# Trying to get a 32.768 kHz crystal to oscillate

I've spent quite a bit of time read up on oscillator circuits and trying to get this circuit to oscillate. I'm trying to build this oscillator for a clock but I won't be putting it into a RTC, so I only need the 32.768 kHz frequency as an output.

Rereading, I'm pretty sure I need 25 pF caps instead as I misinterpreted how load capacitance worked. Regardless, I've tried a few different cap values (which were not 25 pF) and have gotten nothing.

I've also tried swapping and removing the resistors to a number of values as well.

However, I don't know what else is wrong. Moreover, I want to simulate it in LTspice but I don't know how to get the motional parameters as I only know the motional resistance.

I've also tried building this circuit on a breadboard and I get nothing but a flat line on my oscilloscope.

The crystal I'm using is a CFS-206 HZFB and the datasheet is here: http://cfd.citizen.co.jp/english/prod-tech/product/pdf/datasheet_TF/CFS-206_CFS-145_E.pdf

Thank you in advance for any help with this.

• Just my two cents worth: the idea is to make the inverter into an amplifier. That is what the 11.2Megs is for. But afaik the 74S04 makes a lousy amplifier. You need a different inverter. Try a CMOS inverter. – Oldfart Dec 20 '19 at 21:12
• That 270K seems a bit high too. When I've seen series resistors in oscillator circuits, they've been around 10-22 ohms. – Cristobol Polychronopolis Dec 20 '19 at 21:20
• But that is not a conventional crystal but a 32768 Hz tuning fork crystal for watches, so 270 kohm sounds approximately right for it. And the about 10 Mohm bias too, for a CMOS inverter. – Justme Dec 20 '19 at 22:26
• On a breadboard you don't need the loading caps on the crystal, as the breadboard traces have enough stray capacity already. So just test without C1 and C2. However, you will need these caps on a real PCB. – Turbo J Dec 21 '19 at 14:54

You need 25pF caps less any stray board and chip capacitance. You'll probably end up at around 18-22pF.

However, I don't know what else is wrong.

Using a 74S part is probably killing you, for starters, if you're really using it. And if you are, then R2 needs to be much lower than the megaohm range.

Switch over to a 74C04 or a 74HCU04 (the "U" is important -- it stands for "unbuffered", and means that there are fewer stages of amplification internal to the part -- this makes its action less positive as a gate, but more linear -- and you're using it as an amplifier here).

If you don't switch over to a 74C or 74HCU part, then start by choosing a value of R2 that puts the output at around 1.5 to 2V when R1 isn't present (VDD/2 for CMOS -- 1.5-2V is specific to TTL). I'm not sure what that value will be, but I suspect it'll be in the 10-50 kiloohm range. Then see if things work. Find the largest value of R1 that'll work, then cut it by about a factor of 2.

I suspect, however, that to make a 74S part oscillate you'll need to overdrive the crystal. The thing is designed to work on a whisper of power, but the 74S is a brute compared to a typical CMOS gate. I'm not sure if you'll break the crystal outright, but you may have problems with temperature and aging.

Moreover, I want to simulate it in LTspice but I don't know how to get the motional parameters as I only know the motional resistance.

Motational capacitance = teeny, motational inductance = ginormous. It really only matters if you're trying to characterize the thing for start-up time (it's the motational resistance that matters for whether it'll oscillate at all, and whether you're overdriving it). I'd just guess at a Q of between 10,000 and 100,000 and solve for the motational capacitance and inductance from that and the published motational resistance (it's effectively a series circuit, so use $$\X_{Cm} = X_{Lm} = R_m Q\$$).

If you just have to know, and if you have a good enough signal generator you can measure the parameters. It would be too much of a digression to explain my (quirky) methods here; just Google on "measuring crystal parameters" and choose a method that matches the equipment you have available.

• Very nicely covered! I always go with unbuffered and your comment regarding $R_2$ for S parts is right on target as it sets the loop biasing. – jonk Dec 21 '19 at 7:22
• @jonk if I had time (and still had my amateur radio license -- ex KG7LI) I'd be tempted to see how much power I could get from a crystal oscillator based on a 74S gate. I don't think I have any 04's, but I should have a NOR or a NAND around. It'd be a fun experiment, in an "I can make a tuba from a drain pipe!" sort of way. – TimWescott Dec 21 '19 at 15:38
• I have some S parts. They are ancient and I'd have to go find my ancient plastic box of little bins to find them. But I've got too much going on. And besides, you'd be better at it. But these tuning fork crystals, memory serving, are supposed to operate at under $1\:\mu\text{W}$. – jonk Dec 21 '19 at 17:44
• Thank you, everyone, so much for the help. I ended up grabbing some 22pF and 18pF caps along with a few CD74HCU04E CMOS Inverters. Swapping out the inverter with this new one and changing the value of both capacitors to 22pF got it oscillating at exactly 32.7kHz on my oscilloscope. – Eule56 Dec 26 '19 at 16:58
• You will find that you can tune the exact oscillation frequency by changing the caps -- higher capacitance should lower the frequency oh so slightly, lower capacitance should raise it. Changing the PCB layout will change it one way or another (and if you have a ground plane, leaving a "hole" behind the crystal will decrease stray capacitance from the board, which will decrease sensitivity to temperature). – TimWescott Dec 26 '19 at 17:01

74S04 is completely wrong chip for this and would require very different biasing to get it work. Usually a 74HCU04 or 4069UB is used with that kind of circuit.