Sinusoidal oscillator configuration

A minimal hardware circuit could be a tiny PIC micro with an RC filter

- 3 SMD parts. Depends on the specs you actually need of course.

Dave.

I would be surprised if a PIC generating a square wave followed by an RC filter and then a power amplifier and antenna would come anywhere clear to meeting FCC spurious emissions requirements...

I think the O.P.'s suggestion of using a regular oscillator circuit might be a better start -- if the oscillation can start up slowly, it only needs a little more than unity gain, so the sine wave can be relatively pure. It might still need some filtering after the power amplifier depending on just how much power is being used and of course how linear the amplifier is... although at 125kHz this might not be too hard. Still, traditional L-C filtering high power at

125kHz isn't going to be much fun (large components), so keeping it out of the amp in the first place is probably preferable.

If they don't have 125kHz as a standard frequency, you may have to get a bazillion of them before you can get them at a reasonable price.

Assuming that you can get them* the traditional Pierce-ish inverter and two caps oscillator could, indeed, be expected to give you something like a sine wave at the input and something more like a square wave at the output, with the crystal filtering things as they went through from output back to input. You'd have to try this out on your circuit to see how reliably it generated a nice sine wave.

Alternately, you could use (gasp!) discrete components to build a BJT or JFET Colpitts oscillator. This gives you a heck of a lot more control over the critical oscillator parameters, so if you're interested in wringing the most out of your oscillator this may be the way to go. With one of these oscillators you can use very light capacitive coupling from the hot end of the crystal to get a nice sinusoid.

Whatever you do you'll have to make a tradeoff between power output, circuit complexity, component sensitivity and the pureness of the sinusoid. Anything that puts the oscillator pick-off at a point where the signal is a nice pure sinusoid requires that you take little power from the oscillator, which in turn means that you must follow your oscillator with a healthy amplifier, which adds to the circuit complexity.

Were I doing this I would investigate using a cheesy DDS. This sounds fancy, but all you do is use an oscillator running at some multiple of

125kHz followed by a divider driving a resistor network to make a sorta-sinewave that drives an op-amp filter. Something like a 1.25MHz oscillator driving a 4017 would let you hand-pick your ten resistor values to really stomp down on the first five harmonics, which would leave you only needing to filter above 500kHz or so.

• Check Digikey -- many folks have the rule of thumb that if Digikey stocks it then it's probably a standard item. Beware of Digikey part numbers that have no stock behind them -- those are either special buys for someone, or they've been discontinued.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

Using a "cheesy" DDS sounds like the solution I am trying to replace actually. There is a 4 MHz crystal going through a 4060 counter, which brings down the frequency to 125 kHz (square wave at that point), and is then fed into an RLC network to filter out the harmonics, giving a pretty clean sinusoidal. My new design doesn't require a 4 MHz clock and the counter is pretty bulky and generate a square wave incurring the need for the RLC network. This solution isnt bad by any means, I am just investigating different ways of accomplishing the task with the goal of minimizing the hardware.

This is a learning process for me, so I decided to inverstigate the Pierce configuration further. I have a 125 kHz crystal, so thats not a problem. However, I cant get it to oscillate in this configuration.

Schematic:

I am using the following crystal: ?name=3DXC999-ND with 4.7 pF caps at C1 and C2 (also tried 12 pF) and 10M for R1 (also tried 1M) and its just not working. The circuit is very simple, so I am completely lost at what I am doin wrong here. Are my C1 and C2 values wrong? Do I need a series resistor between the crystal and inverter output?

Thanks everyone for the very informative replies!

-Nick

df

e . .

d text -

Howdy,

I think your capacitor values need adjustment. You may find this article helpful

Note that the series combination of the two capacitors should equal the required load capacitance of the crystal and that increasing the value of the capacitor on the gate's output increases feedback. If the crystal is a bit sluggish you'll want more feedback.

If you have a jfet handy, like the mpf102, you'll find it works much better than a gate. Though when a spare gate is handy I'm likely to use it.

Also see,

This is excellent,

Armed with these I'm sure you can get it working if the crystal is functional.

I've looked at the Fairchild paper before and overlooked the series combination part, in turn making an error in my load capacitance calculations.

The load capacitance for the crystal is 12.5 pF. So C1 and C2 should be 25 pF. I tried 24 pF (two 12 pF in parallel) for C1 and C2, no luck. I tried giving it more feedback, as you have suggested, by putting ~27 pF at the output (22 pF in parallel with 4.7 pF), no luck.

I also tried to take into account the gate input and output capacitance (assumed to be around 5 pF) by placing 17 pF (a 12 and a

4.7 in parallel) as C1 and C2, still no luck.

