LT1712: Need help understanding and modifying this oscillator

Question

I'm in need of an analog circuit that can generate a square wave output of 20MHz. My first thought was to use a schmitt trigger inverter with a crystal, but then I thought of using a fast opamp to make it more challenging and a fun to work on project. 
 I came across this circuit made using LT1712 4.5ns comparator from Linear Technologies (now ADI). The circuit below generates a square wave output with a freq. of 1MHz which is filtered to obtain a sine wave of 1MHz. It uses a series resonant crystal to generate a stable 1MHz signal as shown in Fig.1 
 I modified the circuit, by replacing the crystal with a 20MHz one and R3, C1 by using another circuit given in the datasheet of LT1016 for 10MHz to 25MHz signal as shown in Fig. 2. I made a PCB to test the chip. However, my output frequency is around 10MHz with a distorted square wave output. I have tried the circuit in LTspice where it gives proper 20MHz output. 
 Hence, I need help in understanding how the series resonant crystal circuit is working, why the 210 ohm resistor? and how the R3 and C1affect the output freq? As the values are both different in the two circuit diagrams. I have also posted the LTspice circuit simulation and output as well as the measured output of my PCB.

Fig. 1: LT1017 1MHz square wave gen

Fig.2: LT1016 10MHz to 25 MHz generator

DSO output

dang74 · Accepted Answer

I am not an expert on this type of oscillator but I've seen it before. If you look at the negative feedback portion of the circuit, it could be said that when the output is positive capacitor C2 is charged through R3. Conversely when the output is negative, capacitor C2 discharges through R3. Now on the positive feedback side of things I believe the trip point at the noninverting input will be different for the two main phases of the output... and while I don't know the specifics R4 plays a role in this. I suggest that you add two more probes on your simulation circuit... one at the inverting input and the other at the noninverting input. From there play around with the value of R4 to see its effect.

dougw · Answer

I am not an expert on this at all. If you remove the crystal, this circuit will oscillate based on R3 and C2. If these component values are adjusted until the frequency is approximately correct, reinserting the crystal should lock the frequency to the crystal resonance?

jc2048 · Answer

Another non-expert here :) 
 "why the 210 ohm resistor" 
 Series resistor is to reduce the power that the crystal is having to deal with. It may be a small piece of hard rock (quartz), but it will crack if it's driven too hard. You need to choose it according to the spec of the crystal (which will give you a max figure) and a calculation of the power that you're delivering which will depend on the output characteristics of the comparator and its output swing on one side and the bias network on the other.

shabaz · Answer

Hi, it's unusual to use such a circuit (especially that very expensive part). Personally I would have gone for maybe a CMOS logic IC. There are a few (I am mobile, not easy to check the specific part number) logic ICs that also contain built-in dividers, they provide a great quality square wave output. Otherwise, for more fun/challenge, check out Oscillator Design and Computer Simulation by Rhea (It's possible to read it for free by virtually borrowing it on Internet Archive), one particular chapter has various example circuits, the one here would provide a signal, then you'd need to square it using (depending on output level) say a logic IC or another transistor, you'd take the output from the arrow, using a capacitor). The left and right side of this circuit is supposed to be connected together, I.e. connect that arrow to the wire labeled 'loop'.

LT1712: Need help understanding and modifying this oscillator

Top Replies