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.

Crystal maximum ESR versus Rs

Other Parts Discussed in Thread: TMS320VC5506, TINA-TI

How critical is the maximum ESR for the external crystal (e.g. for the C5506)?

I have selected a 12 MHz crystal for compatibility with the USB bootloader in ROM.  Table 5-1 suggests a max ESR of 30 Ω for the 12 MHz to 15 MHz range, and a max ESR of 40 Ω for the 10 MHz to 12 MHz range, the latter recommending a value of 100 Ω for Rs.

I was able to find a 12 MHz crystal with a max ESR of 80 Ω and shunt capacitance of 5 pF.  That seems to violate the maximum of 40 Ω found in the table, but there is a cryptic note in SPRS375C stating "theoretically a crystal with a lower maximum ESR might seem to meet the requirement. It is recommended that crystals which meet the maximum ESR specification in Table 5-1 are used."  What is not mentioned is whether a crystal with a higher maximum ESR would work.  That quoted statement seems to imply that I should avoid going below the value listed, but since I have gone above the suggested value it makes me wonder what the theory would say there.

A second question involves Rs, which is 0 Ω for crystals in the 12 MHz to 20 MHz range, but increases from 100 Ω at 10 MHz to 12 MHz up to 2,200 Ω at 5 MHz to 6 MHz.  I understand that Rs can reduce noise on the board by shaping the external electrical waveform more like a sine wave than a square wave, and that it basically acts as an RC low pass.  I'm tempted, though, to think of the crystal ESR as contributing to the same effect.

Question: If I have a crystal with an excess maximum ESR of 40 Ω to 50 Ω - i.e. 80 Ω (rated) minus 30 Ω to 40 Ω (recommended) - then would this be about the same as having 40 Ω to 50 Ω extra for Rs?

For a 12 MHz crystal, it seems like anywhere from 0 Ω to 100 Ω would be valid for Rs, so should I put a 50 Ω resistor there to combine with the 40 Ω to 50 Ω excess max ESR?

I realize that crystal circuits are highly dependent upon board layout and trace geometry.  I also realize that Rs is on a different node than the ESR of the crystal.  But I still find this an interesting topic to consider, especially since I am looking at a revision of my board.

P.S.  I have two working revisions of the 12 MHz board with C1 and C2 of 32 pF and Rs of 0 Ω.  There do not seem to be any problems, but I have not tested under the most adverse conditions yet.

  • I will take a look, and get back with you.

    Wen

  • Hi Brian,

    I'm also using a 12 MHz crystal with 80 ohms ESR. The ESR is apparently a function of size, so the only way to get the data sheet's recommended value is to spec a really large can style as used on the demo boards. That's just not practical for today's product designs which have footprint and height constraints, and the extra trace length is likely to cause EMI problems as well. So I've decided to go with a modern surface-mount crystal, and I plan to test multiple boards over temperature to be certain the oscillator starts reliably.

    It would be helpful if TI could provide additional guidance on this point. And maybe they need to redesign their oscillator circuit to be tolerent of the smaller crystals that people need to use these days.

    David

     

  • Thanks for the information, David.  I agree that the large can style crystals are not very common in modern designs any more.

    By the way, the part that I am using is the ECX-53B series, 12 MHz ECS-120-20-30B-DU, from ECS Incorporated International.

    Hopefully, Wen or someone else from Texas Instruments can report on any information that has been found.

  • I'm using a different ECS part, ECS-120-20-23B-F-N, but I'm guessing it's probably the same crystal blank in a different package.
    David
  • Wen, did you ever find an answer?

  • Brian;

    Here is the info that I found. Basically, we should try to not go much over 5X negative resistance at the fundamental. According the plot,  if your crystal is 12Mhz (fundamental) with 80 ohm ESR, it should work fine.

    Thanks and regards


    Wen

    uos327_negative_resistanc 9.pdf
  • Thanks.  I must admit that I got a little lost reading that chart without the accompanying test circuit.

