2021 Jun 29, 12:09 PM
No problem for the "NOOB questions". For me it is quit difficult to handle the issue without having the stuff on my workbench. It could be anything and we have to handle it by the well known processes "trial and error" and/or "elimination".
In general, if you want to stabalize power you need a large and a small cap. The large cap handles ripple of lower frequencies the smaller cap is faster and will handle higher frequencies but with a limit. If the noise or spikes are to large than it is better to use a LC (coil/cpa) or RC (resitor/cap) circuit. Therefore the suggestion of the 100 Ohm resistor in series with a 0.1 uF cap behind the R (on the hc595) side. Several cases of unwanted behaviour of the hc595 can be found on the internet and are solved with this RC circuit.
Swapping the Pi psu was a good test to do. If it makes abolutely no difference than you can eliminate the psu (for now). If it did make a significant difference but didn't solve the problem than the psu could still be something to look at.
I was also wandering if all 4 boards are showing erratic behaviour or is the first board different? The data from the Pi is clocked in on the first hc595 (pin 14) and the overflow bits are exitting on pin 9 which is connected to the next hc595. If the noise/spikes are on the data, clock or latch pins than you could a test with connecting these pins to gnd. Of course the circuit does not work but I 'm curious if you pull the black and brown wire from the Pi and connect them to GND what happens if you do what you do (pull fan?) to generate the issue.
The above might not solve the problem but is merely to identify the entrypoint of where the problem enters your circuit.
In general, if you want to stabalize power you need a large and a small cap. The large cap handles ripple of lower frequencies the smaller cap is faster and will handle higher frequencies but with a limit. If the noise or spikes are to large than it is better to use a LC (coil/cpa) or RC (resitor/cap) circuit. Therefore the suggestion of the 100 Ohm resistor in series with a 0.1 uF cap behind the R (on the hc595) side. Several cases of unwanted behaviour of the hc595 can be found on the internet and are solved with this RC circuit.
Swapping the Pi psu was a good test to do. If it makes abolutely no difference than you can eliminate the psu (for now). If it did make a significant difference but didn't solve the problem than the psu could still be something to look at.
I was also wandering if all 4 boards are showing erratic behaviour or is the first board different? The data from the Pi is clocked in on the first hc595 (pin 14) and the overflow bits are exitting on pin 9 which is connected to the next hc595. If the noise/spikes are on the data, clock or latch pins than you could a test with connecting these pins to gnd. Of course the circuit does not work but I 'm curious if you pull the black and brown wire from the Pi and connect them to GND what happens if you do what you do (pull fan?) to generate the issue.
The above might not solve the problem but is merely to identify the entrypoint of where the problem enters your circuit.