I also have a JFET, a 5457, so I attemped to put the crystal in the following configuration:

with 10k for Rd, 10M for Rf, no Rl, no buffer at the output, and 24 pF as C1 and C2, no signs of life. I also tried, J201 and a PN4392 jfets, still no luck.

At this point it seems like I am playing trial and error, and thats pretty disturbing in my mind. =3D(

Thanks,

-Nick

df

If that is good enough performance-wise for the existing circuit then simply replacing the counter and crystal with a 125KHz square wave output from a tiny 50 cent 5 or 8 pin PIC micro is a good solution to minimise your hardware. Just one tiny chip and the existing RLC filter. If you use the smallest SMD parts you can get then the circuit footprint will be very small indeed. Double sided component load would make it even smaller. I doubt you'll get a smaller solution using more traditional means.

Dave.

Howdy,

Do you have a URL for the crystal specification? I'll look for my gate oscillator spreadsheet and drop the values into it.

Have you tried more than one crystal?

Remember that the gate should be biased into the linear operating range to get things started. Sometimes a small resistance in series with the gate output is necessary in addition to the 10M-20M feedback resistor.

Then the pi section (the two caps and the crystal) provides

180 degrees of phase shift, which when added 180 degrees from the inverting gate satisfies the 360 degree round the loop requirement. The other requirement is loop gain greater than unity, but the gate has more than enough gain to compensate for the loss across the pi network.

With low frequency crystals (~100KHz and below) the gate oscillator usually works, so this is a puzzle. If you have a 4049 inverter I'd try that. They make vigorous LF oscillators. With the 4049 use 10M feedback 49pf and 56pf on either side of the crystal (highest value on the gate's output side) Also use

39pF between the gate output and the pi section (where some application notes show a few K-Ohms of resistance.) Of course the 10M bias resistor must be directly connected from input to output of the gate to force it into linear operation.

I have a pierce circuit with an inductor insted of a resistor in the drain(2.5mH because it's what was handy.) Every LF crystal or ceramic resonator that I've tried, mostly salvaged from computer junk, has oscillated if it was ever going to oscillate. I don't like the app note oscillator for LF crystals. The pi section capacitors and feedback resistor are too low in value and the drain resistor would be better replaced by an inductor. But I can't say it wouldn't work.

Don't be too hard on cut and try (trial and error) somtimes a person stumbles onto something wonderful via that path.

g

.pdf

o t a

see

r l g e

ty.

s z

ys.

ed text -

I still think Tim's solution is the way to go. Isn't such a low frequency crystal expensive? In any event, he is not suggesting creating a square wave, but rather a pseudo-sine-wave that will have much reduced harmonics. I did a sine wave generator for telecom apps by simply creating a square wave with a flat spot in the middle (i.e. reduce the size of the jump), then filtering the signal with a switched capacitor filter. There are Walsh based circuits to make pseudo sine waves, but you would need to watch out for patent issues.

If you want to use a dac and can make the sample rate a multiple of the desired frequency, then you just need a simple look-up table. If that timing cannot be achieved, a coordic can be used to compute the sine values. That scheme is stone age enough that there should be any patent issues. The coordic would have the advantage of being able to change the frequency independent of the clock, i.e. you could make

125KHz and 13KHz with the same circuit.

In any event, if you want a clean signal, it is better not to generate the harmonics in the first place rather than to filter off the harmonics.

On a sunny day (Sat, 30 Aug 2008 17:36:27 -0700 (PDT)) it happened Nick Zalutskiy wrote in :

I have used 100kHz xtals in circuits like this: ftp://panteltje.com/pub/xtal_osc_125_khz.gif

The fet I used was likely a BF245 (high gain). I cannot run simulation on this diagram without more info on the xtal. Your solutions use the xtal in series resonant mode. This one uses parallel resonant mode. Give it a try perhaps. The waveform at the source should be almost a sine already, with some flattening at the bottom. In my applicaton startup was really slow...

You can also add a drain resistor of a few hundred ohms, and take the signal from there. Adapt the capacitors perhaps, ratio 1:2 is normally good.

might help on the filter requirements to use two pins and two resistors go get a three step wave form:

_____ _____ _____ pin1 |_| |_| |_ _ _ _ pin2 _| |_____| |_____| |___

_ _ _ combo _| |_ _| |_ _| |_ |_| |_| |_

adjust overlaps to get minimum harmonics

-Lasse

Nice idea!

Dave.

e

ich

nd

a k g I d

The magic number is 60 degrees to get no 3rd.

If your micro has tri-stated pins, you can do much the same by using the tristate condition to make the flat spot in the edges.

yes, but using two pins might make it possible to setup two timers to generate the two waveforms autonomously. Don't think I have ever seen a timer that could tristate a pin on a compare

-Lasse