This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

External memory on a XC16x Board

Hi Everybody,

I'm still experimenting with Infineon's XC161 Starter Kit using the Keil C166 compiler, v4.27 (linker version 4.28). Almost everything I need works OK now, except the external bus controller. Below I'll describe my situation:

Hardware
========

I've connected the RAM to the processor by setting switch S101.8. This is the configuration for the multiplexed mode, so I also switched S103.1 and S103.2 off for 16 bits multiplexed mode.

The CS-lines are enabled by off-switching S102.6 and S102.7. I use 4-bit address lines by setting S102.4 and S102.5.

S102.8 is switched on to use /WR and /BHE as /WRL and /WRH.


Software
========

In the startup file (START_V2.A66) I configured the EBC as follows:


$SET (CONFIG_EBC = 1)
_SAPEN      EQU    4
_CSPEN      EQU    5
_ARBEN      EQU    0
_SLAVE      EQU    0
_EBCDIS     EQU    0
_WRCFG      EQU    1
_BYTDIS     EQU    0
_ALEDIS     EQU    0
_RDYDIS     EQU    0
_RDYPOL     EQU    0
_APDIS      EQU    0
_DHPDIS     EQU    0


CS1 as follows:


$SET (CONFIG_CS1 = 1)
_ADDR1      EQU    0x400000
_SIZE1      EQU    256*KB
_ENCS1      EQU    1
_RDYEN1     EQU    0
_RDYMOD1    EQU    0
_BTYP1      EQU    3
_PHA1       EQU    3
_PHB1       EQU    1
_PHC1       EQU    3
_PHD1       EQU    1
_PHE1       EQU    25
_RDPHF1     EQU    3
_WRPHF1     EQU    3


Maximum bus times to control the RAM, so it isn't a timing problem. In the target options (under 'External Memory") I set the following values for #1:

Start: 0x400000
Size : 0x40000

In my C-code I also set the P6 pins to output.

I've looked over my problem several times but couldn't trace the things going wrong. I also tried other memory configurations (8 bit mux, 8 bit demux and 16 bit demux) but all of them with no success.

Final thing to tell about are the contents of my .M66-file. All data is located in valid areas (I can see vars in the 0x400000 area), but one line is very strange:


  VALUE       PUBLIC SYMBOL NAME               REP   TGR  CLASS   SECTION
=======================================================================
-- cut --
  3F8000H     ?C_NDATABASE                     CONST ---  ---     ---
-- cut --


I don't know if 0x3F8000 should be an address, that is used somewhere else, but IF so, then it's not valid. I didn't configure this value anywhere; it's generated by Keil's Linker.

Well, phiew, that was a lot of writing, so I hope somebody can help me? :)

Kind Regards,
Dennis Ebben

Parents
  • Hi Dennis,

    yes you can only connect CS0 and CS1 at the starter kit board. And CS1 may be, depending on your board version.
    This because the older versions were shiped out with FLASH and SRAM as external memory. Newer versions come only with SRAM and are connected by default to CS0.
    All other setting are irrelevant, without hardware modifying (since the CS lines are wired in parallele way .
    On the other side there is a nice SPI-EEPROM at the solder side, which not many peoples have recogniced. I wrote little bit code for it to read from and write to it.

    SRAM is working in 16-bit demuxed mode allways (FLASH versions to) and you have set the
    external RAM (Options for Target)
    0x10000 Size=0x20000.
    I wrote a short code too, to test the SRAM.

    I have documentation for the board, if you not found on your starter CD-ROM.

    Concerning the NDATA question I can not answer without have a look at your C-code and the MAP-file.
    It is in linker/locators responsibility, where the code would be located.
    It depends from your memory modell.
    May be the DPP's are changed or the segment pointer is modified.
    The most simple way I think is to force the locator to lay it in a memory area, which really is in use.
    For more information I need some details more.

    Stefan

Reply
  • Hi Dennis,

    yes you can only connect CS0 and CS1 at the starter kit board. And CS1 may be, depending on your board version.
    This because the older versions were shiped out with FLASH and SRAM as external memory. Newer versions come only with SRAM and are connected by default to CS0.
    All other setting are irrelevant, without hardware modifying (since the CS lines are wired in parallele way .
    On the other side there is a nice SPI-EEPROM at the solder side, which not many peoples have recogniced. I wrote little bit code for it to read from and write to it.

    SRAM is working in 16-bit demuxed mode allways (FLASH versions to) and you have set the
    external RAM (Options for Target)
    0x10000 Size=0x20000.
    I wrote a short code too, to test the SRAM.

    I have documentation for the board, if you not found on your starter CD-ROM.

    Concerning the NDATA question I can not answer without have a look at your C-code and the MAP-file.
    It is in linker/locators responsibility, where the code would be located.
    It depends from your memory modell.
    May be the DPP's are changed or the segment pointer is modified.
    The most simple way I think is to force the locator to lay it in a memory area, which really is in use.
    For more information I need some details more.

    Stefan

Children
  • Hi Stefan,

    Thanks for your response.

