This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Am1808 Low power mode.

Other Parts Discussed in Thread: AM1808, OMAP-L138

I am trying to understand how AM1808 can be put into lowest power consumption; while still be able to receive Ethernet/internet calls. so to wake/work fully when such call/request comes in and then go back to very low power mode. Any ideas would be very appreciated. thanks in advance.

Yes we can bring down the voltage of the core to 1.0. and going to deep sleep mode; but in that; ethernet mac may not be active. Since the cpu receives the ethernet equest only a few times a day; a good solution has a big effect on the design. Please give your ideas; thanks.

  • Deep sleep mode would probably not work in this situation because it clock-gates the device.  Without a device clock, the EMAC module will not function.  If the AM1808 is placed behind a switch-PHY such that only network traffic destined for the device is passed along, you might be able to tie the DEEPSLEEP pin to the RX_DV signal.  That way the device wakes up when a packet comes in.  Of course, the packet contents will not be read during DEEPSLEEP, but it can ready the device for subsequent communication.

  • Thanks for the reply. the idea of looking at the rx change-state is interesting. I also have another helper-cpu in the vicinity of the am1808. that one is always on  but doesn't have ethernet port.. Can that one do anything? like setting the am18x to sleep and waking it up by hardware signal. and If i look at the rx line's activity; wouldn't the first message that comes from ethernet phy get lost; due to the time the cpu takes to wake up and get ready to receive? 

    I'm also looking into determining about how much power consumption is if we turn off everything except the ethernet?

    this is a 100k yearly volume type product and in today's financial troubles; I dont' want the company closes due to my bad design!

    thanks in advance for your answers.

     

  • asad1 said:
    I also have another helper-cpu in the vicinity of the am1808. that one is always on  but doesn't have ethernet port.. Can that one do anything? like setting the am18x to sleep and waking it up by hardware signal.

    Yes, the helper device can place AM1808 into DEEPSLEEP and wake it up when needed by controlling the DEEPSLEEP pin.

    asad1 said:
    If i look at the rx line's activity; wouldn't the first message that comes from ethernet phy get lost; due to the time the cpu takes to wake up and get ready to receive?

    Yes, the first message would definitely be lost so there would have to be some predetermined protocol to send a junk packet to AM1808, wait for acknowledgement that AM1808 is back from DEEPSLEEP, and then send data packets to AM1808.  If your helper device can operate fast enough to buffer packet data, it's possible that the data from the wakeup packet could be relayed to AM1808 after it has started up again.

    asad1 said:
    I'm also looking into determining about how much power consumption is if we turn off everything except the ethernet?

    Do you have a local FAE contact?  The power numbers can vary significantly depending on various use conditions.  It would be easier to collect the details offline.

  • thanks for yoru answers. I think I'll apply a combination of them. But still need to learn how can I save the ethernet message and then relay it to the am18 after it wakes up.

    I assume that if the helper IC to have an ethernet port/access then it could  cache the ethernet message for the time the am18 wakes up.  I am assuming that all the messanges that come out of the ethernet switch box; are visible by all the Macs that are connected to that switch.

    If  you know of any references, links, ... ; i would appreciate it.

    meanwhile; I'll call the local TI Rep to get a contact to an FAE. I check the distributors like Avnet, arrow; every other day; but no sign of availability! do you have any idea when the AM1808 would be available? LogicPD's saleman; onetime told may end of summer or so.

    anyways;  thanks again and have a nice day..

    .

  • It's possible to have all the MACs share a bus, but that would limit your operation to half-fuplex, which is very slow.  I think the ideal thing to do would be to snoop with the helper device on the MAC bus, but relay the information through another port like SPI, UART, I2C, etc so that you can maintain full-duplex operation with the PHY.

    I'm not aware of any references for this implementation, but I'm sure there are some on the web.

    AM1808 EVM availability at the end of summer sounds reasonable.  They are finalizing the alpha boards right now and should be building up inventory in the coming weeks.  If you have an urgent need of a development platform, the OMAP-L138 has the same ARM9 and a very similar architecture to AM1808.

    -Tommy

  • Asad1,

    If you are doing a very low power design, you should be aware Logic PD released their new Wattson software for the AM1808 EVM which will help you see actual power consumption across all power rails on the CPU and SOM design.  There is an adapter required to connect to the PC available from this website: http://createnewstuff.webs.com/breakoutboardwattson.htm


    The Logic PD software is available from here: http://www.logicpd.com/wattson

    -A

     

  • Hi,

    I want to put my AM1808 based board in low power mode (suspend to RAM mode.)

    But I can not configure the wakeup event on any GPIO based IRQ. It is also not documented any where in the Reference manual.

    So can I implement the some GPIO based wake up signal or I must have to use the hardware DEEPSLEEP signal ?

    Thanks,

    Nirav.

     

  • Nirav,

    Suspend to ram uses the processor deepsleep mode.

    From the processor perspective, deepsleep can exit only with the external DEEPSLEEP signal or via internal RTC alarm.

     

  • Hello Ian,

     

    Is there no other way?

     

    Can we not use the RTC way and poll when to STOP the DEEPSLEEP. Every time the RTC wake up forces a DEEPSLEEP, until the GPIO has been toggled. Is this feasible?

     

    Just wondering. Really not sure if that makes sense. Please advise....

     

    That way I can really wake up on GPIO/keypress. Isn't it?

     

    Best Regards

    Feroz

  • Feroz,

    Technically, I think what you describe might work, but the enter and exit from deepsleep/suspend-to-ram in a linux kernel environment take a considerable amount of time.  In order to realize any reasonable responsiveness, you would spend so much time entering and exiting the deepsleep mode that the current consumption would no longer be close to deepsleep numbers.

    -Ian