DSPRelated.com
Forums
Online · Live + on demand
February 9-11, 2027
Price goes up Jan 1: register now to lock in today's rate
Register Now →

Please help me with this bad DSK 6211 problem!

Started by jeanpierrepoulin ●March 18, 2005
Hello Jean-Pierre,
 
A few things...
1. The EMIF setup and timing [and memory size] is different between 6211 and 6711 DSKs.  A long time ago I ran into a 6211 DSK that had a different memory manufacturer than previous 6211s that i had seen.  I had to tweak one of the settings to get it to work well.
2. I am sure that you have done this, but check your voltages - especially the 3.3v.
3. If EMIF settings don't fix it, you probably have a bad SDRAM chip.
4. If you are ambitious, try to characterize the failure - do all memory locations fail??  Do all patterns fail??  if you get some that work and some that fail, take a look at the data.
5. If you try to slow the clock down [it's fairly easy to do], you will need to recompute your refresh timing.
 
mikedunn

George Wicks <g...@sbcglobal.net> wrote:


Jean-Pierre:

What Jeff is referring to is a GEL file. It will have
a .gel extension on it. The GEL file is typically used
for initialization of the EMIF (External Memory Interface),
and C6x reset.

The EMIF contains registers which define the memory timings
and other characteristics (row size, column size, etc.).

Chances are if you haven't heard too much about the .GEL
file, you're probably using the default one, which I
believe is something either like DSP621x_671x.gel or
DSK6211.gel or dsk6xinit.gel.

Inside the .gel file you're using, for a DSK unit you
should have a function called something like emif_init()
or init_emif():

emif_init()
{
/*--
-*/
/* EMIF REGISTERS */
/*--
-*/

#define EMIF_GCTL 0x01800000
#define EMIF_CE1 0x01800004
#define EMIF_CE0 0x01800008
#define EMIF_CE2 0x01800010
#define EMIF_CE3 0x01800014
#define EMIF_SDRAMCTL 0x01800018
#define EMIF_SDRAMTIMING 0x0180001C
#define EMIF_SDRAMEXT 0x01800020

/*--
-*/
/* EMIF REGISTER VALUES - these should be verified with TARGET hardware
*/
/*--
-*/

*(int *)EMIF_GCTL = 0x00003040;/* EMIF global control register */
*(int *)EMIF_CE1 = 0xFFFFFF23; /* CE1 - 32-bit asynch access after boot*/
*(int *)EMIF_CE0 = 0xFFFFFF30; /* CE0 - SDRAM */
*(int *)EMIF_CE2 = 0xFFFFFF23; /* CE2 - 32-bit asynch on daughterboard */
*(int *)EMIF_CE3 = 0xFFFFFF23; /* CE3 - 32-bit asynch on daughterboard */
*(int *)EMIF_SDRAMCTL = 0x07117000; /* SDRAM control register (100 MHz)*/
*(int *)EMIF_SDRAMTIMING = 0x0000061A; /* SDRAM Timing register */
}The EMIF SDRAM registers are the ones that are most
critical - the EMIF_SDRAMCTL register, as the name implies,
has alot to do with the control settings for the SDRAM
(items like row & column size), and the EMIF_SDRAMTIMING
register contains the value for the SDRAM refresh period.

The above be the right values for either a 6211
or 6711 DSK. Invalid values can lead to problems.

But, like I said, chances are if you've never heard of
the GEL file, you have those values already.

But, I could be wrong.

Thanks Jeff for pointing that out!

