Low Frequency 3rd Overtone Crystal Osc

AFAIK all low frequency "crystals" are made with tuning forks. Yawn. 40Kc oscillator with non-linear multiplier for even multiples, with parametric multiplier for odd multiples. Yawn.

Sure, always buy your xo's, tcxo's and ocxo's. The fact that I have been the analog ASIC designer at Rakon for nigh on 9 years, has absolutely nothing to do with my recommendation :-)

-- Kevin Aylward

- SuperSpice

I don't know. Is that a trick question?

--

Rick C

*** The third overtone will not be exactly three times the fundamental frequency . It will be close, but the difference is temperature dependent which is the basis for MCXO design (oscillator which runs at fundamental and third overtone simultaneously)..... If you need EXACTLY three times the fundamental, use a multiplier.... Usually, tho, the third overtone will be close enough at room temp,,,, typically a few hundred cycles off....

What exactly are you trying to achieve ?

There are a lot of LF-dicipled (16, 60, 77.5 kHz) radio clocks that runs for year(s) with a single AAA battery. These clocks run with reasonable accuracy during the day, hen no on-air reference is available.

One alternative would be running a 120 kHz free running 120 kHz oscillator disciplined with a 32768 Hz crystal.

But so many wonderful things come out of KC -- barbeque, methamphetamines.... oh, well that might explain your reluctance for their crystal. )-;

Ah, but in a simulation you shouldn't expect to measure it anyway; you can't do it on AC Analysis because it's nonlinear, and you can't do it on Transient because it takes forever, at super-high accuracy besides (like RELTOL < 10^-4, and > 10^4 cycles of run time).

Better to wave the hands; if you know the CRLC parameters of your crystal, you can calculate how far it will pull. Reality won't match exactly, for all the usual reasons, but it should be similar at least.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Certainly any crystal would be pullable to compensate for initial accuracy, temperature (small ranges you would find in an occupied dwelling) and aging. For a reasonably good crystal that would amount to less than 100 ppm. The question is what does it take to pull that far?

I found some varactors that will give a 10:1 capacitance range without requiring more than 3.3 volts. But I have no idea how much that will change the frequency of the crystal I can't get anyway. Lol.

How would you calculate that from the crystal parameters? How much can you trust the data the crystal makers provide? I mean, will it be accurate enough to calculate this? I guess the calculations are just the typical filter calculations with the given topology of the crystal and load caps?

--

Rick C

Just wonder if having a strong 120kHz presence is wise if trying to measure a low level 60kHz signal? Would it be better to have a more complex harmonic relationship between high level oscillator and low-level input?

piglet

The 120 kHz is the sampling clock so I'm not worried about it getting into the signal. It gets sampled right back out perfectly.

--

Rick C

An HF fundamental crystal might easily pull 100 ppm, but an third

Are you going to use the crystal in parallel or series resonance mode? IIRC, overtone crystals are often specified in series resonance.

Pulling a parallel resonance can be done with a parallel capacitance, while a series resonance may need a tunable series inductance.

Such low frequency crystal may need a lot of extra capacitance, how much would the maximum capacitance be for that varactor ?

Oh... Usually one wants a *stable* oscillator. i.e. one that doesn't change in frequency over temperature and supply voltage changes.

Ok. This is the way you design and simulate xtal oscillators in spice.

You De Q the xtal.

This means using a c1 (the series capacitance) that is say, 100-1000 times larger than it is in reality. It is absolutely important to keep the esr the same, to keep the loop gain and phase the same at the oscillation point. Let the model calculate the inductance. Even at 1000s larger, results are usually accurate enough.

Typically, make a paramertised .subckt like:

.SUBCKT XTAL 0 1 c1=1p esr=200 f0=1M c0=2p

• Cp 1 2 {c0} Ls 1 3 {0.025330295/f0/f0/c1} Cs 3 4 {c1} Rs 4 2 {esr}

• .ends

Where the top parameter line is settable on the schematic (typical defaults shown). This allows you to easily check say, for variations in esr that will stop it oscillating.

Note that increasing c1 will shift the frequency a tad by c1/2(c0 + CL), CL being the series value of the caps on either side of the xtal, c0 is the parallel cap inherent across the xtal terminals.

