понеделник, 19 септември 2016 г., 20:06:13 UTC+3, Hans de Goede написа: > > Hi, > > On 19-09-16 18:07, TsvetanUsunov wrote: > > Hi, > > > > We make our final touch of A64-OLinuXino PCB and there we add option > eMMC Flash to work on dual voltages 1.8V and 3.3V. > > The eMMC is connected to AXP803 pin.34 GPIO1/LDO. The problem is that > when A64 boots and AXP803 is not initialized it outputs default 0.8V then > after initialization driver takes care to drive it 1.8V or 3.3V. > > This makes impossible to boot from eMMC which is not good. We now think > for solution which to drive eMMC at 3.3V initially when AXP803 output is > below 1.8V but this adds unnecessary hardware complexity. > > For hardware point of view it will be much more simplier if dedigated > A64 GPIO is used and initially is pulled down and after AXP803 is > initialized is pulled up. > > Ok, so what your suggesting is: > > axp803-ldo-io1 -\ > [mux]---> mmc-supply > Fixed-3.3v ------/ | > | > |mux-control > A64 gpio out--------/ > > Note the above ascii-art requires a fixed-width font. > > With a pull-down (or pull-up) to fix the mux in a certain position when > the gpio is in tri-state ? > > As long as we pin the axp803-ldo-io1 at 1.8v then the Linux regulator > framework should be able to deal with, and in u-boot we can just > keep things at 3.3v. > > > How would you suggest us to implement it? Will this additional GPIO > create troubles in eMMC driver philosophy? > > For the Linux mmc driver the mmc-supply is abstracted as a regulator, > and the regulator framework should be able to deal with any setup > you can come up with. > > > For the SDMMC we are still hesitating what to do as we don't know if the > card which will be inserted will support low voltage and higher speeds at > all. > > As long as you default to 3.3v then the kernel's sd subsystem can > dynamically switch voltage (through e.g. the gpio) if the card > advertises it supports low voltage. Note that you're planning > the first board to implement this that I know off, so the sunxi-mmc > kernel driver will need some work to support voltage switching, > but in the mean time things should work fine at 3.3v. > > > Also eMMC Flash and SDMMC card should be driven by separate voltages, as > they may work in any combinations. > > Ack, right, as said both cards should come up with 3.3v and then > a new voltage will be negotiated before switching, so this definitely > needs to be per card. > > > This means we need another AXP803 LDO and another GPIO for the SDMMC > card. > > Right
Micron eMMC chips we use do not support higher clock at lower voltage, so the way we wired the schematic right now makes no much sense. I check for other vendors but also can't find such eMMC chip, if someone knows please let us know to investigate more? So in this case makes sense to move the dual voltage supply to the SD-MMC card only but this rise some more issues :) The card is currently wired to port F which Vcc is internally connected together with port B and H where is WiFi SIDO , I2C UARTs etc which will be lost if we power with 1.8V, so no go. We can swap the SD-MMC and eMMC ports, port F and port C, but in this case we will lose the NAND Flash option i.e. the possibility to run Android. I still can't find SD-MMC card which to work on 1.8 and 3.3V can you point me to some model so we perform test and see if this is really good to have feature, or we will cut this and wire 3.3V permanently :) Tsvetan -- You received this message because you are subscribed to the Google Groups "linux-sunxi" group. To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
