I have built a noise generator using a couple of CD4006 shift registers. The circuit is on page 9 of this PDF:
http://schematicsforfree.com/archive/file/Oscillators and Generators/Pink-White & Other Noise Generators/Pink & White Noise Generators - Misc.pdf

I used a CD4070 quad XOR instead of the older CD4030. I haven't implemented the pink filter or MPSA13 buffer, I'm just monitoring the output directly on the o'scope).

It is on breadboard and all works perfectly until I start varying the supply voltage, between 5V and 15V. It appears that at certain critical supply voltages (e.g. 5.3V, 10.0V and others...), the sequence stops being random and somehow gets stuck in a repeating pulse train of a few kilohertz, even though the clock continues to run normally. In fact, at 6.8V the output stays permanently low, yet the clock is running! Can anyone shed any light on this??

 

the timing changes with supply voltage, delay through the registers changes.

 

hi david,
A lot of the very old CD CMOS had unbuffered outputs, sometimes coded CD4070U. or UB
The newer CD CMOS is buffered on the output, coded CD4070B.

The buffering makes a big difference in performance of some circuits.

The PDF posted is circa 1979
E

 

The chips I'm using all appear to be the 'B' (buffered) type. I will try making use of the 4006's half-cycle delayed output latch and see if that helps.

 

The output latches didn't help at all -just stopped it working. However, it seems that a 100pF cap between the clock input and ground does the trick.