Avr Doper performance

Sparky · Post by **Sparky** » Fri Feb 16, 2007 2:16 am

Hi,

I built up the AVR Doper circuit and it seems to be working. The only problem is that it takes a very long time to program a device. My experiments have seen it take 14 minutes to program and verify 75K bytes of code. This is about 200 bytes per second to program and 170 bytes per second to verify. My system is running avrdude 5.3.1 on Ubuntu linux V6.06. Has anyone seen better performance ?

Thanks

Post by **christian** » Fri Feb 16, 2007 11:59 am

AVR-Doper may be a bit slow in serial emulation mode because the USB driver does not leave much time for the main task, but what you report is definitely out of question. Have you set the low-speed jumper? Or have you set a slow ISP bit clock? And finally: Do you get better performance in HID mode?

Sparky · Post by **Sparky** » Fri Feb 16, 2007 6:32 pm

Hi,

Thank you for the response. Will checkout what you suggested today.

Best regards

Sparky · Post by **Sparky** » Fri Feb 16, 2007 9:45 pm

Hi,

I tracked down the problem. Looks like it is a issue with Ubuntu kernel version. I ran the Avrdude 5.3.1 on my laptop with Gentoo and programming took only 75 seconds and 71 seconds to verify. I scoped out the USB activity after issuing a "read eeprom 0 100" command in terminal mode of Avrdude. Ubuntu was using up 90% of the bus bandwidth while Gentoo was using up 12%. So it would definitely seem that this is a driver issue. I am planning to upgrade my Ubuntu system to 6.10 and will see if this improves.

Thanks for open sourcing the design

Best regards.

Post by **christian** » Fri Feb 16, 2007 10:29 pm

If you use serial emulation mode, this may not even be a bug. AVR-Doper implements bulk endpoints in this mode which violates the spec for low speed devices. This may confuse the bandwidth allocation algorithm in the host driver.

HID mode should not suffer this problem.

sfreak · Post by **sfreak** » Thu Mar 15, 2007 10:37 pm

Hi,

what programming times can I expect from AVR Doper? Programming and verifying a ATmega324P takes about 130s for me. Faster programming would be nice... I am using HID mode with avrdude in Mac OS X.

thx Sebastian

sfreak · Post by **sfreak** » Thu Mar 15, 2007 11:05 pm

Uh... stupid mistake! I still had the pull-down resistor on the slow jumper pin from my first attempts... works perfectly now that I removed it. Sorry for wasting you time!