Hi All,
We have a DM640-based custom board with CPU operating at
400MHz. This board was working fine. Now we have replaced the DM640
with DM642-720 chip. The input PLL clock is same, i.e. input PLL clock
is 33MHz and DM642 CPU is also running at 400MHz. Is this fine? We are
trying to access SDRAM using the same GEL file as used for DM640 board.
Since we have not changed any clock inputs, we are using the same EMIF
settings. But SDRAM is not able to retain the data written to it.
Please let me know if there are any points that we are overlooking? Is
a direct replacement of DM640 by DM642 possible? If not, what points
should be taken into consideration?
DM642 SDRAM issue
Started by ●August 18, 2008
Reply by ●August 19, 20082008-08-19
kalai,
On Mon, Aug 18, 2008 at 9:55 AM, Kalaivani.S wrote:
> Hi All,
> We have a DM640-based custom board with CPU operating at
> 400MHz. This board was working fine. Now we have replaced the DM640
> with DM642-720 chip. The input PLL clock is same, i.e. input PLL clock
> is 33MHz and DM642 CPU is also running at 400MHz. Is this fine? We are
> trying to access SDRAM using the same GEL file as used for DM640 board.
> Since we have not changed any clock inputs, we are using the same EMIF
> settings. But SDRAM is not able to retain the data written to it.
1. Have you checked all of the SDRAM control signals?? If you soldered
the DM642, you could have affected another device on the board.
> Please let me know if there are any points that we are overlooking?
2. Are the devices the same revision??
3. Was you DM640 in a socket??
If not, approach this like a new board bringup. You could have a
solder problem with the new device. Check the power and control
signals on the SDRAM - even though the board was working, "make NO
assumptions!'.
> Is
> a direct replacement of DM640 by DM642 possible?
I do not know the answer - this sounds like a question for TI to
answer. I would expect that they would have an app note, although I
did not see one.
The devices do appear to be 'highly similar'.
Make sure that you check all of the relevant pins for their
connections. Sometimes similar devices have 1 or 2 pins that are
different - this may be an additional ground or possibly a reserved
pin that is tied to the opposite polarity on each device. I would
suggest printing out the portion of the datasheet with the pin info
and having someone not directly involved with the project eliminate
all of the identical pins [make sure that every power and ground pin
is checked individually]. Then you can investigate the rest - one by
one. This will take some time, but in the absence of specific
compatibility info, it seems like the best route.
> If not, what points
> should be taken into consideration?
4. Do you have a working DM640 board??
If yes, you could compare the SDRAM signal timing.
5. If you are always reading 1s from SDRAM, write 0s and check the
data bus for 0s. With a constant 'read loop' from SDRAM, check the
data bus [and control signals] with a scope. If you see 1s, something
is up with SDRAM, if you see 0s, look at the DM642.
mikedunn
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
On Mon, Aug 18, 2008 at 9:55 AM, Kalaivani.S wrote:
> Hi All,
> We have a DM640-based custom board with CPU operating at
> 400MHz. This board was working fine. Now we have replaced the DM640
> with DM642-720 chip. The input PLL clock is same, i.e. input PLL clock
> is 33MHz and DM642 CPU is also running at 400MHz. Is this fine? We are
> trying to access SDRAM using the same GEL file as used for DM640 board.
> Since we have not changed any clock inputs, we are using the same EMIF
> settings. But SDRAM is not able to retain the data written to it.
1. Have you checked all of the SDRAM control signals?? If you soldered
the DM642, you could have affected another device on the board.
> Please let me know if there are any points that we are overlooking?
2. Are the devices the same revision??
3. Was you DM640 in a socket??
If not, approach this like a new board bringup. You could have a
solder problem with the new device. Check the power and control
signals on the SDRAM - even though the board was working, "make NO
assumptions!'.
> Is
> a direct replacement of DM640 by DM642 possible?
I do not know the answer - this sounds like a question for TI to
answer. I would expect that they would have an app note, although I
did not see one.
The devices do appear to be 'highly similar'.
Make sure that you check all of the relevant pins for their
connections. Sometimes similar devices have 1 or 2 pins that are
different - this may be an additional ground or possibly a reserved
pin that is tied to the opposite polarity on each device. I would
suggest printing out the portion of the datasheet with the pin info
and having someone not directly involved with the project eliminate
all of the identical pins [make sure that every power and ground pin
is checked individually]. Then you can investigate the rest - one by
one. This will take some time, but in the absence of specific
compatibility info, it seems like the best route.
> If not, what points
> should be taken into consideration?
4. Do you have a working DM640 board??
If yes, you could compare the SDRAM signal timing.
5. If you are always reading 1s from SDRAM, write 0s and check the
data bus for 0s. With a constant 'read loop' from SDRAM, check the
data bus [and control signals] with a scope. If you see 1s, something
is up with SDRAM, if you see 0s, look at the DM642.
mikedunn
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
Reply by ●August 21, 20082008-08-21
Kalaivani,
Please copy the group on all posts.
On Thu, Aug 21, 2008 at 1:25 AM, kalai vani wrote:
> Mike,
> Thanks for your inputs. We tried out accordingly.
