Oscillator idea come's unstuck

[Magnus Danielson]

>>>>> "B" == BJ  <zzynt at swipnet.se> writes:

 B> jhaible wrote:
 >> Has anybody checked the signals on a 4046? Is the difference voltage
 >> between
 >> the two capacitor pins a true triangle or not ?
 >> 
 >> JH.

 B> Its not!
 B> It seems that you have missed BJ's and Magnus Daniellsons discussions
 B> on the 4046 pll!

Hey, I was about to say that... but you got there before me ;)

 B> Sorry to say, but the signals on each side of the cap is out of
 B> phase and only a half period sawtoot.Therefore you can not get 
 B> a true tri wave just by subtracting the waves.

 B> What you get in a subtraction is a half the wave tri and half the wave
 B> square.

 B> Looks like this:(well how about your gif proposal again?)

  
 B>    /l      /l
 B>   / l     / l
 B>  /  l    /  l
 B> /   l   /   l 
 B>     l   l
 B>     \   l
 B>      \  l
 B>       \ l
 B>        \l

This is correct.

The clue to how this waveform really comes together as it does is
hidden in a RCA application note ICAN-6101 (it is still a very valid
application note on how the 4046 works). On each end of the capacitor
there is one pair of N and P type MOSFETs and a inverter input to the
VCO state machine. When the loading one end of the cap is "hooked" to
earth via the N-channel MOSFET of that end and the other is beeing
feed with a load current via the P-channel MOSFET over in that
end. This load current comes from a P-channel MOSFET-pair acting as a
current mirror. When the voltage reaches the threshold (set by the
CMOS midpoint at 1/2 the Vdd-Vss voltage) the state-machine flips, it
now hooks the previous load point (now having the high voltage end of
the cap) to earth and the previous ground point (now having a low
voltage) to the load current. This will cause the previous ground
point to actually be negative, this is not a place you would want a
input to a CMOS logic and therefore there is a input protection diode
up to earth (there are more of those, but they are not used actively
in the curcuit) so now will the capacitor quickly uncharge via the
protection diode. When the capacitor is empty the loading current will
be in full charge and start loading that end.

In BJ's and mine discussion thread before I forgot the protection
diode discharing the cap. However, the curcuit would work if the
capacitor was not discharged to ground by the input diode, but the
input stage would lock up in the parasitic SCR mode and burn the hell
out of the 4046.

One could take inspiration from the 4046 VCO to make one that does
have a triangular wave. Note that the 9046 has a much improved
stability for supply voltage as well as many other benefits. They use
op-amps instead of 1-fet source followers (the pin 9 VCO input looks
into the gate of a N-channel MOSFET acting in emitter follower action) etc.

The signals are 180 degree out of phase in respect to each other.

 B> Difficult to do a asci of the wave!!

It certainly is, but equations mostly come out pretty readable.
On the other hand, I prefer TeX/LaTeX for equations.

 B> It looks like a "monster chain sawtooth wave" sort of.

Ugh. Yeah.

Cheers,
Magnus