High voltage power supply disables my shift registers

[Mixmaster]

Hello everyone. Long time browser of the forum, but as I stubled upon a problem with my nixie tube clock I decided to make my first post.

So inspired by Daves counter I decided to make a clock. Inspired by a 555 baser 170V power supply I tested this on my breadboard with a nixie tube and got light in all segments. Thats step one completed. Further I went with Daves first solution with ULN2003, 47V zener and 74HC595 shift register as I had a bunch laying around. This was also breadboarded and made some LED light up with the exact solution minus the zener diode. So I got the arduino working and power supply. Next part was laying out the PCB and ordering.

I got PCB made and I started by soldering components for 170V power supply to make sure this part was working. As the breadboard confirmed, this was working great. Soldering the rest of the components and all good. I connected the USB to the Arduino Nano and programmed it. While having the USB connected I measured the outputs of the shift registers and got the same pattern (high and lows) on the output as programmed. Note that this was without the 12V input connected, only 5V was supplied to the board by USB through the arduino.

Problem started when I applied 12V (12V supplied from by bench power supply) to the input which was supplying the 5V regulator and the 170V circuit. I had a nixie-tube connected with two digits only for testing purposes, and the two digits were just lit. I had programmed the outputs so only one digit should be lit. Measuring the outputs revealed 0V on all shift register outputs. This is strange to me as the power-LED on the arduino is lit and a LED is blinking as an "alive"-LED. The 5V is apparently stable per. DMM-measurements (Do not have a scope. Shame on me), and the arduino is apparently not resetting.

I tought this might an issue with the power input, so I soldered out the MOSFET. Now the 170V-circuit was not working and the output of the shiftregisters was the same pattern as programmed. This was with no USB-power, only 12V input. Soldering the MOSFET back and turning on gave me 170V again, but all shift register outputs low.

I attempted to solder a wire from +5V to a input of a ULN2003 to see if it the circuit actually worked, and it did. Connecting a input directly to +5V made a segment on the nixie to light up. By having one input high the other digit went dark and the digit that was supposed to be lit was stronger than when both were lit as described above.

So now I am clueless, if someone could help me on the way I would be very thankful!

Schematic in pdf attached. You might notice some faults in the nixie connections, but those are just screw terminals, not nixies soldered directly to the board. I also spent some time struggling with the hiddens pins on ICs in KiCad, but figured out that all Vcc-pins are connected to Vcc-label. No need to make the connections. So to clarify; 74HC595 is supplied by 5V and the 555 supplied by 12V. All these voltages has been measured and confirmed on my PCB.

[alm]

I do not see any power supply decoupling near the '595s. Obviously I do not know the PCB layout, so you may have a four layer board with power and ground planes and power supply decoupling very close, but especially since you were not able to check the Vcc of the '595 with a scope, this is one place I would suspect.

Are you sure the '595's output are not tri-stated (/OE high)? A Tri-stated output could measure high with a very high-impedance load. With the 2003 disconnected, try putting a 5-10 kOhm resistor from '595 output to ground, and see if the output still measures high if you try to pull ~1 mA from it.

The last option I can think of is that the ULN2003 is sinking too much current, but I find it hard to believe it would pull the output all the way down to 0 V. You could add a small series resistor (something like 10-100 Ohm) between '595 output and 2003 so you can measure the current flowing into the 2003.

[Mixmaster]

Thank you for the tips. Simple two layer boardNo decoupling cap on the 595, so connected a cap between Vcc and ground, no success. Also disconnected the 2003, still the same issue.

The OE is tied to ground. When it is supplied by USB, and not the 12V from bench PSU, the 595 outputs high. This means the 170V supply does not output anything. So it seems that the issue is with the 170V power. The same issue occurs when the +5V regulator is soldered off and the +5V is supplied from USB only and bench PSU only supplies the 170V.

So to me the issue is clearly with the 170V supply. I guess I should get hold of an oscilloscope and check the +5V rail.

[Andy Watson]

Decoupling, or lack of, would have been my first guess. Are you clocking the data into the shift registers on the correct edge of the clock?

[T3sl4co1l]

Any number of things.

Ground plane?

Without, your digital signals are all floating in space, easily disrupted by any nearby source of interference.  Including the signals themselves!

This goes double for supply bypass, and triple for that nasty little source of noise, the 170V supply.  Especially one that's nothing but a hack, and hasn't had any kind of EMC testing.

Tim