From: Gil Y. <Gil_Yetter@SensArray.com> - 2006-01-18 20:51:31
|
I have a desire for the RTC (real time clock) to continue without an Epoch time reset when power is interrupted. For an application requiring that power be disconnected during relocation of the product (with the Gumstix inside) the RTC needs to continue. There may not be a Ethernet port or other means to obtain time to set / sync the internal RTC on each power up. The Gumstix Connex 400xm, PXA255 would require a 2nd 32.768 crystal oscillator for the lowest power mode of operation and a power management scheme to power down and ground 2 of the CPU power supply terminals is needed. The current Connex 400xm board does not have the second crystal and it is not clear or obvious that one could just be added. The power management circuitry to remove power from 2 of the 4 powered connections (VCC and PLL_VCC ) to the CPU is probably not there as best my eyes can see. The use with the single 3.6864 MHz crystal which is divided by 112 (32.914 kHz) and the divider trimming provides the RTC time base but at a performance cost of not being able to fully power down into sleep mode. One possible alternative or workaround to get close to the desired sleep mode with the RTC continuing to operate might be (and here is where I need some help): to rebuild the STUART board with additional circuitry needed for my application but to also add a RTC clock chip such as from Maxim, a coin cell battery, 32.768 crystal and some voltage monitoring detection circuitry. With the addition of a small battery to maintain operation at 3.3 vdc (VCC * VCCQ* VCCN* VCC* PLL_VCC) which is lower than the externally supplied 4.0 (or higher) power supply voltage ( and presumably in parallel in some fashion) then when the external voltage drops below the expected level a signal passed to the CPU via a GPIO line can signal any need for a graceful power down (if needed). Code would need to be added for the CPU to check to see if the Epoch time "midnight 1/1/1970 GMT" is set or observed on power up and then to either obtain the RTC time (set earlier by the CPU) from the Maxim RTC chip on the new version of the STUART board or from an Ethernet connection if that port in online. Using this scheme the CPU can go down hard. Question: Does this idea seem rational ? Question: Are there any other schemes that I can examine for consideration? The Future Gumstix Connex ?? with a PXA 27x? may have a far better capability to achieve this. The 2nd oscillator ( 32.768 kHz) is still needed and a way needs to be available to just solder one onto the new yet to be seen Gumstix board if one is not provided (to reduce cost). The PXA27x has a clean power monitoring and switchover to battery back up coin cell circuitry design in the Intel provided documentation.. figure 3-1. Typical battery and external regulator configuration, page 11:3-5 of the Intel PXA27x Processor Family Design Guide , this design places the CPU into deep sleep mode and continues to operate the RTC. The resumption of operation after power is re-established seems to have been addressed with a nREST line from U11. Question: to what extent is the Intel provided method going to be incorporated in the upcoming new Connex board? Question: will the new connex board with the PXA27x be pin compatible with the current accessory / expansion boards (i.e.. Net CF, STUART, etc) ? |