Maker Pro
Maker Pro

crystal load capacitor sizing

J

Jamie Morken

Jan 1, 1970
0
For this crystal:
"http://www.rocketresearch.org/new/crystal/crystal2.jpg"

The crystal is on the right side of the board vertically
between the two load capacitors, the bottom load capacitors
trace is highlighted red. The board thickness is 1/16" and
the trace width is 10mils.

datasheet:
"http://www.eea.epson.com/go/Prod_Ad...35&categoryId=EEA.QD.Crystals.kHzSMD_Crystals"

(digikey part#: SE2413CT-ND)
(manufacturer part#: FC-135 32.7680KA-A3)

frequency: 32.768kHz
load capacitance: 12.5pF

For selecting the load capacitors to use for the crystal oscillator
circuit I am using the formula:
CL = [(CL1 * CL2) / (CL1 + CL2)] + Cstray

I am wondering what the stray capacitance Cstray is. If I estimate
2pF for the above circuit then I get:

[(21*21) / (21+21)] + 2 = 12.5pF



Also for this crystal:
"http://www.rocketresearch.org/new/crystal/crystal1.jpg"

C12 and C13 are the load capacitors and the trace width is 10mil
and the board thickness is 1/16"

datasheet:
"http://www.abracon.com/Resonators/abm3b.pdf"

(digikey part#: 300-8206-1-ND)
(manufacturer part#: ABM3B-16.000MHZ-10-1-U-T)

frequency: 16MHz
load capacitance: 10pF

If Cstray is 2pF, then:
[(16*16) / (16+16)] + 2 = 10pF



Would the two above crystal oscillator circuits have about 2pF Cstray
given the board layout?

Also if the actual load capacitance is high, the frequency will be
lower, and if the actual load capacitance is low, the frequency will be
higher. Is there a formula to give the percent change in frequency for
a given load capacitance mismatch?

What kind of error in frequency can I expect if I have an actual load
capacitance that is 15pF instead of the specified 10pF for the 16MHz
crystal?

cheers,
Jamie
 
A

Anthony Fremont

Jan 1, 1970
0
"Jamie Morken"
Would the two above crystal oscillator circuits have about 2pF Cstray
given the board layout?
Probably.

Also if the actual load capacitance is high, the frequency will be
lower, and if the actual load capacitance is low, the frequency will be
higher. Is there a formula to give the percent change in frequency for
a given load capacitance mismatch?

None that I know of. It's not that clear cut.
What kind of error in frequency can I expect if I have an actual load
capacitance that is 15pF instead of the specified 10pF for the 16MHz
crystal?

Probably insignificant. What is the crystal's purpose? If it's
clocking a microcontroller, I wouldn't worry too much about it since you
are probably only talking about a few Hz difference. If it's in a model
rocket, you may want to consider a ceramic resonator instead of a
crystal for durability reasons. One thing you don't want is too little
capacitance or you'll have start-up/stability problems. Waking up from
SLEEP mode can be a killer too. It's easy to start an oscillator with
power application, but when a micro is sleeping it can be difficult to
cause enough imbalance to start the oscillator going. In some cases you
might need to have one cap at a different value to help.

As far as the clock crystal goes, what kind of long term accuracy are
you needing? If you are worried about individual pF of load
capacitance, you are only talking about perhaps a 1 or 2 seconds/week
change in accuracy. Temperature changes will probably have a larger
effect. If you need extreme accuracy, then use a trimmer cap.
 
T

Tim Shoppa

Jan 1, 1970
0
Jamie said:
What kind of error in frequency can I expect if I have an actual load
capacitance that is 15pF instead of the specified 10pF for the 16MHz
crystal?

For an AT cut it'll probably be in the very low tens of ppm. In the
total error budget (initial frequency offset, temp variations, etc.)
this isn't entirely in the noise. But that's a good thing: it's why a
trimmer cap gives you enough control to set the frequency exactly.

All of what you said about parallel load caps and strays is pretty much
on target, BUT in the real world the capacitor values are usually
subject to the constraint that they ensure rapid startup of the
oscillator and that the oscillator run stably and with a "good"
amplitude (not too high to avoid damage to the crystal or accidentally
run it in an overtone mode, but high enough that the microprocessor
actually clocks). If it so happens that this is somewhere near the
crystal's load capacitance, that's great.

Actually ensuring reliable crystal oscillator startup and operation
over a wide temp/Vcc/process variation range is not generally an easy
problem.

Tim.
 
W

Wes Stewart

Jan 1, 1970
0
[snip]
Also if the actual load capacitance is high, the frequency will be
lower, and if the actual load capacitance is low, the frequency will be
higher. Is there a formula to give the percent change in frequency for
a given load capacitance mismatch?

What kind of error in frequency can I expect if I have an actual load
capacitance that is 15pF instead of the specified 10pF for the 16MHz
crystal?

If you characterize the crystal(s) to determine the motional
parameters then you can calculate the effects.
 
T

Tim Wescott

Jan 1, 1970
0
Tim said:
For an AT cut it'll probably be in the very low tens of ppm. In the
total error budget (initial frequency offset, temp variations, etc.)
this isn't entirely in the noise. But that's a good thing: it's why a
trimmer cap gives you enough control to set the frequency exactly.

All of what you said about parallel load caps and strays is pretty much
on target, BUT in the real world the capacitor values are usually
subject to the constraint that they ensure rapid startup of the
oscillator and that the oscillator run stably and with a "good"
amplitude (not too high to avoid damage to the crystal or accidentally
run it in an overtone mode, but high enough that the microprocessor
actually clocks). If it so happens that this is somewhere near the
crystal's load capacitance, that's great.

Actually ensuring reliable crystal oscillator startup and operation
over a wide temp/Vcc/process variation range is not generally an easy
problem.

Tim.
If you had to get reliable startup characteristics as well as accuracy
you could probably change the distribution of the two capacitances --
but it'll be work.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google? See http://cfaj.freeshell.org/google/
 
Top