> We checked the SDRAM control signals on our DM642 board with our
> working DM640 board. The SDCAS, SDRAS and WE signals look similar. But then
> when we probed the data lines, we found some anomaly. When we wrote 1's to
> all 32-bits, the voltage levels of the some SDRAM data lines was1.2V instead
> of 3.3V. When we wrote 0's to all 32-bits some SDRAM data lines had voltage
> levels of 1.2V as against 0V.
> Any suggestions?
I assume that something is connected or shorted to the core voltage [I
assume is 1.2/1.4v].
Do the 1.2v data lines look like a power supply voltage or do they 'toggle'??
You need to be very precise about which pins are working correctly and
which are at 1.2v. The signal names/pin numbers could be a clue.
Have you checked the DM640 vs. DM642 pin assignments??
I am currently thinking that -
[1] The DM640/642 are not interchangeable
or
[2] There is a problem with the workmanship on the installation of the
DM642. Can you get the board xrayed?? [That would show an alignment
problem and some types of shorts].
mikedunn
> Thanks,
> Kalaivani.S
>
> --- On Tue, 8/19/08, Michael Dunn wrote:
>
> From: Michael Dunn
> Subject: Re: [c6x] DM642 SDRAM issue
> To: "Kalaivani. S"
> Cc: c...
> Date: Tuesday, August 19, 2008, 8:34 PM
>
> kalai,
>
> On Mon, Aug 18, 2008 at 9:55 AM, Kalaivani.S
> wrote:
>> Hi All,
>> We have a DM640-based custom board with CPU operating at
>> 400MHz. This board was working fine. Now we have replaced the DM640
>> with DM642-720 chip. The input PLL clock is same, i.e. input PLL clock
>> is 33MHz and DM642 CPU is also running at 400MHz. Is this fine? We are
>> trying to access SDRAM using the same GEL file as used for DM640 board.
>> Since we have not changed any clock inputs, we are using the same EMIF
>> settings. But SDRAM is not able to retain the data written to it.
>
> 1. Have you checked all of the SDRAM control signals?? If you soldered
> the DM642, you could have affected another device on the board.
>> Please let me know if there are any points that we are overlooking?
>
> 2. Are the devices the same revision??
> 3. Was you DM640 in a socket??
> If not, approach this like a new board bringup. You could have a
> solder problem with the new device. Check the power and control
> signals on the SDRAM - even though the board was working, "make NO
> assumptions!'.
>> Is
>> a direct replacement of DM640 by DM642 possible?
>
> I do not know the answer - this sounds like a question for TI to
> answer. I would expect that they would have an app note, although I
> did not see one.
> The devices do appear to be 'highly similar'.
> Make sure that you check all of the relevant pins for their
> connections. Sometimes similar devices have 1 or 2 pins that are
> different - this may be an additional ground or possibly a reserved
> pin that is tied to the opposite polarity on each device. I would
> suggest printing out the portion of the datasheet with the pin info
> and having someone not directly involved with the project eliminate
> all of the identical pins [make sure that every power and ground pin
> is checked individually]. Then you can investigate the rest - one by
> one. This will take some time, but in the absence of specific
> compatibility info, it seems like the best route.
>
>> If not, what points
>> should be taken into consideration?
>
> 4. Do you have a working DM640 board??
> If yes, you could compare the SDRAM signal timing.
> 5. If you are always reading 1s from SDRAM, write 0s and check the
> data bus for 0s. With a constant 'read loop' from SDRAM, check the
> data bus [and control signals] with a scope. If you see 1s, something
> is up with SDRAM, if you see 0s, look at the DM642.
> mikedunn
>>
>> --
> www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
Please copy the group on all posts.
On Thu, Aug 21, 2008 at 1:25 AM, kalai vani wrote:
> Mike,
> Thanks for your inputs. We tried out accordingly.
> We checked the SDRAM control signals on our DM642 board with our
> working DM640 board. The SDCAS, SDRAS and WE signals look similar. But then
> when we probed the data lines, we found some anomaly. When we wrote 1's to
> all 32-bits, the voltage levels of the some SDRAM data lines was1.2V instead
> of 3.3V. When we wrote 0's to all 32-bits some SDRAM data lines had voltage
> levels of 1.2V as against 0V.
> Any suggestions?
I assume that something is connected or shorted to the core voltage [I
assume is 1.2/1.4v].
Do the 1.2v data lines look like a power supply voltage or do they 'toggle'??
You need to be very precise about which pins are working correctly and
which are at 1.2v. The signal names/pin numbers could be a clue.
Have you checked the DM640 vs. DM642 pin assignments??
I am currently thinking that -
[1] The DM640/642 are not interchangeable
or
[2] There is a problem with the workmanship on the installation of the
DM642. Can you get the board xrayed?? [That would show an alignment
problem and some types of shorts].
mikedunn
> Thanks,
> Kalaivani.S
>
> --- On Tue, 8/19/08, Michael Dunn wrote:
>
> From: Michael Dunn
> Subject: Re: [c6x] DM642 SDRAM issue
> To: "Kalaivani. S"
> Cc: c...
