Re: [avrdude-dev] new USB-based programmer with Igor USB

[Jan-Hinnerk Reichert]

On Tuesday 25 November 2003 15:17, Jeff Epler wrote:
> Because "it could be done", I tried out the avr-based USB device
> code written by Igor Cesko, and quickly had a very versatile,
> USB-controlled device. (http://www.cesko.host.sk)

This site looks quite nice ;-)))

> With 8 available I/O pins, I realized that one use of the device
> might be as a usb-attached AVR programmer using SPI.
>
> My initial design used only the basic I/O capabilities of Igor's
> firmware, with one USB packet for each pin change or pin read.  It
> was terribly slow, taking over 10 minutes to program and verify a
> 270-byte program. But it proved the concept was sound.
>
> The problem was the number of packets: even running over-clocked at
> 12MHz, the Igor USB takes a considerable length of time to handle
> one USB request.

AFAIK, this is not specific to Igor USB. The number of packets per 
second is simply limited to 125 for some reason. Have read something 
about this in the in the I/O-Warrior datasheet a few days ago. Can't 
find anything about it in the USB spec.

However, this seems to fit with the timing data you give.

> But it takes 128 pin changes or reads to do a
> 4-byte SPI packet, and multiple packets to write one byte of flash.

If you don't use polling it should be two 4-byte-commands. 

>  So I added to Igor's firmware a command to perform 8 consecutive
> bits of SPI I/O (getting rid of 31 out of every 32 requests) and
> the time to program dropped to 30 seconds, which is an acceptable
> speed for me.  Removing support for the status LEDs removed another
> 10 seconds from programming time, showing that the number of USB
> packets is still the greatest concern.

IMHO this won't be acceptable for most people ;-(
Just think of the programming times for larger projects.

However, the concept is still interesting. Unfortunately, I don't have 
the time to dive deeper into USB right now.

/Jan-Hinnerk