• State TI Thinks Resolved
  • Locked Locked
  • Replies 9 replies
  • Answers 3 answers
  • Subscribers 65 subscribers
  • Views 1058 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

Oscillator start up timing spec for external osc. feeding HFCLKIN on TPS65920

Jeffs_Fairway
Intellectual 290 points
Other Parts Discussed in Thread: TPS65920, TPS65930

I stumbled across something on page 93 in the data manual for the TPS65930/TPS65920 concerning oscillator startup timing for the external oscillator feeding the HFCLKIN pin.  When using the CKEN pin as the enable for the oscillator, the timing diagram gives a maximum of about 5.2 to 5.3mS for the oscillator to start.  Specs for all of the CMOS 1.8V oscillators I've come across so far give 10mS as the startup time.  Even if the enable pin is tied to VIO (as on the BeagleBoard) that still only gives it about 6.5mS before HFCLKOUT is expected to be stable.  So what am I missing here in the startup timing, or have you lot all been using some magic oscillators that start within 5mS?  Or is everyone just rolling the dice and keeping their fingers crossed?

Regards,

-Jeff

  • 0
    TI__Genius 15100 points

    Hi Jeff,

    This requirement is necessary because OMAP may not function properly if the HFCLK is not provided before reset goes high. I have seen some customers adding delay on reset so the clock is stable before the reset signal goes high. 

    Here is one oscillator with 3ms start-up time - SG-210 series from Epson Toyocom.

    Hope this helps.

     

    Regards,

    Gandhar.

     

  • 0 in reply to Gandhar Dighe
    Intellectual 500 points

    Hi Gandhar,

    Do you have another oscillator with a proper startup time ?

    I cannot use Epson Toyocom components, and all the oscillators I have found from others manufacturers have a 10ms startup time.

    Thanks for your help

    Philippe

  • 0 in reply to Philippe MAIRE
    TI__Genius 15100 points

    Hi,

    I think there are two on the beagle board design - http://beagleboard.org/hardware/design

    You can refer those.

     

    Regards,

    Gandhar.

     

  • 0 in reply to Gandhar Dighe
    Intellectual 500 points

    Hi Gandhar,

    On Beagleboard two oscillators are proposed:

    - SG-210 which I cannot use

    - EP16E7E2H from Ecliptek, which has a startup time of 10ms !

    http://www.ecliptek.com/oscillators/ep16e7/

     

    I guess the maximum startup times are a bit exagerated and the typical value is probably less than 5.3ms, but the boot is not guaranteed...

    If you ever hear of another reference with guaranteed startup time less than 5.3ms, please let me know.

    Regards

    Philippe

  • 0 in reply to Philippe MAIRE
    Intellectual 290 points

    Hi Philippe,

     

    On Beagleboard I think they are just keeping their fingers crossed and hope it will work for most situations.  Obviously this is not the answer for production.  Statek makes a 1.8V oscillator that has a guaranteed startup time of 5mS maximum, but they were so expensive (like USD$12 in quantity) that I instead used a 2.8V oscillator from NDK, part number NT3225SA-19.2M-DJA3002A.  2.8V, 19.2MHz, with guaranteed startup of 5mS maximum.  I used a little 2.8V LDO regulator from Micrel (MIC5232-2.8YD5) to power it.  Hopefully adding the LDO doesn't significantly add to the startup time.  But it seems at present the choice of oscillators in 1.8V is quite limited.

    Regards,

    -Jeff

  • 0 in reply to Jeffs_Fairway
    Intellectual 500 points

    Hi,

    I can't afford another LDO because of space constraints...

    For reference, I have found another reference from Abracon: ASE3 – 26.000 – L – K. (The startup time is not in the datasheet but has been confirmed to me by the technical support)

    Unfortunately, the lead time is very long, and incompatible with my production calendar...


  • 0 in reply to Philippe MAIRE
    Genius 3685 points

    Are there any special reason not to generate the HF clock from the OMAP using a crystal instead of an external oscillator connected to TPS659xx? I know there was a problem with this approach initially (chip rev 2.1 and earlier - Advisory 2.1.1.120 Peripheral Boot Issue When Using External Crystal), but isn't this fully fixed?

    I as well know that most designs use the external oscillator approach, but would like to hear if there is any valid reason no to use a crystal connected to OMAP3 instead, since I think this would save both cost, space and power? Are the external oscillator design approach just an old left over from the original working designs (due to the old problem :-) or are there still valid reasons to keep using this method? I currently don't see any?

    Best regards and thanks for clarifying
      Søren

  • 0 in reply to Soren Steen Christensen
    TI__Genius 15100 points

    Hi Soren,

    I can answer this is some cryptic way as we do not a platform that we have tested this exhaustively.

    "Ideally" nothing stops one from using the crystal. OMAP will support a 19.2MHz crystal which will be ok for the device.

    All TI platforms and probably software were always built based on an oscillator approach. There may be some customers who would have tried this low cost method, however, I am unaware of anyone.

    Since we have not done tests with this topology we do not know if there would be any issues.

    I hope this helps.

     

    Regards,

    Gandhar.

     

  • 0 in reply to Gandhar Dighe
    Genius 3685 points

    Hi Gandhar,

    Thanks for the quick answer. Basically I'm glad that you can confirm my point, that it's supposed to work. I of cause had hoped that somebody at TI had done real testing with this feature, since I think it's (seen from an overall system point of view) a better/cheaper approach than using an external oscillator :-)

    Unfortunately it seems that nobody have apparently done this yet. I would else had hoped/imagined, that some of the mass volume OMAP3 customers had done this in order to bring down the BOM :-). I will let you know in case I take the time to try it out at some point in time...

    Best regards and thanks
      Søren