choosing crystal when tracks are very close together

This is a circuit I am having trouble with. I made 20 units with the exact same setup.

Three units with this setup managed to work with 33pF ceramic caps with 20% tolerance and a crystal from Ebay.

The remaining units are failing yet nothing is smoking or showing discoloration. I even replaced the 20% ceramics with 5% npo caps and still nothing.

I think now the lucky units were working because the crystal load capacitance and ESR happened to be spot-on.

What I want to try to do since people here consider ebay as a flea market is buy crystals from digikey. But there are two types I could possibly choose from.

One with 18pF load capacitance and one with 20pF load capacitance. The ideal frequency I'm looking for is 22.1184Mhz.

When I look at my tracks between the crystal leads they're probably too close together and I should probably increase the parasitic capacitor value when calculating my crystal load capacitance value.

One source on the internet claims this math for crystal load:

 CL = C/2 + StrayC


Because the caps I will use are 5% 33pF, its actual value would be between 31.35pF and 34.65pF

Now diving those capacitor numbers by two, we get 15.675pF and 17.325pF respectively.

People say the normal stray capacitance value is between 2 and 5pF but isn't mine more due to tight track spacing? I don't understand. So if it was 2pF then values would be:

  17.675pF and 19.325pF


Which means an 18pF load capacitance crystal is my best bet, but my stray capacitance is probably higher since I'm using a socket for my micro controller.

If its 5pf stray, then values would be:

20.675pF and 22.325pF


Which means a 20pF load capacitance crystal is my best bet.

Somehow I think the 20pF load crystal is better for my situation but I could be wrong.

Am I best to choose a crystal with 20pF load capacitance or 18pF load capacitance? and yes I will still use a standard IC socket for the micro (AT89S52) that uses this crystal.

• The "crystal from Ebay" is likely the problem. Commented Jun 10, 2019 at 15:12
• If one were bound and determined to use eBay crystals one could make a test oscillator. With care you could cook up an oscillator that'd indicate crystal activity, to boot. Just set it up next to a trash can with your bag of crystals and start sorting good from bad. Commented Jun 10, 2019 at 15:23
• Mike, you've produced twenty units of this, and rely on ebay-sourced crystals? Why? You've now got half about half a dozen electrical engineers that tell you that using crystals from ebay is a bad idea. Maybe, just maybe, spend USD 11.50 on twenty oscillators that are of good quality instead of hunting bugs that have nothing to do with your design, but all with the fact that you make bad sourcing decisions. Commented Jun 10, 2019 at 15:27
• (Edit: it's not half a dozen, but only five engineers telling you the problems are most likely related to your component quality in this and your previous question: @TimWescott, Matt Young, Sunnyskyguy EE75, Chris Stratton and myself.) Commented Jun 10, 2019 at 15:32
• @MarcusMüller I'll bring it up to half a dozen..... Don't use eBay crystals if you want reliability!
– MCG
Commented Jun 10, 2019 at 15:34

If you follow the recommended oscillator circuit, a good quality active crystal should work with a load capacitance that's off a bit. The problem isn't that it won't oscillate at all, the problem is that it'll oscillate at the wrong frequency.

Looking at your layout your oscillation frequency may be extra sensitive to temperature variations because FR-4 is a crappy dielectric, but it should work.

• ok. The expected temperature of the circuit will be no more than 25 degrees celcius because it will be in a plastic enclosure and used in a game. And by wrong frequency, I'm guessing if I pick the lower load capacitance crystal and the math results in the need of a higher capacitance crystal, then the actual frequency is lower than the frequency printed on the crystal can? and yes I'm following the suggestions from the microcontroller datasheet itself. I'm going to guess that the remaining unused crystals I bought from ebay are trash now. Commented Jun 10, 2019 at 15:38

• ground plane adds more unwanted stray capacitance (to gnd) than coplanar tracks. Maybe <150 ppm lower f, if C is too high and <150 ppm high if C is too low.
• no thermal solder pads for SMD and XTAL parts. Each pad should be a thermal island of heat when soldering so you do not have to heat up long fat tracks.
• XTAL Solder time 3s max.. fat tracks. maybe you damaged the XTAL if too hot and too long.
• no datasheets, no idea if it is series or parallel resonant, no idea if it is fundamental or harmonic (overtone) type.
• no series current limiting R=4k7 typ. to limit unknown xx uW rating of XTAL
• no test measurements.
• measure Voltage on pins 18 (XTAL2=out) and 19 (XTAL1=in) with DVM or better with scope
• if Vout = 5V, input is shorted to Gnd
• if Vout ~ 2.5V may be oscillating but if no clock , then ESR is too high or XTAL is damaged.
• This is weird. one board is rather wacky. The voltage between VCC and ground is 5, but when I measured crystal directly, voltage is 5.4. If I measured the other way, I get 0.8V but the crystal pins aren't grounded. I measured that with my meter to make sure. I'm going to guess a broken crystal? Commented Jun 10, 2019 at 22:08
• There is a CMOS inverter across those pins, how did you measure it? P18 is output Commented Jun 11, 2019 at 0:38
• Try 1Meg across Xtal and verify output. The DMM leads will affect capacitance. The DMM has a hard time reading this RF but you get weird DC effects. A CMOS Logic Probe is better. Get one Commented Jun 11, 2019 at 0:39
• Now I'm wondering if 6V is too much for the Ebay crystals to handle. My digikey crystals are currently in a postal truck. Because after 6V, the voltage measurement across the crystal is 0V which makes me think that the crystal is a short circuit at some frequencies, but when I measure manually for a short, there isn't one. I'm going to try digikey crystals and limit the voltage to 5.5 (via LM2940 regulator) like I do with the previous board that did work. Maybe that extra 0.5V was enough to destroy the crystal. Commented Jun 11, 2019 at 1:41
• mike if your are measuring across the crystal, that is not what I asked, it should be with respect to 0V Commented Jun 11, 2019 at 2:58