• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 79 subscribers
  • Views 1248 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.

MSP430F2252, 32.768KHz Crystal, 6pF load

Bob Bongiorno
Prodigy 30 points

Other Parts Discussed in Thread: MSP430F2252

We are finding what appears to be an oscillator capacitance issue with the MSP430F2252. We used 6pF load crystals that were tested by the manufacturer to be very close to 32.768KHz. with the proper load. After trying a few of these crystals, the processor is adding about 6pF capacitance to each crystal lead in addition to the actual internal capacitance value selected. In other words, to provide the proper 6pF load to the crystal for 32.768KHz operation, the processor oscillator capacitance selection of 1pF internal  + 4pF internal phantom + 1 pF  external capacitance is required.

 We believe it’s not a PCB issue. We tried multiple boards with different layouts including modified board layouts and finally, just the crystal across the processor pins and it all points to the processor. For our case, a 2.2pF cap on each crystal leg, with the processor set for 1pF load capacitance, produces exactly 32.768 KHz. It should require 8pF on each leg, not 2.2pF!

 As long as this characteristic is repeatable and doesn’t vary lot to lot, we can live with it. I’d just like to know what’s going on with these parts. Has anyone brought this up before?

 We are using test code specifically written to change the internal oscillator cap value to 1pF or 6pF and outputting the buffered clock frequency for measurement.  

 We've experimented with setting 1pF and 6pF internal and with and without external caps and it's consistently adding more capacitance than specified.

Thanks...any help would be appreciated.

 Bob B.

  • 0
    Guru 227245 points

    Bob Bongiorno said:
    We used 6pF load crystals

    That measn the capacitance on each crystal side has to be 12pF.


       --------                 --------          
       !      !                 !      !          
       !     ---  12pF          !      !          
      ---    ---               ---    ---         
      ###     !-----GND  -->   ###        6pF load
      ---    ---               ---    ---         
       !     ---  12pF          !      !          
       !      !                 !      !          
       --------                 --------          

    If you select the internal "6pF load" setting, then it already includes the pin capacitance (2pF per pin) and adds two 10pF internal silicon capacitors.

  • 0 in reply to Jens-Michael Gross
    Prodigy 30 points

    Our settings for the processor are for minimal (0pF) internal capacitance. For some reason, we only need to add a 2pF cap on each crystal leg to get exactly 32.768KHz. As I explained on my first post, we have reduced the layout to just the crystal across the part, so it’s not a case of stray capacitance.  I believe the part is adding more capacitance than desired and few designers are using the part for absolute accuracy. In our application, we’re trying to get as close to 32.768KHz as possible. Most users probably don’t care whether the clock is off a few hertz but we do. Here’s a link to an application report from T.I. that suspiciously shows the frequency in their own tests to be off and blames the capacitance of the socket. See pages 5 and 7 and note the frequency error from 32768. If the parts are consistent, we can live with this. I just don’t want to find a variance batch to batch.  http://www.ti.com/lit/an/slaa225/slaa225.pdf

  • 0 in reply to Bob Bongiorno
    Guru 227245 points

    Bob Bongiorno said:
    Our settings for the processor are for minimal (0pF) internal capacitance.

    YOu still have the 2pF pin capacitance (resulting in 1pF load).

    Also, if you have a GND layer beneath the MSP, this adds a lot of capacitance too, even if the crystal is directly soldered on the pins.

    But I agree, i wouldn't expect that much.

    Bob Bongiorno said:
    blames the capacitance of the socket

    Socket capacitance sure is high. However, it's difficult to see whether it's only the socket or something else.

    Don't forget that typical watch crystals are specified for 25°C and the temperature coefficient changes sign at this point, so the frequency drops for higher and lower temperatures. This is different from normal high-frequency crystals which have a rather constant positive temperature coefficient.

  • 0 in reply to Jens-Michael Gross
    Prodigy 30 points

    For a standard watch crystal, you can see about a 2 ppm shift (0.06Hz), which is not much, from 20 to 30 degrees C but our temp is constant 22. Besides, we're seeing much more shift than than that.

    If I had to do it over, I'd choose a 12pF load crystal because it's less affected by stray capacitance and internal variations which is even mentioned by T.I. in that same app-article. I don't think I'll get a straight answer about this one. I'm just looking for consistency...

  • 0 in reply to Bob Bongiorno
    Guru 227245 points

    Bob Bongiorno said:
    I don't think I'll get a straight answer about this one. I'm just looking for consistency...

    I understand your need. However, I fear that not much mor than the already discussed information is available.

    Maybe a TI engineer may dig into it further or locate internal information that just wasn't release d(but isn't confidential).
    Or you do your own investigations if the results are so important for you.
    I always trust my own experiences more than any datasheet. Especially the bad ones ;)

**Attention** This is a public forum