Crystal Serie VS Parallele

There are two issues here. One is the series/parallel distinction. All crystals looks like a capacitor below its series resonance and also above its parallel resonance. In the very narrow region between, it looks like an inductor. So if you make an ordinary LC oscillator, and replace the inductor with a crystal, you'll get a crystal oscillator running at a frequency somewhere between the two resonances. (This assumes that the capacitors are in the right range of values--20 to 50 pF usually.)

The frequency marked on a 'parallel resonant' crystal is the resonant frequency of the tank circuit built from the crystal and the specified capacitance. (Parallel resonant crystals always have a capacitance spec as well as a frequency spec, for this reason.)

The second issue is startup. It's very possible for a poorly-designed crystal oscillator not to start reliably, or to start up at the wrong frequency. There are two classes of wrong frequencies: (a) overtones, and (b) LC resonances.

Crystals, like guitar strings, have more than one resonance. Generally the higher overtones are harder to use, because the parallel capacitance of the electrodes in the crystal tends to swamp the inductive reactance more and more at higher and higher overtones. So if you design your oscillator so that it needs a decent amount of inductance, and doesn't have too much gain, it'll usually work reliably at the fundamental.

But you have to really design it. Calculate how inductive your crystal gets (minimum and maximum Q specs apply); choose the capacitor values accordingly; and calculate the losses, so you can choose the right amount of gain in the active device: enough to start reliably from zero signal, but not much more than that.

If you don't do the design carefully, you're liable to find that your gizmo can oscillate at many frequencies--a few crystal resonances, the LC resonance of a poor layout, or the (much higher) frequency where the propagation delay and phase shift due to the capacitive load add up to

180 degrees. (Low and stable gain is your friend.) Crystal oscillator startup is one place where SPICE isn't that much help, so do the algebra. If you don't know where to begin, find the crystal parameters from the data sheet and post them. One of us will probably be able to help.

If you have a 3 MHz crystal, the easiest way to get 1.5 MHz is to use a flipflop as a divide-by-2.

Cheers,

Phil Hobbs

Thanks for the explaination. I used the ECS-30-S-1X Here a link for the crystal:

I am not sure I am familiar with the computation require to match the crystal to the 74LVC1404

For test, I plan to use:

Rs= 2.2m Rf = 1K C1 & C2 = 22pf

I was planing to use a simple flipflop to divide the 3Mhz by 2.

Bye Jacques

Well, they don't seem to give much info about the circuit's gain, do they? They do say that the gain is a function of frequency, so for a one-off, you could try changing the supply voltage and adjust Rs to get reliable starting and frequency keeping over the whole available range. A hair dryer and some freeze spray will help ensure that it works over the temperature range you care about. (Don't blast it too hard with the freeze spray, or you're liable to cause cracks in some of the components.)

The crystal wants a 32-pf load capacitance, so your capacitor values are probably too low. The load capacitance seen by the crystal is basically C1 || (C2+Cin), where Cin is the input capacitance of the oscillator chip, so you probably want something more like 56 pF for C1 and 68 pF for C2, which works out just about exactly with Cin=6 pF (typical value from the datasheet). You can adjust the gain by trading off the C1/C2 ratio.

The datasheet says to try the circuit with the crystal replaced by the equivalent series resistance (200-300 ohms in your case), which is a good idea.

You might very well need to increase the value of Rs since you're running at a low frequency, where the chip's gain is high.

If you need to adjust the frequency to be exactly right, you can reduce the tank capacitor values a little (say 56 and 56 pF) and put a 1-5 pF trimmer cap in parallel with the crystal. Watch out for the capacitance of the pads and traces, which also have to be factored in, and don't use a trimmer that's anywhere near the value of the tank capacitors, or you'll have problems.

Cheers,

Phil Hobbs

That doesn't make any sense, the parallel mode crystals are far more popular than series. Brush up on your search skills. The two types are identical in construction but tuned differently. Series mode crystals are designed for circuits that resonate at the series frequency, where the crystal appears as a pure resistance of low value and is loaded by a low resistance load at its output. There is 0o phase shift through the crystal at this frequency, meaning the buffer is non-inverting. The parallel mode crystal is intended to be loaded by a high impedance, it is combined with other reactances to produce 180o phase shift and requires an inverting buffer. You can go ahead and stick the series resonant crystal in the parallel mode circuit, but the frequency will be off. Otherwise, stability and immunity to spurious oscillations should be the same, if you're having these problems with a series mode crystal, you will also have them with the parallel mode type, in the same circuit. The usual method of eliminating the spurious and overtone response is to RC filter the gate drive into the crystal, hence the popularity of the Pierce configuration.

Apparently you have a knowledge gap here, those are parallel resonant mode crytals, this is inferred from the "load capacitance" specification.

There is technically no difference between a parallel and series crystal. Each crystal operates at or near its (series) resonance point. If the oscillator circuit is designed to look capacitive to the crystal, it will operate on the high, inductive side of it resonance point. The (bad) terms parallel mode crystal or parallel mode oscillator refer to this capacitive load aspect. The crystal manufacturer incorporates this by manufacturing the crystal a bit lower in frequency so it will operate at the rated frequency using a certain capacitive load. The proper load is specified in the data sheet.

The popular Pierce circuit is such an oscillator. Like Phil said, start with about double the specified load capacitance as values for C1,C2 and tune (one of) them if you really need to be exactly on frequency. Your Rs of 1K is a good starting point to keep the crystal power below the 1mW of the data sheet.

Joop

If it works, I would not worry too much. The crystal types are basically the same. The only difference is perhaps that the frequency of a series crystal might be a bit high in a Pierce circuit. How much accuracy do you need? About your problems in the past, this might have been with cheap computer crystals that may have had a bit high series resistance. Increasing the drive level could probably have fixed that. In general keeping the drive level low enough makes sure the crystal survives forever and does not drift much. Making it too low might give trouble starting or make the oscillator a bit noisier. What is your application?

Joop