    On the board I can see the following text: "XC16Board REV 200". This board is shipped without flash. There is an ability to place the flash components anyway.

    A few days ago, I also set my memory settings to [start: 0x10000] and [size: 0x40000]. After that, the board started correctly and everything seems to work OK. This morning I write some vars to the RAM and the returning value also seems to be OK. So I think it works. But it's still weird, I only succeeded doing this on /CS0 and not on /CS1.

    Why did you mention the SPI-EEPROM? Any special reason?
    Talking about EEPROM's: I'm using FRAM's from Ramtron (FM24C256). Used them before and very satisfied with them.
    If you don't know them: They are pin-compatible with 24xxx-EEPROM's, controlled by an I2C-bus, and can be written/read as much as you like. :)

    Does the RAM only work in 16 bits demux?? I thought the board could work in all 4 modes. When I look to the wiring in the schematics of the board, I would say this shouldn't be a problem. My board was shipped with a CD-ROM and I do have the schematics and also a Hardware Manual.

    I don't know what it is with the weird NDATA-values, but they appear everytime I configure external RAM. If I don't configure external RAM, the values is 004000H. I think that's also an undefined area, cause it isn't in internal processor RAM.

    Here's a dump of the weird value when I DO configure external RAM:

          C00000H     ?C_NCONSTBASE                    CONST ---  ---     ---
          0000C0H     ?C_NCONSTBASESEG                 CONST ---  ---     ---
          008000H     ?C_NDATABASE                     CONST ---  ---     ---
          000000H     ?C_NDATABASESEG                  CONST ---  ---     ---
    

    "?C_NDATABASE" has value 008000H now. That's outside my RAM-area, which starts at 0x10000. When I configure the RAM to 0x400000 (instead of 0x10000) "?C_NDATABASE" has the value 3F8000H. It's always: [start-address] – 0x8000. That's very strange, because that's almost always an undefined area.

    Fortunately "?C_NDATABASE" isn't used anywhere (as far as I can see). So I don't care about the weird value anymore.

    Kind Regards,
    Dennis Ebben

  • Hi Dennis,

    there was no special reason for the SPI_EEPROM, but I have metioned it, because, with only RAM and one LED on the board you have not so many options to "start up".
    Now I use external hardware too.

    The boards as shipped out are working with
    16 bit demultiplexed bus.
    CS0 is wired to the SRAM and CS1 to the (optional)FLASH.

    
    C00000H     ?C_NCONSTBASE    CONST ---  ---
    0000C0H     ?C_NCONSTBASESEG CONST ---  ---
    008000H     ?C_NDATABASE     CONST ---  ---
    000000H     ?C_NDATABASESEG  CONST ---  ---
    

    This all are internal variables from the linker.
    ?C_NCONSTBASE address for calculating and linking locating is in this case C00000, each step this will increased to show you correct lines, addresses and so on.

    ?C_NCONSTBASESEG is the value for the base code segment
    in your case it is C0
    - you find it in C0'0000h (C0 is the segment pointer)
    ?C_NDATBASESEG is 000000 what means it shows to segment 00'0000 which is the systemsegment, where you find the internal RAM (with DATA's) and the sfr's.
    The DATABASE I do not know , what could it mean exactly, but it is used for calculating something with datas.
    But definitly you have not to take care for it, since these are internals.

    (Thanks for your memory hint!)

    Stefan

  • Hi Dennis,

    me again.
    I found out what the ?C_NDATABASE means.
    Have a look at system manual on memory organization and find in segment 0 that the
    area from 0x0000 to 0x8000 is external memory area.
    0x8000 is a base adresse for calculating addresses of internal RAM and SFR for linker MAP file.

    Stefan

  • Hi Stefan,

    Yes, you're right about the starter kit. The kits that have been made in the past (by Phytec) all had more I/O on board, like I2C-components, more LED's, etc. I'm missing these things on this board. It's indeed not easy starting up, this way.

    I had looked in the system manual already and saw the area you mentioned. But that still doesn't explain the 3F'8000 value for ?C_NDATBASESEG in some cases, because there's nothing on that address. No memory, no I/O. Don't know why Keil creates this value. Because my software works OK, in spite of these values, I've decided to ignore them.

    Stefan, thanks again (second time now) for all your help and time. Things are clear to me again. :)

    Kind regards,
    Dennis

  • Hi Dennis,

    thanks for this, but no matter.
    OK I found out what is the meaning of
    ?C_NDATABASE and ?C_NDATABASESEG.

    I configured my program as yours regarding the hardware.
    For the following it can be usefull to have a look to system user manual section 4.5. Data adressing.

    As you know the code is accessed by segments
    ( a 64kBytes each ) and data memory with pages ( a 16k each ).
    So you have for pages per segment.
    First you should ensur that your
    ?C_PAGEDPPx are the following values:

    000000H         ?C_PAGEDPP0
    000301H         ?C_PAGEDPP1
    000100H         ?C_PAGEDPP2
    
    I try to explain here:
    
    3F'8000h is
    0011 1111 1000 0000 0000 0000
    where       00 0000 0000 0000 is Page offset
                                  (14-bit)
              10                  selects DPP2
    in DPP2 you find value 100    (10-bit)
    01 0000 0000
    now we will assemble the real address
    (it results in 10-bit DDP2 value and 14-bit
     page offset)
    
    0100 0000 0000 0000 0000 0000 ( 24 - bit )
    in
    HEX it is 40'0000h your NDATA area in RAM !
    
