By the way I have updated how the LM334 is biased on my PaintCAD.

​

 

Using the 1N4454's here's what I measured.

Input Voltage	Current
3.0 V	2.5µa	.............................................................................................................
3.5V	2.5µa
4.0V	2.6µa
4.5V	2.6µa
5.0V	2.6µa
5.5V	2.7µa
6.0V	2.7µa
12.0 V	2.9µa
29.0V	3.5µa

Much better, I find it interesting that the diodes leaked when I got it to the higher voltages. Looks like I am almost done here. Now to add the 555.

 

Next brain fart:

​

I'll build it tomorrow and let everyone knows how it worked.

 

There seems to be a relationship between forward voltage and leakage:
Germanium - low forward voltage drop - lots of leakage
Silicon schottky - slightly higher forward voltage drop - a bit less leakage
silicon - less leakage - higher forward voltage drop.
It seems to model as a parallel resistance, so more voltage = more leakage.

 

There is another way of generating a triangle wave:

It does rely on both LM334s being identical and R1 being exactly twice R2 for perfect symmetry.

 

Or you could do this:

Symmetry would still be perfect, the square wave would be perfect too.
I'm going for one resistor and one capacitor in the design.

 

Or you could do this:
View attachment 247193
Symmetry would still be perfect.

Unless the LM334 has reverse leakage problems!
(Datasheet says it will withstand 20V reverse bias, but doesn't say if any current might flow)

 

Then you would put a diode each leg.

 

Back to basics:

Nothing I do seems to get this circuit to work. So I am going to use step-by-step experiments to see if it works I think realizing that the diodes leak backwards was a major breakthrough for me. So next step I will try this experiment and I will post the word waveforms.

​

why 10KΩ? Because it should draw about 100 µa when plugged into the CC circuit. We shall see.

I left the other components in there because there's no sense in tearing this thing down step by step.

Pin 3 measurement:

pins 2 and 6.
Ignoring the diodes the calculated frequency of.7/RC is 700Hz, the displays show the measured frequency

 

So now we have established that the diodes are not killing the oscillator. Next step is to try this:

​

so after correcting my drawing, I got this:
3.0V=2.4µa
4.0V=2.4µa
5.0V=2.5µa
6.0V=2.5µa
Which is close enough for experimental purposes. Obviously my understanding of how the current programming resistor is flawed. So now I will try this circuit with these values back to the previous experiment.

 

Back to square one. It simply does not work. I get intermittent pulses, but nothing repeatable. Frustrating. I'm not giving up on this but it is time to move on to other projects.