1
\$\begingroup\$

I'm working on a Nixie clock based on a STM microcontroller. It consists of two boards: one includes the microcontroller and a boost converter and the other includes the tubes and two HV5530 high voltage shift registers.

I noticed that the clock works fine when one tube is on, but when I try to turn more tubes on, it behaves strangely and some tubes turn off. The code was trying to display '123456' and to update the registers every 100 ms. Here is a video of this problem.

https://www.youtube.com/watch?v=nN-ojr2xXTM

After some hours of investigation I found the cause: ringing in the boost converter which adds noise on the ground and on the traces that go from the microcontroller to the HV5530 chips. Here's a scope capture of that. Channel 1 is the switch node of the boost converter and Channel 2 is the clock signal probed right at the input of one of the HV5530 chips.

enter image description here

Here is the schematic: enter image description here

And here is the layout: enter image description here enter image description here

I added a snubber between the switching node and ground consisting of a 100 nF capacitor and a 1.2 K resistor. The improvement is only marginal and the resistor gets really hot.

And finally, my questions are:

  • Is there an obvious schematic or layout error that could cause the ringing?

  • Is it wrong to have a ground plane under the boost converter? I'm thinking about the fact that it adds stray capacitance to the traces.

  • How should I calculate the RC snubber?

\$\endgroup\$
  • \$\begingroup\$ Did you try a simulation? Which assumptions did you not test or verify? Ground noise could also be your probe ground inductance. How do you control, test, compute impedance? There is so much you aren't telling us about clock, power and regulator design specs. Test Power & clock loading regulation in isolation. (without EMI) with equivalent R impedance. Does it work? \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Nov 27 '19 at 17:55
2
\$\begingroup\$

I finally found the root cause: the HV5530 chips really need 12 V logic in order to work. Initially I gave them 5 V logic and that made the communication really susceptible to noise. I made a circuit to raise the logic level and the display looks perfect even with the snubber from the boost converter removed.

| improve this answer | |
\$\endgroup\$
0
\$\begingroup\$

I've been trying to understand what you think the root cause of the problem is so please bear with me if I'm not following right:

  • scope capture shows certainly pretty consequent ringing on Channel 2, which is... VCC scoped at the input of the boost regulator? (eg. scope shows 5V)
  • what is the rated voltage and current range of one Nixie tube? (eg. is the boost circuit sized accordingly?)
  • about the ringing, where did you scope and did you use a short ground lead (also connected where?) and high-speed probe? Sometimes it is an artifact of the probe ground loop picking up noise from the circuit
  • it looks to me like the boost regulator circuit is located in the south-east corner of your board, correct? Could you highlight/zoom-in on all parts of this circuit next time so it's easy to spot?
  • you mentioned the two decoupling capacitors on VCC but they are shown separated from the circuit in your schematic hence on first view it felt like you did not have them. I recommend to connect directly to the IC next time.

How did you determine ringing was the root cause of the issue? Have you eliminated options like:

  1. have you isolated each Nixie tube (turning them one by one with all possible values) to make sure they are all functioning fine? If yes, how many tubes does it take to re-create this behavior?
  2. boost circuit output sag > can you monitor the HT node while turning on all Nixie tubes? Have you try staggering the turn-on time for each tube? (eg. could be inrush current protection kickin-in)
  3. verifying your 5V source while monitoring MCU stability using printf or external LED control? (or any other way)
| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ - Channel 1 is the switch node of the boost converter and Channel 2 is the clock signal probed right at the input of one of the HV5530 chips. \$\endgroup\$ – Stelian Saracut Dec 3 '19 at 9:05
  • \$\begingroup\$ - One nixie tube draws a maximum of 7 mA at 170 V, but the resistors limit the current to about 5 mA. It needs at least 200 V to strike. The boost converter provides 230 V and was sized to handle the current. The actual voltage from the boost converter drops by about 6 V when all tubes are lit, but it is still more than enough for the tubes to strike. It's a good question, but still I'm sure it's a layout/grounding issue. \$\endgroup\$ – Stelian Saracut Dec 3 '19 at 9:15
  • \$\begingroup\$ - I used a regular 1 X probe and a regular ground lead. I scoped the switching node at the transistor drain and I connected the ground to the thick ground trace of the boost converter by removing the solder mask and soldering a thick copper wire on the trace. I attached the ground alligator clip to that wire. \$\endgroup\$ – Stelian Saracut Dec 3 '19 at 9:19
  • \$\begingroup\$ - The boost regulator circuit is located in the south-east corner of the board on the bottom side. So you can see it in the first layout screenshot with green traces. \$\endgroup\$ – Stelian Saracut Dec 3 '19 at 9:25
  • 1
    \$\begingroup\$ @StelianSaracut Concerning the Boost switcher circuit, dropping 6V may be nothing relative to 230V but it is significant for a switcher which such a low-current output. I would not rule the switcher design out yet. Can you show more waveforms: switching node vs HT output with different time scale (to see the PWM waveform from the switcher) and with all Nixie off then on? \$\endgroup\$ – eeintech Dec 3 '19 at 12:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.