g.
----- Original Message -----
From: "Jeff Brower"
To: "Jean-Pierre Poulin"
Cc:
Sent: Friday, March 18, 2005 5:35 PM
Subject: Re: [c6x] Please help me with this bad DSK 6211 problem!>
> Jean-Pierre-
>
> > From my CCS V 2.20 installation, the filename used for the DSK6211 is
called
> > 'dsk6x11spp378.cfg'. Does this file give you the information you need?
>
> No... you have to look at the last screen of the CCS Setup wizard and find
the
> current Gel file setting. There's a way to check that in CCS also, maybe
a CCS
> super-guru like Bhooshan will jump in and help.
>
> -Jeff
>
> > -----Original Message-----
> > From: Jeff Brower [mailto:j...@signalogic.com]
> > Sent: Friday, March 18, 2005 4:09 PM
> > To: Jean-Pierre Poulin
> > Cc: c...@yahoogroups.com
> > Subject: Re: [c6x] Please help me with this bad DSK 6211 problem!
> >
> > Jean-Pierre Poulin-
> >
> > > Testing out the dsk6xtst application, the test does indeed fail,
mostly
> > > during HPI write.
> > >
> > > Compiling the source code and tracing through the application, the
board
> > > does indeed appear to have SDRAM access problem, and I'm not sure what
to
> > do
> > > at this point. (Maybe I'll look for spare SDRAM parts I have lying
about
> > > and replace the things)
> > >
> > > Beware what you buy on eBay!!
> > >
> > > Thanks for taking the time. It set me off in a new direction and I
can
> > now
> > > do more scientific testing!
> >
> > What .gel file are you using with CCS; i.e. if you go into the CCS Setup
> > program,
> > what's the Initial Gel file set for?
> >
> > -Jeff
> >
> > > -----Original Message-----
> > > From: s...@dlep30.itg.ti.com [mailto:s...@dlep30.itg.ti.com]
On
> > > Behalf Of Jagadeesh Sankaran
> > > Sent: Friday, March 18, 2005 10:42 AM
> > > To: jeanpierrepoulin
> > > Cc: c...@yahoogroups.com
> > > Subject: Re: [c6x] Please help me with this bad DSK 6211 problem!
> > >
> > > Are you sure that there is no problem with the underlying memory SDRAM
> > > in external memory. You can run dsktxt.exe which comes with CCS.
> > >
> > > Regds
> > > JS
> > >
> > > > jeanpierrepoulin wrote:
> > > >
> > > > I've purchased an old DSK6211 on eBay, and I'm having great
> > > > difficulty running the samples. The samples compile and link, but I
> > > > one or more `Data Verification Failed at address 0x8000xxxx' (right
> > > > in the DSK SDRAM memory space) and the program refuses to run
> > > > correctly.
> > > >
> > > > Rebuilding the samples to turn off the L2 cache and fitting all
> > > > sections into the on-chip L2 memory, the samples (those I can make
> > > > fit into 64K ram) work just fine and I can program, run and debug as
> > > > usual if I only use on-chip memory. But as soon as the loader needs
> > > > to load into SDRAM memory, I get these error and nothing works as it
> > > > should.
> > > >
> > > > Checking things further (with a simple program that fits into L2 to
> > > > read / write into SDRAM memory), I find that 1/4 of my writes are
> > > > not being written correctly into SDRAM. If, for example I write a
> > > > pattern such as 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 etc
> > > > into SDRAM memory past 0x8000000, I will see 0x01, 0x01, 0x02, 0x03,
> > > > 0x05, 0x05, 0x06, 0x07 and so on.
> > > >
> > > > If on the other hand I write the same memory pattern backward
> > > > (starting from end going toward 0x8000000), then the pattern is
> > > > written fine!
> > > >
> > > > Looking into the GEL file, I've noticed small differences in the
> > > > EMIF timings between 1.05 and 2.21, but switching back and forth
> > > > makes no difference!
> > > >
> > > > I'm about to pull my hair out as I've been struggling with this for
> > > > at least 30 hours! Any ideas anyone??
> > > >
> > > > A million thanks for any advice?
> > > >
> > > > Jean-Pierre
> > > >
> > > > (The same problem occurs with Code Composer 1.05 that came with it,
> > > > and with CCS2.21 as well. This is a revision 1 board with 4Mbyte of
> > > > RAM, and the DSK jumpers appear set to their default)>

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/c6x/

<*> To unsubscribe from this group, send an email to:
c...@yahoogroups.com

<*> Your use of
Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/




Hi George,

Many thanks for that gracious offer with the DSK 6211. I must admit that it
would save my bacon right now, because of the large amount of 'things to do'
piling up and this unforeseen problem requiring so much time and causing so
much frustration. Why is TI getting so much problems building these
boards?? Is it the chip's EMIF that is fragile or is the board poorly laid
out?

Of course, I'd be more than happy to buy you that beer (or two or three...
but no more than ten!) if you ever get around to Ottawa, Ontario! ('tis
always fun to meet fellow engineers.) Of course, I'll be more than happy to
pay you for shipping, and if you tell me what 'old part' you're looking for
yourself, I'll look in my (increasing pile) of unused development boards if
I can send you something you need in return.

Or, if you're interested in a couple hours of professional time to research
/ review something you're doing, I'd be happy to compensate with that as
well. (I'm very good with FPGAs, embedded systems and professional-level
large-scale software development on PCs and embedded machines) and I have
collected quite a pile of reference designs and schematics of all sorts.

Thanks again!

Jean-Pierre -----Original Message-----
From: George Wicks [mailto:]
Sent: Saturday, March 19, 2005 11:12 AM
To: jeanpierrepoulin
Subject: Re: Please help me with this bad DSK 6211 problem!

Hi Jean-Pierre:

Yes, it very well could be that the internal EMIF is busted, in which case
attempting to swap out the SDRAM would have no effect.

