Crystal specs and configurations.

[kvresto]

Hi Everyone.

Its been  while since I've posted a basic question, and as a novice hacker its these type of question I cant Google. This time it revolves around crystals. I read a couple of app notes on selecting crystals and load capacitors, and I'm fine with that, but looking through Digikey and others I find in the SMD range there are two main types, two pad SMD and four pad SMD (not the CAN type).

Why the two and what are the pros and cons of either. Also I notice the CL spec varies from about 8pF to about 30pF, is this just a manufacturers design/process, or is there a specific reason for large differences, and what is my consideration if I want to use either to clock say a PIC, AVR or other? for example cpu data sheet for the stm32f103 gives a recommended load capacitance of 30pf, but several of ST's designs use a crystal to clock the same micro with a data sheet spec of 8pF?, and then recommend a CL in the range of 5 to 25pF? (page50 from Doc ID 13587 Rev 11 September 2009)

Thanks
kvresto.

[T3sl4co1l]

More pads is just a packaging difference. I don't know that they make any electrical difference.  That said, often the pads are connected to the case, and it's preferred to connect these to ground as shielding.  Usually the datasheet will suggest this.  Leaving them NC isn't a problem.

Capacitors vary due to frequency range, crystal cut, geometry and impedance.  Use the crystal datasheet recommendation.

Oscillators should never recommend loading capacitors.  That particular recommendation probably goes the opposite direction, that the oscillator is expecting a certain impedance, and so a crystal which requires that amount of capacitance.

Also, if they are recommending a full example, including a specified crystal (manufacturer and part number), then that's pretty obviously a single case example.

Note that the loading capacitance is the total parallel equivalent, as seen by the crystal.  The physical capacitors added, act in parallel with the capacitance of each OSC pin (which should be listed in the datasheet somewhere).  And those capacitances, in turn, act in series as seen by the crystal.  So the total capacitance per pin needs to be double the amount shown in the datasheet.

(I think. Unless they're actually saying to use the per-pin capacitance, in which case don't double it.  I was never very clear on which one it is...)

If the capacitors are improperly dimensioned, the oscillator may not start up correctly, or it runs at an unusual amplitude (consequences..?), or the frequency is out of tolerance (the crystal might be rated for 50 ppm accuracy, but can be "pulled" off frequency by using more or less capacitance).

Tim

[kvresto]

Thanks Tim, I might be asking the same question here and you may have answered it, but as the stm32 data sheet recommends

                           "Recommended load capacitance versus equivalent serial resistance of the crystal (RS) = (typ: 30pF)",

and my crystal selection CL spec is CL=8 or CL=18pF to mention two I'm looking at, do I need t take this micro data sheet spec into consideration in any way?,
do I just size the caps as per the crystal data sheet spec for CL,,,, or other?

 I know you already mentioned to "Use the crystal datasheet recommendation", but some clarification for me here would be great.

cheers

[CaveMannDave]

Hi:

Why the two and what are the pros and cons of either. Also I notice the CL spec varies from about 8pF to about 30pF, is this just a manufacturers design/process, or is there a specific reason for large differences, and what is my consideration if I want to use either to clock say a PIC, AVR or other?

As an educated guess, I would say that the style of Cut (bar, plate, or 'tuning fork' shape), package profile (flatter 4-pad = lower and sturdier, narrower 2-pad = less real-estate and less sensitive to board flex), stray capacitance, designer preference, intended use, Temperature Coefficient curve, & etc. all come into play, and more.
Technology plays a big role, too -- time was when crystal "holders" were the size of your two thumbs, side by side, the 'slab' inside was larger than your thumbnail.

... for example cpu data sheet for the stm32f103 gives a recommended load capacitance of 30pf, but several of ST's designs use a crystal to clock the same micro with a data sheet spec of 8pF?, and then recommend a CL in the range of 5 to 25pF? (page50 from Doc ID 13587 Rev 11 September 2009)

Physics, and resonant circuit theory and practicalities.  The range is there because AIUI, it is: a) Hard to control, b) Has limited effect, c) Capacitors have wide tolerances, d) Except in precision systems, who wants to Pay for 0.001 % without the need?

Plus, the I/O pins on a uP are not all that well-characterized or controlled.

ETA: If you want to get closer to 'Spec', you can parallel a ceramic trimmer cap.

GPS?  Use Caesium, Rubidium, or OXCOs, because it Matters.
Heart rate monitor?  Use a decent crystal, but heartbeats aren't stable, and move around every beat.
LED flasher or microwave oven timer uP?  Who would ever POSSIBLY know or care?  Use a 20% resistor and a +100/-25 % cap @ 500KHz.  Will it matter?

Hope this was useful.
Cheers,

Dave

[kvresto]

Thanks Dave, yes, very useful.

[T3sl4co1l]

Thanks Tim, I might be asking the same question here and you may have answered it, but as the stm32 data sheet recommends

                           "Recommended load capacitance versus equivalent serial resistance of the crystal (RS) = (typ: 30pF)",

Highlighted is the part they're actually interested in.  A crystal with lower ESR is expected to use larger Cp.  And will have a smaller resonant voltage when placed in a current-limited oscillator circuit.

Also, do they give a circuit, and do they indicate whether the load capacitance is "per pin", or "crystal equivalent"?

Tim

[kvresto]

Tim, I extracted the page 50 from the data sheet, I think everything is there (I hope), I hope its easy? to answer what these figures mean in selecting my crystal?

[Dragon88]

http://www.st.com/web/en/resource/technical/document/application_note/CD00221665.pdf

Also:

[krapht]

Hi,

getting crystals to oscillate accurately and with high performance correctly is a chore and requires accurate measurement equipment. As a practical matter, unless you want to get down and dirty with timing, either copy the manufacturer's recommended pcb layout and BOM exactly, or use a discrete oscillator IC.

The textbook that I use when I have to deal with this topic is http://smile.amazon.com/Discrete-Oscillator-Design-Nonlinear-Transient/dp/1608070476/ref=asap_bc?ie=UTF8

[kvresto]

Thanks everyone, very helpful, that video explains some details very well.

cheers

[uncle_bob]

Hi

Yikes ...

The load specified on a crystal is there for only one reason - to tell you what load the manufacturer used when they set it on frequency. That is the only significance of the number. Motional resistance, Q, and motional capacitance have no direct (or indirect) relation to the load capacitance. There is no way to guess from the load what the other parameters of the crystal might be / may be / could be. There is also no way to guess the amount of drive the crystal will tolerate.

So why so many load capacitances? The simple answer is that there are a lot of IC's out there. Different IC based oscillators present different load capacitances. The textbook says to match them up, so people do.

Does this matter? Well, that depends. If it's a 0.1% crystal, no it does not matter. The hundred or two hundred ppm frequency error from a 12 pf to a 32 pf load is nothing compared to the finish tolerance.

One other odd tidbit .... it does not matter if you put the load in series with the crystal or in parallel. The frequency you get out will be the same. (Yes there are a few other qualifiers, but it's why the whole series / parallel thing is nutty).

Bob