[sdiy] Zener-oscillators

[Czech Martin]

Hi Magnus,
I understand that you want to build an oscillator with
a zener diode from shelf and a capacitor.
I think that this will not work. The zener diode
as such has a continuous V/I curve, even in avalanche
region. It has no current amplification, hence no snap 
back or other switching action like the npn in avalanche
(emitter-base breakdown).
I think that the supposed circuit will only run into
the avalanche limitation - and stay there for ever.

The other statements are correct, the npn relaxation 
oscillator will have sometemperature coefficent issues
as well as leakage issues. Therefore the range of operation
is limited.

m.c.

-----Original Message-----
From: Magnus Danielson [mailto:cfmd at swipnet.se]
Sent: Montag, 31. März 2003 14:42
To: synth-diy at dropmix.xs4all.nl; micke at bmh.nu
Subject: [sdiy] Zener-oscillators


Dear All,

The reversed-transistor/minimum-component oscillator is as close to a minimum
component current-controlled oscillator core we can get - a capacitor and
a Zener-diode.

The current simply charges the capacitor and the Zener uncharges it when it
gets "lit" by the voltage of the capacitor. The Zener-diodes have a temperature
coefficient of almost 0 mV/C at 5.6V. Actually, the tempco-curve crosses 0
in the neighborhood of 5.6V since there is two different phenomenas that cause
similar effects - quantum-mechanical tunneling and avalance-reaction. These
have different signs on their tempco and for zeners below 5.6 the tunneling
effect (and the negative tempco) dominates and above 5.6 the avalance effect
(and positive tempco) dominates. So 5.6V is somewhat of a sweetspot tempco
wise.

If you put a zener across a classic sawtooth core (like the ASM-1) you can drop
the comparator for the sake of oscillation. You might want to keep the setup
for the sake of oscillator sync, but then the feedback-loop from the output of
the oscillator needs to be removed. Actually, with a 5.6V Zener almost no
changes needs to be done to the ASM-1 VCO setup, since it is supposed to kick
out a 5 Vpp ramp and now it gets a 5.6 Vpp ramp.

I haven't dug into a few issues, but they display some of the thoughts of
problems that I have come up with:

1) Temperature dependence? Solved: 5.6V close to zero. Say we have +0.56 mV/C
   at 5.6V, then we have +560 u / 5.6 = +100 u which is +100 ppm/C which isn't
   bad at all. The uncompensated exp would be far worse, and other effects
   like capacitors tempco probably overshadows this component anyway.

2) Tracking error due to constant resistive leakage?
   Can we expect a linear (resistive) leakage which needs to be compensated?

3) Tracking error due to constant current leakage?
   Can we expect a linear (current) leakage which needs to be compensated?

4) Tracking error due to non-linear current knee near threshold?
   Can we expect a tracking error since the leakage changes near the knee
   such that lower frequency (which takes longer time to reach the knee) will
   become lower?

5) Tracking error due to non-linear turn-off?
   Can we expect a tracking error due to the non-linear turn-off mechanism of
   the tunneling/avalance effects?

6) Tracking error due to discharge speed?
   Will discharge speeds be so long that they cause (classic) high frequency
   tracking errors? Solution is to be found in Franco-compensation (series
   resistor).

7) Discharge spike emission?
   Will the discharge speeds be so rapid that the current spikes cause
   problems? If so, solutions is found in slowdowns. A smart solution would be
   to use a Franco-compensation resistor to act both as slowdown/current
   limiter and frequency compensation for the extended reset-time.

8) Energy analysis - overheating of Zener?
   We store a certain amount of energy in the capacitor, this energy is then
   almost all converted into heat by the Zener-diode for EACH cycle, so for a
   100 kHz waveform this happends 100 000 times a second. We know the energy
   E = U * Q and we know that Q = C * U so E = U^2 * C. For a 5.6 V zener and
   a 2.2 nF capacitor this transforms into E = 68.992 nJ and then P = E * f
   = 6.8992 mW. So, for this case we are more then well home safe. I think we
   are home safe for almost any application, since the maximum frequency is
   expected to go down as the capacitor size goes up.

Any comments?

I have not made any real-life tests just yeat, but I do have 5.6V zeners around
so it should be a simple thing to mod an oscillator and test-fire it.

Cheers,
Magnus