I have some problem with my nixie project.

1 - my mc34063 boost converter have poor performance. While i use active pull down D3 and Q2 for driving my mosfet then its enough for only one nixie tube (z566m - ~2-3mA). But when i remove D3,Q2 and use 330ohm for R12 then efficient increase but for only 4 nixies. It should have at least 20mA (with D3,Q2) but now its less than 15mA efficient.

2 - when i set voltage above 180V then my avr very often hangs (it probably hang because noise on I2C line, it also hang more often when i connect my 4x20 I2C lcd using 20cm wires), and it also count very high interrupts form RTC - im not sure is it weak pullup or noise on supply line. Currently i do not have osciloskope so its very hard to guess. Also when voltage is above 180V then very often I'm unable to send new hex to my avr (failed connection, mishmash or something else). I already burn 2 avr during programming form unknown reason (i cant rescue them even with avrdoctor - HVPP)

3 - any objections to HV layout? i never project this kind projects so here may be some mistakes in routing.

Thans for any help

enter image description here enter image description here enter image description here

enter image description here

  • 2
    \$\begingroup\$ No ground plane, switch mode converter and micro controller? I would give your chances of any signal integrity as slim to none. \$\endgroup\$ – winny Mar 18 '18 at 14:04

Here are some things you could do:

  1. You arranged the components of your boost converter in a neat row. That's a very bad thing to do in a switching converter. You should instead try to put the components as close together as possible to minimize the length of the traces on your board. You could actually lay out the components just like in your schematic, with a wire bridge for MC_FB.

  2. Add a ceramic and electrolytic capacitor to the supply of the boost converter. Its pins should be connected via short traces right to the +12V side of R11 and the source of Q1. This should help to minimize the loop area of the path that the high-frequency switched current takes.

  3. Put a small resistor in series with D3, maybe 4.7 Ohms. That'll reduce the switching noise a bit but increase the power dissipation in Q1.

  4. Put the boost converter on a separate board.

  5. Use a star ground.

  6. Don't be afraid to use wire jumpers if it helps to keep traces short. If you think jumpers look bad, use 0 Ohm resistors.

  7. Put ferrite beads into the supply lines for the boost converter.

  8. Decrease R12 considerably (100 Ohms 2W should do), the turn-off time of Q1 is likely quite bad with a R12 of 1k. Replace Q2 with a BC327 (that transistor can drive higher currents). Maybe also replace D3 with an UF4007 (not 1N4007).

The most important thing is that you keep the high-current hight-frequency path in the boost converter as short as possible. The loop "input cap positive - current sense resistor - inductor - switching transistor - input cap negative" should have a really small area on your board.

  • \$\begingroup\$ 1 - So it is better to route MC_FB across board instead HV line? In my opinion MC_FB is more sensitive to noise because of its impedance, or Im wrong? 3 - this resistor should be before or after PNP transistor? 8 - what is adventage of UF4007? When I compare them then trr for 1n914 is 10times better \$\endgroup\$ – Dobijasek Mar 18 '18 at 17:03
  • \$\begingroup\$ @Dobijasek 1: You shouldn't route traces like those around "obstacles" anyway (that'll make them too long) but use a wire jumper instead. 3: Directly in series with the diode, leaving the emitter of Q2 attached to the FET's gate. 8: You're operating the 1N914 above its maximum rated pulse current, possibly damaging the diode. The UF4007 has a higher current rating. The switching frequency of your circuit isn't high enough for that difference in trr to matter. (A regular 1N4007 is too slow though) \$\endgroup\$ – Jonathan S. Mar 18 '18 at 17:16
  • \$\begingroup\$ 1 - I mean something else. Is it better to put voltage divider near Cout and then trace 1,25V signal or its better to trace HV signal from Cout to the voltage divider that is closer to IC? (it is more general question for future knowledge). Any modyfication I co do at least tommorow because I do not have any tools at home. \$\endgroup\$ – Dobijasek Mar 18 '18 at 17:26
  • \$\begingroup\$ @Dobijasek Better place it close to the IC because that's where the reference voltage is generated. You could change your layout and show it here before you modify the board so I can look over it. \$\endgroup\$ – Jonathan S. Mar 18 '18 at 17:42

I feel bad for saying this considering the amount of work that went into this board, but the layout of the DC-DC is terrible.

As Jonathan says you should be able to salvage this by putting the HV switching converter on a separate board. I'm going to explain how to build it.

First look at the pictures in this answer.

Get a bit of blank copperclad which will be your ground plane. It should be roughly the same size as your current DC-DC converter so you can put it in place later, it looks like 3x6cm or something like that.