> Date: Tuesday, August 19, 2008, 8:34 PM
>
> kalai,
>
> On Mon, Aug 18, 2008 at 9:55 AM, Kalaivani.S
> wrote:
>> Hi All,
>> We have a DM640-based custom board with CPU operating at
>> 400MHz. This board was working fine. Now we have replaced the DM640
>> with DM642-720 chip. The input PLL clock is same, i.e. input PLL clock
>> is 33MHz and DM642 CPU is also running at 400MHz. Is this fine? We are
>> trying to access SDRAM using the same GEL file as used for DM640 board.
>> Since we have not changed any clock inputs, we are using the same EMIF
>> settings. But SDRAM is not able to retain the data written to it.
>
> 1. Have you checked all of the SDRAM control signals?? If you soldered
> the DM642, you could have affected another device on the board.
>> Please let me know if there are any points that we are overlooking?
>
> 2. Are the devices the same revision??
> 3. Was you DM640 in a socket??
> If not, approach this like a new board bringup. You could have a
> solder problem with the new device. Check the power and control
> signals on the SDRAM - even though the board was working, "make NO
> assumptions!'.
>> Is
>> a direct replacement of DM640 by DM642 possible?
>
> I do not know the answer - this sounds like a question for TI to
> answer. I would expect that they would have an app note, although I
> did not see one.
> The devices do appear to be 'highly similar'.
> Make sure that you check all of the relevant pins for their
> connections. Sometimes similar devices have 1 or 2 pins that are
> different - this may be an additional ground or possibly a reserved
> pin that is tied to the opposite polarity on each device. I would
> suggest printing out the portion of the datasheet with the pin info
> and having someone not directly involved with the project eliminate
> all of the identical pins [make sure that every power and ground pin
> is checked individually]. Then you can investigate the rest - one by
> one. This will take some time, but in the absence of specific
> compatibility info, it seems like the best route.
>
>> If not, what points
>> should be taken into consideration?
>
> 4. Do you have a working DM640 board??
> If yes, you could compare the SDRAM signal timing.
> 5. If you are always reading 1s from SDRAM, write 0s and check the
> data bus for 0s. With a constant 'read loop' from SDRAM, check the
> data bus [and control signals] with a scope. If you see 1s, something
> is up with SDRAM, if you see 0s, look at the DM642.
> mikedunn
>>
>> --
> www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
--
www.dsprelated.com/blogs-1/nf/Mike_Dunn.php
Reply by ●August 25, 20082008-08-25
The last time I checked, the DM642 was _not_ pin-compatible with the DM640/1. I
recall there were some differences in the video port area, but probably other
pins as well. You'll have to take a close look at the datasheets.
-Shyan
-Shyan
Reply by ●September 21, 20082008-09-21
Dear All,
We successfully replaced DM640 using DM642 with no changes at
all. We found that the issue was with the DSP soldering. We
resoldered DM642 on a fresh board and it worked like magic. Thanks a
lot for all your inputs.
I just wanted to share my experience with the migration from
DM640 to DM642. We had a DM640 board working at 400Mhz and SDRAM at
100MHz. We then had to upgrade to DM642-720MHz. We read from the
datasheet that the DSP Core voltage must be 1.4V. We did this and
this was the only change we did. If your board worked well for DM640,
it would work well for DM642 as well with the core voltage changed to
1.4V.
We then noticed that the Video port pins were different on
DM640 and DM642. That is video data pins VP0-7 on DM640 is VP2-9 on
DM642. We then digged up the C6x Video Port Reference Guide (spru629)
and to our delight we found that Video port is 8-bit on DM640 and
uses data pins VD0-7 and DM642 has a 20-bit Video port with data pins
VP2-9 used for 8-bit Video capture/display. Hence DM640 and DM642 are
actually pin-compatible if you use 8-bit for video and not use HPI
and PCI.
Infact we did not even change any GEL file settings. Thanks
again for all your inputs.
Thanks,
Kalaivani.S
We successfully replaced DM640 using DM642 with no changes at
all. We found that the issue was with the DSP soldering. We
resoldered DM642 on a fresh board and it worked like magic. Thanks a
lot for all your inputs.
I just wanted to share my experience with the migration from
DM640 to DM642. We had a DM640 board working at 400Mhz and SDRAM at
100MHz. We then had to upgrade to DM642-720MHz. We read from the
datasheet that the DSP Core voltage must be 1.4V. We did this and
this was the only change we did. If your board worked well for DM640,
it would work well for DM642 as well with the core voltage changed to
1.4V.
We then noticed that the Video port pins were different on
DM640 and DM642. That is video data pins VP0-7 on DM640 is VP2-9 on
DM642. We then digged up the C6x Video Port Reference Guide (spru629)
and to our delight we found that Video port is 8-bit on DM640 and
uses data pins VD0-7 and DM642 has a 20-bit Video port with data pins
VP2-9 used for 8-bit Video capture/display. Hence DM640 and DM642 are
actually pin-compatible if you use 8-bit for video and not use HPI
and PCI.
Infact we did not even change any GEL file settings. Thanks
again for all your inputs.
Thanks,
Kalaivani.S
