Good design guide for oscillators

Depends on the load, but you might not need the buffer - I once did a variant of the (Peter) Baxandal class-D oscillator where ALC loop kept the output voltage very stable (way better than 20% - how much better I don't know, because I couldn't measure any variation).

------------- Bill Sloman, Nijmegen

RF? Randy Rhea's book is good. But it sounds like you need a low Zout buffer amp after the oscillator.

John

I think Hayward does a very respectable job discussing filter design as well, by emphasizing the idea of coupling between stages. At least at the Universities I've attended, filter design was taught with a much more mathematically based approach, and while this works fine for various 'cookbook' filter designs, it makes it difficult to understand how you'd go about empirically tuning 'non-traditional' filter mechanisms such as, e.g., coupled waveguides.

(Hmm... Randy Rhea is also a guy with good books on both oscillators and filters...)

---Joel Kolstad

Hi,

Any good URLs on LC oscillator design? I need an oscillator where the load varies 10:1 but the voltage doesn't (20% would be ok). And harmonics at -60dB.

TIA

--

Malcolm

Malcolm Reeves BSc CEng MIEE MIRSE, Full Circuit Ltd, Chippenham, UK ( snipped-for-privacy@fullcircuit.com, snipped-for-privacy@fullcircuit.co.uk or snipped-for-privacy@iee.org). Design Service for Analogue/Digital H/W & S/W Railway Signalling and Power electronics. More details plus freeware, Win95/98 DUN and Pspice tips, see:

NEW -

I read in sci.electronics.design that Malcolm Reeves wrote (in ) about 'Good design guide for oscillators', on Sun, 23 Jan

I don't have any URLs, but you solve the first one with a good buffer, so the osc never sees the varying load, and you solve the second one with a well-designed automatic level control (ALC) loop.

It wouldn't be a bad idea to tell us what the frequency is and whether it has to be varied (a lot or a little).

the

vary

You can't directly load the tuned circuit:

w (omega) = 2*PI*f = 753982 Reactance XL = w*L = 181

Worst case Q = 500/181 = 2.76 which is far too low.

You need a buffer stage: a FET source-follower might do it.

How accurate does the 120 KHz need to be? Can't you get near enough with the CPU? An LC oscillator will only get you within a few percent at best.

Is this for X10 by any chance?

to

timer

What CPU is it?

"op-amp oscillator", that's what I'd do. Use the BP version of my gyrator-type filter (on my website)... you can get tremendous Q's.

...Jim Thompson

DOh! Frequency is 120kHz. I already have a LC filter output stage and I was going to get the CPU to produce the frequency then filter it, but the CPU oscillator is all wrong and I wonder since I already have an LC whether I can't build that into a colpitts and so simplify the circuit. Current L is about 240uH but if I go custom I could vary that. Load across the L is about 500-5k.

The other alternative I have is to stick with my current output stage and replace CPU clock and filter with op-amp oscillator (same PCB area) to give 120kHz sinewave. I think that might be easy (more certain) to achieve -60dB harmonics. But the colpitts sounds like less components, if it can be made to work.

Which is what I was thinking, unless there is something really clever :-).

The LC output stage is adjustable so there is an advantage to have just one adjustment. Drift should be less than 1% and 1% (or more) would be ok. CPU frequency would 125kHz at the nearest. But I could use external resonator. However CPU only has clock out so I'd need to divide it otherwise CPU has no time for anything else. Internal timer just generates an interrupt and running that at 240kHz is just not on :-).

So CPU route is add resonator, divider, then filter (-60dB) to give sinewave. Alternate is generate a sinewave, which sounds like it might take less parts, if it can be done.

Yes, but in UK to meet EN standards which specify output levels and maximum harmonics.

Wes Hayward's book : Introduction to RF Design : discusses the issues involved in LC oscillator design, and the principles hold at all frequencies. It enabled me to know what to look for when tweaking oscillators on the bench, design my own, and make sense of other oscillator literature. Oscillator design is so un-intuitive that many engineers I respect do not grasp the basics as set out by Hayward.

You can ALC your oscillator or its buffer amplifier to get constant output, or depending on impedances and power levels, you can use a low output impedance or negative feedback or swamping resistance.

These days, straight LC oscillators are not terribly popular - mostly they are built into PLLs, or replaced by DDS.

Roger

One of the PIC range, and yes an alternative is a different PIC with a PWM output.

We're drifting off topic a bit. I know why. The most logical design would seem to be CPU clock, to 120kHz SQ, filter to 120KHz sinewave. But my question was about LC filters incase there was a simpler solution I was missing.

Hello Roger,

In part, the reason for that might be that younger engineers feel very uncomfortable rolling their own oscillators. I have many many who don't even know the difference anymore between buffered and unbuffered logic chips for oscillators.

The best example was a client design that contained several 'crystal oscillators in a can' and they had a tough time reducing cost. So I replaced them all with versions where they just had to buy the bare crystals. The engineers were very uneasy about that at first but happiness set in when the bill of materials came in just under target cost. This anxiety is understandable, considering that they did not really learn practical oscillator design at university.

Regards, Joerg

Hi Malcolm,

We have X10 and all I can tell you is that LC in our case is pretty marginal. They drift a bit much.

I'd consider a crystal, for example the usual 3.57MHz divided by 15, then by 2. This would result in 119,32 kHz. 6MHz crystals are even better and popular at least in the US, for under $1. Divide by 25, then by 2 and bingo you get 120kHz at an even duty cycle. If you can't sacrifice timer circuitry a little logic chip should do.

Regards, Joerg

Yeah! Randy Rhea wrote the book on oscillators...