Atmel flash memory issue, maybe?

Hello folks,

I am using Atmel flash AT49BV322DT in a custom dsp board i am building.
the dsp is a TI 320DM641. I have written test software that erases, writes
and reads from the flash.

When I send the sequence of erase commands to the flash, I am able to see
write signals on the scope. Shortly thereafter, I also see the Ready/Busy/
output from the flash go low, as expected. Same thing while writing. I cam
also read the data back from the flash. so far, so good.

yesterday, all of a sudden, after half a day's experimenting with flash
timing, the thing stopped working. the READY/BUSY signal would not respond
to any command sequence.

The only thing I can think of is that one must __ALWAYS__ do an erase
before a write. While exprimenting with flash timings, I performed quite a
few (8-10-20) writes. unfortunately, i did not scope the signals during
each write, so i do not know exactly when the thing stopped working.

So, my question is: do I always need to erase before reprogramming the
flash? Is this true of all flash memories?

tomorrow i am going to replace the chip.

Any help will be greatly appreciated.

Thanks!

--Vishnu

Vishnu Swaminathan wrote:
> Hello folks,
>
> I am using Atmel flash AT49BV322DT in a custom dsp board i am building.
> the dsp is a TI 320DM641. I have written test software that erases, writes
> and reads from the flash.
>
> When I send the sequence of erase commands to the flash, I am able to see
> write signals on the scope. Shortly thereafter, I also see the Ready/Busy/
> output from the flash go low, as expected. Same thing while writing. I cam
> also read the data back from the flash. so far, so good.
>
> yesterday, all of a sudden, after half a day's experimenting with flash
> timing, the thing stopped working. the READY/BUSY signal would not respond
> to any command sequence.
>
> The only thing I can think of is that one must __ALWAYS__ do an erase
> before a write. While exprimenting with flash timings, I performed quite a
> few (8-10-20) writes. unfortunately, i did not scope the signals during
> each write, so i do not know exactly when the thing stopped working.
>
> So, my question is: do I always need to erase before reprogramming the
> flash? Is this true of all flash memories?
>

Not the whole flash, but yes u have to erase those specific sectors
where you are going to rewrite. If you use any of flash writers, they
actually erase the required sectors before reprogramming.


> tomorrow i am going to replace the chip.
> 
> Any help will be greatly appreciated.
> 
> Thanks!
> 
> --Vishnu

Thanks Sudip.

But is rewriting without erasing so bad that it will destroy the entire
flash chip?

--Vishnu

Sudip wrote:
> Vishnu Swaminathan wrote:
> > Hello folks,
> >
> > I am using Atmel flash AT49BV322DT in a custom dsp board i am building.
> > the dsp is a TI 320DM641. I have written test software that erases, writes
> > and reads from the flash.
> >
> > When I send the sequence of erase commands to the flash, I am able to see
> > write signals on the scope. Shortly thereafter, I also see the Ready/Busy/
> > output from the flash go low, as expected. Same thing while writing. I cam
> > also read the data back from the flash. so far, so good.
> >
> > yesterday, all of a sudden, after half a day's experimenting with flash
> > timing, the thing stopped working. the READY/BUSY signal would not respond
> > to any command sequence.
> >
> > The only thing I can think of is that one must __ALWAYS__ do an erase
> > before a write. While exprimenting with flash timings, I performed quite a
> > few (8-10-20) writes. unfortunately, i did not scope the signals during
> > each write, so i do not know exactly when the thing stopped working.
> >
> > So, my question is: do I always need to erase before reprogramming the
> > flash? Is this true of all flash memories?
> >
>
> Not the whole flash, but yes u have to erase those specific sectors
> where you are going to rewrite. If you use any of flash writers, they
> actually erase the required sectors before reprogramming.
>
>
> > tomorrow i am going to replace the chip.
> >
> > Any help will be greatly appreciated.
> > 
> > Thanks!
> > 
> > --Vishnu

swams wrote:
> Thanks Sudip.
>
> But is rewriting without erasing so bad that it will destroy the entire
> flash chip?
> 
> --Vishnu
> 

Not at all. What about reading the Data Sheet?

On Wed, 01 Nov 2006 20:55:16 -0800, swams wrote:
 
> But is rewriting without erasing so bad that it will destroy the entire
> flash chip?

Production code in automotive engine controllers rewrites flash without
erasing.  We use certain bytes to indicate certain areas are erased,
programming in-progress, programmed, and invalid.  As long as we program
0's to bits which are 1's, there is no problem.  The four states above
would be:

    $FF -> $BF -> $AF -> $AB

We allocate fixed size blocks in a flash sector and use the bytes to
indicate the state of the blocks.  When all the blocks in a sector have
been used, we transfer to a second sector and erase the first sector.  The
controller writes a new block every power-down and invalidates the
previous block.  


    ~Dave~