synchronization issue was discovered after the first round of testing on different parts. I have resolved the issue and sent the update out for verification. I have personally tested the library on a Tiny461, Mega8, Mega48, Mega644, and Mega2560. 12MHz and 15MHz are working and I should be able to easily add others speeds soon.

I'll post this version of the library within a couple of days.

glena · Joined: 25 Jul 2007 Posts: 284 Location: PA near Philly

When you say other speeds may be possible, do you think 18.432MHz might be possible? I use it for a lot of things since its the fasted that is baud rate friendly.

-Glen

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

Hi Glen,

I'm fairly certain I can make that work. I think I even purchased an 18.432MHz crystal to try it for you. I'll scrounge around tonight and see if I can find it.

Last night I finished my HID keyboard example and tested it -- it's working great now. I want to do a mouse next and put them on a webpage or perhaps submit application notes.

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

I'm attaching version 1.06 of the library for now. I've tested it on several devices and believe it to be suitable for public distribution. I'm still working on the mouse example and adding 16MHz support. If I have time, I'll post the keyboard example over the weekend.

ollopa · Joined: 03 Sep 2007 Posts: 233 Location: California

I'm attaching my keyboard example (login to download).

This was asked for a long time ago and I hope it can still be of some use now. AndersL kindly posted a similar demo before me. Most things are the same, however my version has a routine to convert ASCII to a keyboard usage code and another routine to "print" strings to the keyboard. I did not spend time on optimizations. I hope it will serve as a useful example for somebody who will improve the implementation and post it back Smile