I didn't bother mentioning the GEL files and checking the EMIF settings,
because, well, I went through that pain to begin with when I had first
received my first bad 6211 DSK, and I had to go through the entire GEL
file checking with the TI engineers, which you'll probably still end up
doing -- and I saw that the symptoms that you were experiencing were exactly
like I got, hence I decided to "cut to the chase".

I mean you can check through the EMIF settings, and compare them against
the parts on the board - I have found that TI pretty much used the same
parts on the DSKs, so the GEL file and the DSK should be in sync. Mike
says he's had seen different parts of DSKs - I don't know, he must be right
if he's seen it.

There is a benefit to playing with the EMIF settings, even if it doesn't
solve
the immediate problem - at a later point, if you're doing any embedded work,
you'll need to become familiar with the EMIF for interfacing with SDRAM,
Flash memory, etc.

You <can> try changing the timings - refesh period, ECLKIN, number of
clocks for read & write, etc - and I had already done that with the first
bad
6211 DSK I had (to no avail).

Listen, sitting on my desk is my old 6211 DSK - I'm not using it anymore.
This is the one that had been fixed by TI, and it was working, but the only
caveat I have is that the last time I used it was in 2000. If you give me
the
shipping info, I'd be happy to send it to you. It doesn't have the cables,
the
power supply, etc. - just the board. Now, when you get it, maybe you'll
out it's also got a fault in it - in which you now have 2 bad boards. If, on
the other hand, it works - problem solved, and you can continue with your
work.

Oh, one more thing: you know that when you take the DSK out of the static
bag, you need to put the DSK (or any other board product) on a flat,
non-conductive surface? I've made that mistake in the past, as have many
others - common to do.

Anyways, it's yours if you want it - if I'm ever in your neighborhood, you
can take me out for a beer.

Cheers,
George Wicks

----- Original Message -----
From: "jeanpierrepoulin" <>
To: "George Wicks" <>
Sent: Saturday, March 19, 2005 8:34 AM
Subject: Re: Please help me with this bad DSK 6211 problem! >
> Thanks for taking the time, all.
>
> I've doubled (triple) checked the GEL's EMIF settings and even spent
> several hours deciphering the EMIF control registers from the data
> sheet and EMIF specification to try to 'slow down' the timing and to
> no avail...
>
> Thinking about this problem, it's odd when I write bytes in
> ascending order that one of every four bytes gets ovewritting by the
> one next to it. (If, for example I write a pattern such as 0x00,
> 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07 etc into SDRAM memory past
> 0x8000000, I will see 0x01, 0x01, 0x02, 0x03, 0x05, 0x05, 0x06, 0x07
> and so on.)
>
> From the days I wrote my own SDRAM controller for a FPGA, I remember
> the importance of timing and what to do to write an individual byte
> into a wide SDRAM circuit by using the byte lane signals.
>
> What is odd about that is that these occur in different 'byte
> lanes'. I'd understand what to do if I got this type of problem
> writing 32-bit words that it could be timing related, but writing
> bytes accross different byte lanes?? Maybe one of the BE# lines is
> damaged or even worse, the internal EMIF!
>
> This is why I'm not sure if it would bring any benefit to change the
> SDRAM parts. However, I'm happy to hear that I'm not the only one
> having SDRAM problems with these boards.
>
> From the people who have had SDRAM problem with the DSK6211, would
> it help if I reduced ECLKIN to 75Mhz?
>
> (Thanks again for the time!)
>
> Jean-Pierre > --- In , "George Wicks" <georgewicks@s...> wrote:
> >
> > Jean-Pierre:
> >
> > What Jeff is referring to is a GEL file. It will have
> > a .gel extension on it. The GEL file is typically used
> > for initialization of the EMIF (External Memory Interface),
> > and C6x reset.
> >
> > The EMIF contains registers which define the memory timings
> > and other characteristics (row size, column size, etc.).
> >
> > Chances are if you haven't heard too much about the .GEL
> > file, you're probably using the default one, which I
> > believe is something either like DSP621x_671x.gel or
> > DSK6211.gel or dsk6xinit.gel.
> >
> > Inside the .gel file you're using, for a DSK unit you
> > should have a function called something like emif_init()
> > or init_emif():
> >
> ...
> > The EMIF SDRAM registers are the ones that are most
> > critical - the EMIF_SDRAMCTL register, as the name implies,
> > has alot to do with the control settings for the SDRAM
> > (items like row & column size), and the EMIF_SDRAMTIMING
> > register contains the value for the SDRAM refresh period.
> >
> > The above <should> be the right values for either a 6211
> > or 6711 DSK. Invalid values can lead to problems.
> >
> > But, like I said, chances are if you've never heard of
> > the GEL file, you <probably> have those values already.
> >
> > But, I could be wrong.
> >
> > Thanks Jeff for pointing that out!
> >
> > g.
> >