RE : calculating the output power of a Quartz oscillator

Dear david j james,

david j james said:

PSS simulation gives a Fundamental frequency of 10.0275 MHz, Harmonic balance gives Frequency= 1.0028e+07. That's a 0.0001% difference!

The difference appears to be about 40 ppm - which is actually a "large" frequency difference in the field of bulkwave quartz oscillators.

In any case,I suspect there may be a simulation accuracy issue or a possible power calculation issue as the shooting simulation result is indicating the steady-state oscillation frequency is closer to series resonance than the result of the Harmonic Balance simulation. Therefore, the quartz crystal unit power dissipation of the shooting PSS simulation should be higher that for the Harmonic Balance PSS simulation. Your data suggests the opposite. If you can, might you share the exact method you are using the compute the quartz crystal unit power dissipation and your PSS simulation settings (the two input.scs file simulation lines will contain the information)?

Shawn

In my example, with 3Vpp across the crystal and recommended ( by manufacturer) of 10pF external caps this gives Power Dissipation >> 100uW, which is TWICE the Cadence simulation result. The 3Vpp is not flexible before you ask. Is it 3Vpp unusual to use rail to rail drive?

The two simulations (pss and hb ) give different (slightly) results.

Do the following setups make sense?

PSS input.scs

simulatorOptions options psfversion="1.4.0" reltol=1e-3 vabstol=1e-6 \

    iabstol=1e-12 temp=27 tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 \

    gmindc=1e-15 rforce=1 maxnotes=5 maxwarns=5 digits=5 cols=80 \

    pivrel=1e-3 sensfile="../psf/sens.output" checklimitdest=both  \

    sqldb_violating_ratio=yes

pss  (  vout  gnda  )  pss  flexbalance=yes  fund=10M  harms=100

+    errpreset=moderate  oscic=lin  oscmethod=onetier  tstabmethod=trap

+    annotate=status

Calculator function : ymax(leafValue( hb_spec_Pout "Design_Point" 1 )) gives 86uW

Harmonic Balance : input.scs

simulatorOptions options psfversion="1.4.0" reltol=1e-3 vabstol=1e-6 \

    iabstol=1e-12 temp=27 tnom=27 scalem=1.0 scale=1.0 gmin=1e-12 \

    gmindc=1e-15 rforce=1 maxnotes=5 maxwarns=5 digits=5 cols=80 \

    pivrel=1e-3 sensfile="../psf/sens.output" checklimitdest=both  \

    sqldb_violating_ratio=yes

hb  (  vinv_xto_inv  vinv_xti_inv  )  hb  saveinit=yes  autoharms=yes

+   autotstab=yes  oversample=[1]  fundfreqs=[(10M)]  maxharms=[5]

+   errpreset=moderate  oscic=lin  oscmethod=onetier  annotate=status

Calculator function : ymax(pvi('pss "/vinv_xto_inv" 0 "/V_xosc/PLUS" 0)) gives 76.4uW

You mentioned that the output is pretty square, so I rather doubt that 5 harmonics would be enough - although you have "autoharms" set to yes, so I don't know how many it actually chose (the log file would show that). Understanding this is hard without seeing it, so I suggest you contact customer support - that's what it's for (submit a support case after logging in). Then maybe you can share the example, we can look at log files, and see what accuracy controls are affecting the difference (your expressions are different between the two and you have them the wrong way around, and the hb expression doesn't seem to actually include the output power expression).

Andrew

My comment was misleading. The output of the osc, the clock to the digital, is quite square but I case about power into the crystal. But the Quartz waveform is quite sinusoidal. The calculator functions are from the Main Results Viewer.

Results → Direct Form → Main form

Select pss (of hb), select power, select Terminal ( an ideal zero volt source) on input to crystal.

Is measuring the Power into a Quartz crystal by simulating the power into a terminal (of an ideal voltage source) a valid method in pss or hb?

It doesn't really matter whether the node in question is sinusoidal if the rest of the circuit has higher harmonic content - the hb solution needs to have enough harmonics to accurately represent the solution everywhere (otherwise the missing energy tends to "leak" into the other harmonics). 

I can't really answer the question about how to measure the power for a Crystal oscillator without trying this out - to be honest, I've never done this with this kind of circuit. Another reason to contact customer support - then somebody can spend the time on it.

Shawn, however, does have experience in this area (I believe, from some of his earlier posts).

However, I just noticed (looking back at your reply in this thread where you included the analysis statements): https://community.cadence.com/cadence_technology_forums/f/rf-design/57185/re-calculating-the-output-power-of-a-quartz-oscillator/1389616#1389616 that you're actually running harmonic balance in both cases. One is via PSS but with 100 harmonics chosen, and the other is with hb with auto harmonics chosen. I'd assumed that you were using shooting PSS and comparing against harmonic balance. There's no fundamental difference between hb-in-pss and hb (it's the same engine). I suggest you pick 100 harmonics in the hb analysis and you should get the same answer from both approaches.

Regards,

Andrew

Dear david j james,

david j james said:

In my example, with 3Vpp across the crystal and recommended ( by manufacturer) of 10pF external caps this gives Power Dissipation >> 100uW, which is TWICE the Cadence simulation result.

I can't reproduce your result as I do not know your quartz crystal unit's parameters.

1. Is the 10 pf the specified load capacitance on its data sheet?

2. Is that the 10 pf you used in the workbook calculator or did you add something to it?

3. Did you enter a voltage of 3 V in the workbook? The units adjacent the entry box indicate the voltage is the rms voltage and not the peak-to-peak voltage. If your driving waveform is "CMOS" like (i.e., analogous to a square wave), the correct value is 1.50 Vrms. If it is closer to a sinusoid, the correct value is 1.067 Vrms. If you used the peak-to-peak value and the waveform is square, that might explain the difference in the workbook result and your Cadence result as the use of the peak-to-peak voltage will double the power esult relative to the power when the correct rms value is used in the workbook.

david j james said:

The 3Vpp is not flexible before you ask. Is it 3Vpp unusual to use rail to rail drive?

It is a little large, but not unheard of.

Shawn

Dear david j james,

Since Andrew has already commented on your simulator settings, I will default to his expertise.

david j james said:

Calculator function : ymax(leafValue( hb_spec_Pout "Design_Point" 1 )) gives 86uW

I noticed in your Calculator functions you are measuring the maximum power - this is not the value that a data sheet for a quartz crystal unit will specify. The maximum power dissipation is usually specified as an rms value. This is also why the Excel workbook uses the rms value of the voltage across the quartz crystal unit.

david j james said:

Is measuring the Power into a Quartz crystal by simulating the power into a terminal (of an ideal voltage source) a valid method in pss or hb?

I am not sure I understand enough about the topology of your measurement. How is the ideal voltage source configured relative to your quartz crystal unit and what is the opposite terminal of your quartz cyrstal unit connected to? Is your oscillator a Pierce or Colpitts topology?

Shawn

Thanks. I was not aware that hb-in-pss and hb are the same thing using the same engine.