Typically, for a fundamental 10MHz xtal, c1=5f, rs=50 cp=1p. A 3rd overtone xtal might have a c1 of 0.3ff, so much higher Q. Also, typically, a 3rd overtone is 10% off from 3X. It is never exactly 3X.

DeQing keeps some characteristic of the oscillator the same, as in gain and phase and whether it will oscillate or not. However, lowering the Q, multiplies up things like capacitor tuning pulling sensitivity and voltage sensitivity, by Q. That is, if you change the supply voltage on the deQed oscillator, the frequency will change by Q times what it would have done at full Q. The transient response will also be Q times quicker. This allows fairly accurate estimation of pullability, and startup time in the full Qed circuit by simply dividing changes by the DeQ factor.

There are some Spice's that have an inbuilt "calculate frequency" of the time domain trans run in its TRAN setup GUI :-)

An example 3rd overtone SC cut xtal from the SuperSpice lib is:

.SUBCKT XTAL3rd_XN !0 !1 cp=1p f0=35M cs0=10f rs0=20 f3=100M cs3=4f rs3=40 f2=110M cs2=3f rs2=30

• _SS_Symbol [Discretes.ssm] [Crystal] V!1 !1 2 0 V!0 !0 1 0 Cp 1 2 {cp} Ls0 1 3 {0.02533029591/f0/f0/cs0} Cs0 3 4 {cs0} Rs0 4 2 {rs0} L3 1 5 {0.02533029591/f3/f3/cs3} C3 5 6 {cs3} R3 6 2 {rs3} L2 1 7 {0.02533029591/f2/f2/cs2} C2 7 8 {cs2} R2 8 2 {rs2} .ends

Note the nominal unwanted B mode frequency being close to the wanted f3

100MHz C mode. Typically, a mode trap is required to supress it. A mode trap is typically a series LC with parallel C, which has a pass and reject centred on the C and B modes.

-- Kevin Aylward

- SuperSpice

the frequency of the crystal I can't get anyway. Lol.

Formula for shift from fs, and spice techniques given in my other post.

-- Kevin Aylward

- SuperSpice

OK. What is your initial frequency accuracy and how much do you want to pull it?

Has anyone suggested ceramic resonators yet? There are loads of cheap

480kHz ceramic resonators and they will be much more pullable than quartz, and start-up quicker too.

piglet

If it helps at all crystals with a fundamental at 5.0 Mhz are extremely common, and while a 5.120 Mhz fundamental isn't a very well-known value, my usual outlet Mouser has around a thousand of them at about 20 cents in small quantities (it allows binary division to 10kHz.)

Might be able to work with that?

Sorry, they're "on order" but the ship date is next week.

Thanks, saved for future reference - I could never seem to get xtal oscillators to sim correctly in LT

What are the LEDs for? People like bling so it's good...

It helps spending the last nine years as the analog asic designer at Rakon

-:)

Email me, and I will send a "complementary" SS password file, and an additional oscillator example, that has been set up to automatically plot AC loop gain and loop phase along with its transient run.

I will try and set up a 3rd overtone SC cut with LC mode trap, example as well. I have tons at work, but I will have to do a custom, generic one for inclusion in the SS libraries, for obvious reasons.

-- Kevin Aylward

- SuperSpice

I didn't see this in the thread -- why not a 1.5MHz or 3MHz crystal and a divider?

I simply don't know if the tuning fork crystals will work nicely in overtone, or if they can be rubbered. I do know that with an AT-cut crystal, the range of frequencies over which you can rubber the crystal goes down dramatically if you're running an overtone oscillator.

Even with an AT cut running in fundamental, don't expect to get a well- behaved oscillator that varies by more than +/- 500PPM of the center frequency. You can push it farther, but the farther you push it the more it becomes an LC oscillator that's decorated with a crystal and the less it stays a crystal oscillator.

Where the overtones are in a crystal are is a combination of electrical and mechanical effects, and to some extent the near 1:3 ratio (it's not exact) in an AT-cut crystal depends on the simplicity of the geometry of the cut (it's just a disk, the thickness of which determines the frequency). I could easily imagine a tuning-fork cut to have overtones that do not fall on convenient ratios, or that vary considerably from manufacturer to manufacturer, or even lot to lot.

--
www.wescottdesign.com