    I placed Rs in series with X1, as depicted on page 73 of the TMS320VC5506.PDF, but the plot mentions 100 Ω in series with X2.

    I have 32 pF from X1 to VSS (on the far side of Rs, closer to the crystal than X1) and another 32 pF from X2 to VSS.  The plot documents 20 pF for these, but I don't know whether I can assume that my 32 pF would make a difference.  I got 32 pF from the calculations on page 73 as well.

    My crystal has a 5 pF shunt, which is below the 7 pF maximum on page 73.  It's CL is 16 pF, which matches the typical value in the chart.

    I think that I will maintain pads on my next board revision for Rs, but populate with a 0 Ω jumper unless I specifically find that I need to increase that towards 100 Ω.  Maybe 20 Ω would be safest.  I realize that I may be worrying about this way too much, but we do not have the budget for a dozen board revisions or expensive testing, so it seems prudent to design prophylactically.

    Question: Does it matter whether Rs is on X1 or X2?  I thought that the internal oscillator circuit was different for each pin and that it would make a difference, so I followed the data sheet.

    Question: Should I just connect a 'scope to X1 and X2 and look for a sine wave, ideally?

  • Brian,

     

    I can see the chart Wen provided is not very helpful without the test environment. Still, we are trying to get more data from the oscillator designer. 

     

    As far as your questions:

    1) Place Rs on X1. X1 is the output of the oscillator circuit. X2 is the input to the oscillator circuit. I believe that plot has X1 and X2 labels interchanged.

     

    2) Connecting a scope to X1 and X2 should show you a nice sine wave. However, the scope's probe capacitance will change the loading on these pins. This in turn can change the characteristic of the oscillation.

  • Brian,

     

    We got some information from SPICE runs on the oscillator cells. For 12MHz with a 20pF total load capacitance (includes board trace capacitance, package capacitance, decoupling capacitors, etc. This can be about 2-3pF more than the CL caps), the following configurations work fine:

      Maximum ESR       shunt capacitance

         100 ohms                     3 pF

          80 ohms                      5 pF

          60 ohms                      7 pF

         40 ohms                     11 pF

    In fact, a higher ESR would extend the oscillator stabilization time.

     

    Seems the datasheet values are too conservative, we will update these on a future revision of the datasheet.

  • Thanks.  I was just wondering if SPICE could handle a crystal...

    Can you share those SPICE files?  I have the Tina-TI SPICE that Texas Instruments makes available, and then I could plug in the 32 pF values I'm using for the decoupling (?) caps to see what affect that has compared to the default 20 pF (or is all the capacitance combined into one element?).

    So, startup time might take longer with higher ESR, but that seems like a decent tradeoff for reduced EMI.

    If you're planning on revising the datasheet, then we can all get an advance look if you also append the revised text to this thread.

  • Brian,

     

    Sorry, We cannot share the SPICE files due to proprietary information.

     

    For 20pF, I meant total capacitance. Thus for your 32pF caps, you are looking at CL = (C1*C2)/(C1+C2) = (32*32)/64 = 16. To this add 2-3pF for board traces and package capacitance, then your total capacitance is 18-19pF. This is basically similar to your system.

  • Pedro,

    Thank you for spelling out the equivalent capacitances.

    To recap my question, I have a choice for Rs between 0Ω (a jumper) to possibly 100Ω maximum according to the chart.  My board is "working" now with 0Ω (a trace) for Rs.  I would like to know whether I could lower EMI by choosing, say, 20Ω for Rs, so that the total might be 100Ω when considering the 80Ω ESR of my crystal.  (I'm still not sure whether Rs would be added to the crystal ESR quite so literally as I suggest).

    Or, to phrase it from the opposite direction: What value should I choose for Rs for minimum EMI, considering a 12 MHz crystal with ESR of 80Ω and shunt capacitance of 5 pF and CL of 16 pF?  My board is sensitive to EMI, so all of my questions have the purpose of minimizing interference between my own circuits, with the side benefit of reducing EMI for other electronics.