Then take a bit of perfboard, say 2x5 cm. Put this on top of the copperclad, upside down so you see the copper pads. Use 4 bits of resistor tails in the corners to solder it to the ground plane and keep it in place.

Now you can assemble your circuit on the perfboard. The component leads won't be able to go through because there is grounded copperclad on the other side, but that's not a problem. Make the important connections as short as possible. That's mostly the current path through the input cap, inductor, FET, diode, and output cap. Optimum layout has the input cap, output cap, and FET ground pins connecting to ground at the same point.

The important thing here is that the copperclad gives you a very low inductance ground plane and you can solder the ground pins of all components on it directly. It will also act as a shield and reduce EMI radiation.

Now about the component choices... let's run the equations for your boost converter.

Vin=12V / Vout=180V / Iout=20mA

Duty cycle D = 0.93 (which is way too high for a boost, a transformer-based converter would be better here but let's go on...)

Frequency is hard to read on MC34063 datasheet graph but let's say F=50kHz.

T = 1/F = 20µs

Toff = (1-D)T = 1.33µs

Ton = DT = 18.66 µs

(Toff is a problem as I'm not sure the wimpy gate driver will be able to do a full on/off cycle in 1.3µs but let's go on...)

Inductor ripple current = Ton * Vin/L = 0.68A YIKES!!!!

At 180V/20mA out, you should have around 300mA average inductor current but since the ripple current is more than 2x that, something will have to give somewhere...

Now let's check the caps:

Ripple current rating of 2.2µF/400V capacitor on the output = 34 mA !... This cap is useless anyway, as it has an ESR of 40 ohms (random datasheet from Mouser). So this cap will shortly die, and it won't smooth the output voltage.

Problem with the high ESR of this output cap is that it will result in huge voltage spikes on the output when the FET switches. Plus the layout inductance which doesn't help. This could be what sends your microcontroller in a coma. With 40 ohms ESR when the FET turns off and the inductor dumps its current into the output cap you're looking at a 20-30 volts spike with a rise time equal to the FET turn-off time, which is rather fast...

Fix: use a low-ESR cap with a proper ripple current rating on the output. This is going to be a bit hard to find, unless you use ceramic or film, but at 400V these will be huge and therefore inductive. (This is also why a transformer-based converter would work a lot better, output current ripple is much lower).

The input cap (on 12V) also needs to be able to provide 0.68A ripple at 50kHz so it needs to be a beefy low-ESR one. Check the datasheet. Add ceramic caps in parallel.

I checked this dude who sells a kit with the exact same schematic, and it doesn't seem to bother him at all that the output cap runs at more than 20x its rated maximum ripple current. I wonder how long it will last...

In light of this I would rather advise you get a transformer-based converter, maybe this one, I never had it in hand so I can't vouch for it though. It's available on ebay for about $10.

  • \$\begingroup\$ my output capacitor is: NICHICON UPS2G2R2MPD 2,2u 400V. So does it metter that caps voltage rating is much higher than I need? Any suggestion what should be output capacity? \$\endgroup\$ – Dobijasek Mar 18 '18 at 18:32
  • \$\begingroup\$ It's rated for a couple tens mA ripple current... I'd use ceramic or film but better use the transformer based converter... \$\endgroup\$ – peufeu Mar 18 '18 at 19:13
  • \$\begingroup\$ Then should I replace my Cout or just add it in parallel? \$\endgroup\$ – Dobijasek Mar 19 '18 at 18:14
  • \$\begingroup\$ You can replace it... if you have some film caps in stock with proper voltage, you can use that, like 2.2µF or two 1µF in parallel. I'm not sure it'll work, the layout problems still remain \$\endgroup\$ – peufeu Mar 19 '18 at 18:20
  • \$\begingroup\$ Are there any difference between AC and DC capacitors? Currently I have only few MKT 275VAC 0,47uF but Ill buy them layter. \$\endgroup\$ – Dobijasek Mar 19 '18 at 18:24

I tried to change diode D3, transistor Q2, add resistor with after D3, replace inductor, Cosc or output capacitor with almost any combination, sometimes it help but only little. In the end i just put potentiometer directly on mosfet Gate (currently its resistance is ~230ohm) and my converter revive, even my coil and mosfet are less heating, I just gues it was ringing but without osciloscope it is almost impossible to confirm. Now my avr does not hang, I have no problem during programming and my interrupt count properly even if I touch everytching fith my finger. However Im now not able to go higher than 195V. I also recalculate resistors for my nixies and colons to operate around 170-175V and now its probably fine.

In this case I do not have other choice because I have to finish my nixie clock in few days because its gift for someone ;). Thanks for any answer and suggestion. enter image description here


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.