xtal oscillators

https://www.researchgate.net/post/What_result_does_selecting_the_wrong_load_capacitance_for_a_crystal_have
Mercury TN-021 technical note: https://www.researchgate.net/profile/Hazim_Tahir/post/What_result_does_selecting_the_wrong_load_capacitance_for_a_crystal_have/attachment/59d632fdc49f478072ea1e61/AS%3A273638526259200%401442251916272/download/tn-021.pdf
Pierce-gate oscillator crystal load calculation (Ramon Cerda): https://www.researchgate.net/profile/Hazim_Tahir/post/What_result_does_selecting_the_wrong_load_capacitance_for_a_crystal_have/attachment/59d632fdc49f478072ea1e62/AS%3A273638526259201%401442251916516/download/PierceGateLoadCap.pdf

Pierce oscillator

critical parts:
 amp/gate         usually a NOT or NAND gate; beware of the device's rise/fall times and prop delays
 xtal (well, duh)
 Rfeedback        1..10 MΩ (100k..22MOhm)
 biases the gate input at approx. Vcc/2 volts lower values increase feedback, but also increase circuit load lower for higher freq (10MOhm for 32kHz, 100kOhm for 10..20 MHz) integrated in xtal osc inputs of chips connect directly, not through Rseries
 Rseries          0..few kΩ, important for low frequencies to avoid overdriving, put on amp/gate output!
 Cx1, Cx2         loading caps, low dozen pF range, control voltage across xtal; don't forget to add parasitics (ballpark 5pF/pin + circuitboard, typ. 3pF)
 Cx1 typ. 0..30 pF (with an inverter), in some applications can be variable to trim freq
 Cx2 typ. 5..50 pF
 typical total load Cl are 12/15/18/20/22/32pF, typ. use Cx1=Cx2=Cl*2


pins OSC1,OSC2 or X1,X2
usually an inverting gate acting as amplifier/gain source
has input (sensing, 1, in, EXTAL), output (driving, 2, out, XTAL) (don't rely on numbering)

Cx1,Cx2 (Cin,Cout, Cx,Cy,...) usually same value, but higher Cout/Cin ratio can increase gain
higher capacitances:
 better freq stability
 higher current consumption
 longer startup time
 longer propagation delay
 lower loop gain
 less trimming range
 lower frequency of xtal (adjust to trim)

amplifier/gate
 usually NAND or NOT gate
 older parts (CMOS 40xx) are slow at low voltages, even moderate frequencies can be a problem (MC14049B doesn't want to work at 4V and 3.58 MHz!!!)
 gain critical for higher freq

driving powers

1 uW for tuning forks at kHz range
100 uW for UM1/UM5 in MHz range
1 mW for HC49/U in MHz range (up to 5 mW)
2 mW for HC51/U in MHz range

physically smaller xtals (rectangular strips vs bigger circular cuts) have lower max drive power, lower ESR, are more sensitive to mechanical shocks

example of Cx1,Cx2 effects:
 Values shown are for a 4.9 MHz crystal, a typical M68HC11 drive circuit, and Vdd = 5V.
 Voltage Changes for Varying Stabilizing Capacitors Cx and Cy:
Cx1=	Cx2=	V@EXTAL		Crystal Power Dissipation
56 pF 	56 pF 	3.3 Vpp 	100 µW
33 pF 	56 pF 	8.0 Vpp 	199 µW
47 pF 	56 pF 	6.1 Vpp 	207 µW
68 pF 	68 pF 	2.8 Vpp 	102 µW

tips

http://cache.freescale.com/files/microcontrollers/doc/app_note/AN1706.pdf

overdriving xtal (too much current flowing through)
symptom: frequency decreases or loses stability with increasing supply voltage
solution: add series resistor or lower supply voltage

too low Cload: won't start oscillating, the xtal ESR dampens oscillations
too high Cload: voltage swing reduced, may be below threshold for amp/inverter

power supply noise:
 can be greatly amplified by the amp/gate
 if harmonic to xtal, xtal may not start
 tying Cx1,Cx2 to power rail may be more stable reference than GND, espec. if ground has noise from return currents from power devices
 when problem, the Vout may be stuck at Vcc/2

long traces:
 coupling to ground
 coupling to noise
 parasitic inductances

contaminants:
 usually introduces stray resistances
 lowers gain by pulling pins down to GND across caps
 lowers Rfeedback, increases circuit load
 critical around gate input signal side

temperature:
 if not starting at higher temp, amp overloaded; check contaminants and loading caps
 test also at low temp (-40c) if xtal starts

freq instability:
 usually overdriving or underdriving the xtal

high freq, over 10 MHz:
 gate gain drops with frequency (espec. important if feeding other circuits)
 stray capacitances more important

low frew, below 50 kHz:
 impedances very high
 amp may be too strong, tendency to overdrive
 take care about Rseries

silicon geometry:
 die shrink: https://www.digikey.com/en/blog/why-my-crystal-does-not-start-up
 gate gain may depend on process used
 parasitics of the case can change
 adjust Cx1,Cx2


power variation test:
move Vcc from 3 to 5.5 volts (for 5v logic)
 freq should very slightly increase with voltage
 if freq decreases with voltage, crystal is overdriven!
 if no oscillation, lower Rfeedback

xtal freq test:
 measure xtal freq off-circuit, with good Cx1,Cx2
 put xtal in circuit, measure freq; if too different, suspect stray capacitances, reduce Cx

transients test:
 run board with exercising power switches, look for transients in wavefoms and on power/gnd rails

if not starting at all:
 check Cx1 side voltage; if Vcc/2, Rfeedback too high or noise prevents oscillations

fast-rising power can kick xtal on to oscillate



resonant freq: Fr, antiresonant (higher): Fa
between Fr and Fa xtal appears inductive (reactance positive)
series cap: pulls Fr up
parallel cap: pulls Fa down

thinner cuts for higher freq: 0.15mm at 15MHz

consider MEMS oscillators (SiTime, SiLabs)
 consider programmable parts + programmer (Time Machine II)