    Thats where the DPP2 points by default after startup.
    Let's do for DPP1 (301h)
    
    11 0000 0000 ( 10-bit )
    
       01 00 0000 0000 0000
       01 is DPP1
          00 0000 0000 0000  is Page offset.
    for Segment : Offset
        8-bit   : 16-bit
      1100 0000 : 0100 0000 0000 0000
       C    0      4    0    0    0
       C0'4000
    
    the DPP1 points to you MAIN routine, you
    find in map file as
    C04000h    ......   ?NC?MAIN.
    
    OK, how to calculate the obscourious
    startaddress ?
    40'0000h is your selected address for RAM.
    DPP0 0-16K
    DPP1 16-32K
    DPP2 32-48K
    DPP3 48-64k ( in each segment ).

    Easy subtract 32k ( 8000h ) for DPP2 from
    40'0000h and you get 3F'8000h.

    DPP0 theoretical (really points to segment0)
    3F'8000h - 3F'C000h
    DPP1 theoretical (really poinst to C0'4000h)
    3F'C000h - 40'0000h
    DPP2
    40'0000h - 40'4000
    ....

    Best regards
    Stefan

  • Hi Stefan,

    I never really understood the DPP's. I think that it's really clear now (also to other readers of this thread). Thanks again. :)

    What version of the datasheets do you use?? You mentioned chapter 4.5 (which is correct in version v1.1, feb 2002). Since the release of that manual there have been released 2 new versions:

    * v2.0, 3-2003
    * v2.1, 6-2003

    The last one is the one I use now. In this version it's chapter 4.7 that handles about data addressing.

    There are still some things that cannot be found in the new manuals, like the ADC-bug that's also known from the XC164 (ADBSY is set after x clock cycles, not immediately after starting an ADC-conversion).

    Kind Regards,
    Dennis

  • Hi Dennis,

    cool that you understand now the DPP's.
    It is really not complicated.
    I need at my start two times and than I calculated three examples and now I have it.
    ( If questions concerning to this, do not hesitate to ask me )

    I used the XC167 manual for this, since the register mappings are similar.
    So if you write a program for a module, which is available on the other XC16x too,
    you can be sure in 98% of cases it will run there.
    So at XC161 the CAPCOM6 will not work and at
    XC164 the I2C module.

    Bugs, problems and workaround (if any) you can find in the errata sheets.
    Please have additional a look of the step code of your device.

    But you are right with the manuals specially for XC161.
    There in the new version is the complete description of SCU missing.
    I reported to Infineon and it will be updated with next release.

    Stefan

  • Hi Stefan,

    I know that the other XC16x devices are almost the same, but it's weird that there isn't any link on the site on the XC161 page to errata sheets of the other (similar) devices.

    Also the starter kit isn't described on the Infineon site. The info about this controller and applications are really poor. But perhaps this will change in the future, for it's a new controller.

    Does Infineon also have a forum, like Keil? If so, maybe we can discuss this there. :)

    Kind Regards,
    Dennis

  • Hi Dennis,

    Infineon unfortunately has no discussion forum, so we have to discuss here, or we can do it by mail.
    The kit indeed is only described at the
    CD-ROM you find at the kit.

    But I wonder, that the eratas are not on the web, since I got them from there.

    I can mail you the last I have, if you are interested.

    Stefan

  • Hi Stefan,

    I've only found XC164 and XC167 errata sheets. No XC161. Searched for them again a few minutes ago, but still couldn't find them. :S

    Kind Regards,
    Dennis

  • Hi Dennis,

    I send you my Erratas concerning Xc161CJ
    step AA,AB and AC.

    Please check your account.

    Stefan

  • Hi Stefan,

    I received your mail with errata sheets. I really can NOT find them on Infineon's site. Even when I log in to My Infineon I still can't find them.

    Seems like you have more permissions to view/download files than I have. :)

    Thanks a lot for de documents. I'll read them tomorrow morning.

    Kind Regards,
    Dennis

  • Hi Dennis,

    it is not your mistake. At the moment there are really no erratas available for XC161CJ.
    I got them some time before and save them all for such times. ;-}

    Stefan

  • Hi Dennis,

    I did a mistake.
    The board works with 16-bit muxed bus.
    Not demuxed.

    Sorry for this,
    Stefan

  • Hello Stefan,
    I've red the thread very carefully, but at the moment I'm a little bit confused about the right configurations for accessing the external RAM.
    I've an XC164 Starterkit, Rev. 200 without FLASH and I got the RAM not working yet. Do you have an working configuration and because everything depends from everything, I need the DIP Switch configuration, the right settings in Start_A2.a66, the Target configuration and if I have something to code in main(),like special PORT configuration, it was also interesting, to get it.
    Thank you very much in advance
    Steffen