Dividing a 32768 Hz crystal frequency

I usually see an unbuffered inverter used, and then you have to mess with drive levels and load caps. Unless power is an issue is it worth he hassle when you can get an oscillator you know will work for a dollar?

just the right amount of "flat spot" at the zero crossing (aka. "modified sine") will take of much of the third and fifth harmonic

To get 10kHz/(2*pi), why not start with a 13MHz or 26MHz xtal oscillator and divide by 8192 or 16384? (-2192 ppm) - that is closer than 32768/20 (+29437 ppm).

Actually, a CD4017 gives a ten-divide, with one-of-ten driven outputs, and a 4013 does a two-divide, AND... a LM13700 section can (in conjunction with nine resistors connected to 4017 output pins) take a half-wave sinusoidal current into its current-program pin. The divide-by-two after the divide-by-ten can be wired to the differential input of the LM13700 to change the current sense every halfcycle so the LM13700 output is the right twenty-point current fit to a sine.

With +/- 12V supplies, all the CMOS on +12, it takes a level translator transistor (PNP grounded base) to drive the LM13700 input correctly, with resistors (from 4017 outputs to PNP emitter) being "0" output: open (infinite resistance, programs LM13700 for zero current out) "1" output: 12000 *(sin(18degrees)) ohms = 3708 ohms "2" output: 12000 *(sin 2*(18)) = 7053 ohms "3" output: 12000 *(sin 3*(18)) = 9948 ohms

... and so on. Twenty-point fit to sinewave, using nine resistors.

A load on the LM13700 output (1k ohm to ground) will keep the output biased correctly, then you connect an LC tank filter.

There's uncommitted darlingtons and another section of the LM13700 free, you can buffer (or maybe oscillate) with that.

I had to sleep on it to remember the part but the 4060 s probably what you had in mind

You have mentioned that you want to end up with a decent sine wave and the 4017 does offer one way of getting close to that.

If you hang ten resistors - one on each of Q0 to Q9 - and feed them all into a summing junction - you can get a ten-step step-wise approximation to a sine wave, and a low pas filter can clean out the higher harmonics pretty effectively.

Using the outputs to drive 4016 transmission gates can give you better controlled step heights.

Buying an Analog Devices Direct Digital Synthesis Chip give you many more steps, but isn't all that cheap.

and I'm only looking a power of two divisors.

Depending on how much is not much then for a spot frequency integrating it once and diode shaping followed by a filter will get all the higher harmonics well down. HP patent on this trick is long out of date.

I'm curious - what is special about 1591.55 Hz ?

To do that you actually want to divide 32768 by (almost)

More likely how your news client is set up. Thunderbird has "Reply" (to author) as one of its buttons.

Updates sometimes promote it to being the default action too!

-------------------------------------

How about a one ( cheap) IC, linear RC oscillator with near perfect sine output - allows adjustable f to spot on your desired number. Using 1% MF resistors and polystyrene caps, temp stability can be suprisingly good.

BTW nice to see you back.

......Phil

Thanks, Martin and all the others who have given helpful advice --- too many to reply individually. The project is tutorial in nature, so I don't want to use too many 'integrated' fixes, and SMD-only devices are not an option. Regarding the frequency, it does need to be near 10k/2pi, but other things can be adjusted to suit the exact frequency, which needs to be stable within ±1% and not require a counter to determine it.

I will try 'Reply to list' instead of 'Reply' to see if that works.

OK. How about a Wein bridge sine wave oscillator instead then? eg.

Some years ago, we had a George Herold thread on making sine waves. The eventual solution was a 4017 with resistors forming a weighted sum of the outputs. This cancels the second through ninth harmonics, making filtering much easier.

Cheers

Phil Hobbs

-----------------------------------------

** OK - we finally have an actual " spec" for your 1591 Hz sine wave oscillator.

This simple topology will do the job very easily:

THD is about 0.15% and frequency can be trimmed by adjusting one or both RT values as shown in fig 2. Amplitude stability depends only on the tempco of the 4 diodes.

Essentially it is my design and many hundreds have been built. Way better than a Wein bridge topology since there's no need for ( now unobtainable) thermistors, tiny lamps or fussy FETs.

.... Phil

Why the requirement to not use a counter??? I don't know what that means in this context, since you are proposing on using a counter. Can you explain?

"not require a counter to determine it" I would assume that means it needs to be on frequency as build so you don't a frequency counter to verify and adjust it

Looks very good, but too complicated for my project.

Yes, correct.

Is ther any way of retrieving that interesting information?

So, you are saying no oscillators tuned by an RC, for example. Ok, that's not hard. What you are talking about, using a crystal and a divider should work.

This thread is very long at this point. What are the problems? You seem to reject a lot of solutions because they are too complex. I don't know what to recommend at this point. You seem to be looking for a Goldilocks solution. Maybe you could summarize the solutions that come closest at the moment?

I really don't want any more help, with thanks to all who have tried. Many of teh suggestions open up other streams of thought.