gEDA-user: challenge board ALMOST works

[Dan McMahill]

DJ Delorie wrote:

> However, the two oscillators tend to self-synchronize.  If I let it
> run, one LED is on and the other is off.  If I put scope probes on the
> buffered outputs (pins 6 and 8), I can see that the oscillators are
> running, they're just in lock-step.  Not even out of phase.
> 
> If I put a scope probe on either feedback line (pins 1 or 13), they
> drop out of sync and the two LEDs pulse back and forth (about 17Hz).
> 
> Also, if I put probes on both xor outputs, they drop out of sync, too.
> Sometimes they stay unlocked if I remove the "on led" led's probe but
> leave the "off led" probe there.
> 
> Ideas?  The only thing I can think of is inductive or capacitive
> coupling, but at 400 Hz?

Welcome to the world of injection locking.  That's the name for what 
you're seeing.  There have been a number of papers published describing 
conditions in a good bit of detail.  Some TV's used injection locking 
for some of the sync circuits.  I see papers with injection locked 
frequency dividers from time to time in IEEE JSSC.

The bottom line is an oscillator can be extraordinarily sensitive to 
signals around its fundamental or harmonics and subharmonics of that. 
It takes a very small amount of coupling.

My guess is that it is supply coupling and probably the culprit for much 
  of the supply coupling is the internal layout of the inverter chip. 
You may be able to get some relief by making sure you have a really low 
impedance at the supply pin at 400 Hz (and several harmonics of that) 
but I'll bet thats not quite enough.

Unfortunately, this circuit wants the two oscillators to be very close 
in frequency to get visible flashes, but that is also the very recipe 
for injection locking.

I guess you could try using 1M and 0.001 uF with the idea of reducing 
currents in the output driver.  I wouldn't bet the farm on that but you 
might get lucky.  And just to cast some doubt on this having any chance 
of working...  I'll bet that what happens is, both oscillators are 
nearing a transition.  The first one to get there causes some fairly 
fast edges and it glitches the internal supply.  Even with 0 ohms 
presented at the supply pin, you still have bondwire inductance and 
resistance and on chip inductance and resistance.  This glitch is likely 
enough to wiggle the threshold enough to cause the 2nd oscillator to 
switch too.  And now both oscillators will hit the next transition at 
about the same time.  The first one to get there will cause the 2nd to 
transition.

I think the only real cure is 2 packages with attention being paid to 
decoupling their supplies from each other.


-Dan