See blow. Gerald
On Mon, Oct 14, 2013 at 9:21 AM, Alberto Potenza < [email protected]> wrote: > Dear Gerald, > I was reading again your answers and there are some points not fully clear. > > In the "BeagleBone Black System Reference Manual" there is written: > "GPMC bus may NOT be available due to the use of those signals by the > eMMC." (pag. 37). > That is correct. You need to disable the eMMC and boot from uSD to access those signals. > > At pag. 61 there is a schematic view of the connection between AM3358 and > the flash. Now, only 10 MMC1 connections are used (from MMC1_DAT0 to > MMC1_DAT7, MMC1_CMD and MMC1_CLK). > By using the Schematic A6 file (here > http://circuitco.com/support/index.php?title=BeagleBoneBlack), at pag.3 > it is possible to see that those MMC1 connections correspond to GPMC_AD0 -> > GPMC_AD7, GPMC_CSN1 and GPMC_CSN2. > Summarizing: > MMC1_DAT0 -> MMC1_DAT7 == GPMC_AD0 -> GPMC_AD7 > MMC1_CMD == GPMC_CSN1 > MMC1_CLK == GPMC_CSN2 > > Correct. > Now, if we take again the "BeagleBone Black System Reference Manual" at > pag.82 there is written how it is possible to set the connector P8. We can > suppose to be in mode 0 (which is different from the mode 0 boot mode...I > wrote it just to avoid confusion). > In that connector I have again GPMC_AD0 -> GPMC_AD7 but I cannot used them > because they are already sent to the flash (I guess...). > Incorrect. See above. > However, I still have some GPMC lines that, in principle, I could use > (GPMC_AD8 -> GPMC_AD15) to send data. > Accoridng to the "AM335x ARM® Cortex™-A8 Microprocessors (MPUs)" TRM, GPMC > AD lines are for both address and data while GPMC A lines are for address > only. > Consequently, I could set an address by using: > GPIO1_16 (which is GPMC_a0) > GPIO1_17 (which is GPMC_a1) > EHRPWM1A (which is GPMC_a2) > EHRPWM1B (which is GPMC_a3) > Moreover, I have a line GPIO2_1 (which should be a clock -> gpmc_clk_mux0) > that I could use as clock and 8 GPMC lines to send data (GPMC_AD8 -> > GPMC_AD15). If I have an external memory which does not need an address to > send out data, I could set fake a fake address on the GPMC_A lines and send > real data on the 8 GPMC AD lines. > Is that right or, as soon as I use some GPMC lines for the eMMC, the full > GPMC bus becomes inaccessible? > Moreover, does the GPMC bus work with 8-bit data or, according with the > description in the TRM, it requires only 16 bit data (because the ARM takes > always GPMC_AD0 -> GPMC_AD15, but the first lines of the bus are connected > to the eMMC and it could mess up everything)? > > Now, we can suppose that I do not want to use the eMMC boot mode. In this > way I should free the GPMC. If I set the connector P8 in mode-1, at pag.82 > of the "BeagleBone Black System Reference Manual" I can see GPMC_a for > addressing (from a0 to a19) but I do not see GPMC_ad for data. Where my > data goes? > Address and data are multiplexed onto the same pins. Use the GPMC_ALE signal to latch the Address. > Also here, where I am wrong? > Now, if I am in mode-1 and I want to communicate with an external device > (it could be a flash memory or a sd card or something else) by using the > mmc1_dat lines (from 0 to 7) how can I do it, because I do not see the > mmc1_cmd signal and the mmc1_clk signal. Now, why there is the possibility > to send out the mmc signals without a clock for them? Am I missing > something also here? > > > Thanks again for your help. > > Alberto > > > > >> >> >> >> Depends on your definition of HIGH. GPMC is 100MHZ and up to 16Bits wide. >>> >>> Gerald >>> >>> >>> >>> On Fri, Oct 11, 2013 at 9:58 AM, Alberto Potenza >>> <[email protected]>wrote: >>> >>>> Look, >>>> I should have 32-bit data generated with a frequency f =100kHz. >>>> Which means a data throughput of 3.2Mb/s. >>>> With SPI I should have the possibility to work with a data throughput >>>> of 48Mb/s (which is more than 10 times faster). >>>> I already know your question: >>>> "But in the previous post you said that you needed GPMC for high data >>>> throughput...but it seems to be not...why?" >>>> Because I will probably will asked to manage an higher data throughput >>>> in a couple of years, so I decided to save time now... >>>> >>>> But right now it seems that SPI is the best compromise between >>>> performance and realization-time. >>>> >>>> Thanks a lot for your effort. >>>> >>>> Regards. >>>> >>>> Alberto >>>> >>>> >>>> >>>> >>>> >>>> >>>> Il giorno venerdì 11 ottobre 2013 14:26:36 UTC+2, Gerald ha scritto: >>>> >>>>> Is that going to be fast enough for your application? >>>>> >>>>> Gerald >>>>> >>>>> >>>>> >>>>> -- >>>> For more options, visit http://beagleboard.org/discuss >>>> --- >>>> You received this message because you are subscribed to the Google >>>> Groups "BeagleBoard" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to beagleboard...@googlegroups.**com. >>>> For more options, visit >>>> https://groups.google.com/**groups/opt_out<https://groups.google.com/groups/opt_out> >>>> . >>>> >>> >>> -- > For more options, visit http://beagleboard.org/discuss > --- > You received this message because you are subscribed to the Google Groups > "BeagleBoard" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > -- For more options, visit http://beagleboard.org/discuss --- You received this message because you are subscribed to the Google Groups "BeagleBoard" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
