"Bill Bowden" <wrongaddr... at (no spam) att.net> wrote in message

news:3eb67024-6caf-48c6-90b6-0a34a80e1319 at (no spam) y22g2000prd.googlegroups.com...



I made a low frequency crystal oscillator using a couple CMOS
inverters (CD4069) and a very old 31.5KHz crystal. It's a fairly large
crystal package that measures about 3/4 by 3/8 inch from the 1970s.

The first inverter feeds the second, and the output of the second
connects the crystal back to the input of the first inverter. The
first inverter has a 100K resistor connected from input to output and
there is a 470pF cap from the input to ground. Works well, but I'm
wondering what the best RC values should be? I was thinking the
reactance of the cap should be about equal to the resistor, but that
works out to about 50pF which doesn't work. Lots of different
combinations will work, such as 22K and 1000pF.

How does one determine the best values?

-Bill

The resistor controls loop gain. I would find the lowest value for which
oscillation starts reliably and multiply it by 10.

The loading capacitance will affect frequency. Since you are using a
crystal, presumably you care about frequency (if not, you could use a simple
RC oscillator) so choose the capacitor which gets you closest to 31.500 KHz.

I was recently playing with miniature 32.768 KHz tuning-fork watch crystals,
which require extremely low drive levels; but I presume your large 1970s
crystal is not of this type.

On May 10, 2:38 am, "Andrew Holme" <a... at (no spam) nospam.co.uk> wrote:
"Bill Bowden" <wrongaddr... at (no spam) att.net> wrote in message

news:3eb67024-6caf-48c6-90b6-0a34a80e1319 at (no spam) y22g2000prd.googlegroups.com...



I made a low frequency crystal oscillator using a couple CMOS
inverters (CD4069) and a very old 31.5KHz crystal. It's a fairly large
crystal package that measures about 3/4 by 3/8 inch from the 1970s.

The first inverter feeds the second, and the output of the second
connects the crystal back to the input of the first inverter. The
first inverter has a 100K resistor connected from input to output and
there is a 470pF cap from the input to ground. Works well, but I'm
wondering what the best RC values should be? I was thinking the
reactance of the cap should be about equal to the resistor, but that
works out to about 50pF which doesn't work. Lots of different
combinations will work, such as 22K and 1000pF.

How does one determine the best values?

-Bill

The resistor controls loop gain. I would find the lowest value for which
oscillation starts reliably and multiply it by 10.

The loading capacitance will affect frequency. Since you are using a
crystal, presumably you care about frequency (if not, you could use a simple
RC oscillator) so choose the capacitor which gets you closest to 31.500 KHz.

I was recently playing with miniature 32.768 KHz tuning-fork watch crystals,
which require extremely low drive levels; but I presume your large 1970s
crystal is not of this type.

Thanks for the reply.

Yes, the little 32768 tuning fork watch crystals work well with little
drive. I'm just playing around with this old "heavy duty" crystal, I
really don't need to use. My understanding of crystals is they look
like a high Q parallel LC circuit at resonance, and to excite them,
they just need feedback at 180 degrees. But that doesn't seem to work,
and some phase shift is needed for reliable operation, and so the
addition of the RC components to shift the phase slightly so the
feedback is not exactly 180 degrees.

Using 100K and 470pF yields a frequency of 31493 which is only a 7
cycle error out of 31500.

On Sat, 10 May 2008 19:16:49 -0700 (PDT), Bill Bowden



wrongaddr... at (no spam) att.net> wrote:
On May 10, 2:38 am, "Andrew Holme" <a... at (no spam) nospam.co.uk> wrote:
"Bill Bowden" <wrongaddr... at (no spam) att.net> wrote in message

news:3eb67024-6caf-48c6-90b6-0a34a80e1319 at (no spam) y22g2000prd.googlegroups.com...

I made a low frequency crystal oscillator using a couple CMOS
inverters (CD4069) and a very old 31.5KHz crystal. It's a fairly large
crystal package that measures about 3/4 by 3/8 inch from the 1970s.

The first inverter feeds the second, and the output of the second
connects the crystal back to the input of the first inverter. The
first inverter has a 100K resistor connected from input to output and
there is a 470pF cap from the input to ground. Works well, but I'm
wondering what the best RC values should be? I was thinking the
reactance of the cap should be about equal to the resistor, but that
works out to about 50pF which doesn't work. Lots of different
combinations will work, such as 22K and 1000pF.

How does one determine the best values?

-Bill

The resistor controls loop gain. I would find the lowest value for which
oscillation starts reliably and multiply it by 10.

The loading capacitance will affect frequency. Since you are using a
crystal, presumably you care about frequency (if not, you could use a simple
RC oscillator) so choose the capacitor which gets you closest to 31.500 KHz.

I was recently playing with miniature 32.768 KHz tuning-fork watch crystals,
which require extremely low drive levels; but I presume your large 1970s
crystal is not of this type.

Thanks for the reply.

Yes, the little 32768 tuning fork watch crystals work well with little
drive. I'm just playing around with this old "heavy duty" crystal, I
really don't need to use. My understanding of crystals is they look
like a high Q parallel LC circuit at resonance, and to excite them,
they just need feedback at 180 degrees. But that doesn't seem to work,
and some phase shift is needed for reliable operation, and so the
addition of the RC components to shift the phase slightly so the
feedback is not exactly 180 degrees.

Using 100K and 470pF yields a frequency of 31493 which is only a 7
cycle error out of 31500.

---
Crystals can look like either series or parallel resonant circuits,
depending on how they're driven.

Here's a pretty good explanation:

http://download.intel.com/design/mcs51/applnots/23065901.pdf

JF

I was recently playing with miniature 32.768 KHz tuning-fork watch crystals,
which require extremely low drive levels;

On Sat, 10 May 2008 10:38:20 +0100, "Andrew Holme" <ah at (no spam) nospam.co.uk
wrote:

I was recently playing with miniature 32.768 KHz tuning-fork watch
crystals, which require extremely low drive levels;

Hi

Maybe you can hep me. I tried a 32.768 with 8051 microcontroller. I
used 11 pf capacitor and a 330k resistor in series and it worked well
on the breadboard. But when I made a pcb and soldered all the parts
exactly as was in the breadboard the crystal refused to oscillate.

What is missing?

Thanks for your time and attention
Archer

probably another 20-30 pf...

try something around 68 pf as a start.