| |
In lugnet.robotics.nxt, David Wallace wrote:
> I'm hoping that Atmel's 10,000 number is conservative. I also hope that the NXT
> firmware doesn't reprogram the same location. If I'm working on a program and
> reflashing it many times, I think I'm going to keep changing the name so it gets
> stored at a different location until the flash is full, and then erase and start
> over.
When downloading a program to the NXT using NBC if there is a file with the same
name on the NXT already it is first deleted. Then a new file is created and
written using the same filename. I don't know whether that means the firmware
will write the file to a different location or not.
Executables and icon files have to be created using OpenWriteLinear which means
they have to be contiguous. The VM actually reads the byte codes directly from
flash as the program executes.
There was a discussion some time back in which, if I recall correctly, it was
said that the NXT firmware doesn't do anything fancy, like wear leveling, for
the flash memory.
> Any knowledgeable people out there in this area?
Dick Swan comes to mind, but it may fall under the category of protected
intellectual property.
John Hansen
|
|
| |
> John Hansen wrote:
>
> > Any knowledgeable people out there in this area?
>
> Dick Swan comes to mind, but it may fall under the category of
> protected intellectual property.
I researched this topic about eight months ago. I no longer worry about
it. I must admit I didn't know about the "lock-bits" specification and
was concerned only with the "flash rewrite specification".
I have also had other conversations on this topic which unfortunately
are covered by NDA.
Here's some public domain information tidbits that are relevant.
1) Limit on flash memory cycles is across the complement of extreme
temperature ranges and voltage battery levels of the device. I double it
anyone intents to operate their NXTs at these temperature extremes.
2) I never kept the web reference, but there was a similar question
posted in a newsgroup for a robotics design contest that was held a few
years back. The contest product used an Atmel AVR series CPU.
There was a response from an Atmel Support Engineer -- his reply was
along the lines of "our specs are conservative", "specs are for
temperature extremes". The most interesting comment was "the main
failure mode is a reduction in the data retention time". I think most
NXT users are willing to accept a reduction in the 40-year data
retention specification for flash.
|
|
|
