High Voltage Nixie Power Supply

A modern DC/DC converter design capable of delivering current in excess of 40mA at 170V

Similar projects worth following
Starting from
tonyp7 has 10 orders / 0 reviews
Ships from Singapore, SG
Building a high efficiency nixie power supply

My nixie clock power supply died. It was one of these cheaply built Chinese modules so it doesn’t come off too much as a surprise. Looking at how to replace it, I realize there is a dire need for a better high voltage design out there. Most of the kits seem to be based of a traditional boost converter based on a MC34063 switching controller. This has several issues:

  • The 34063 was never meant for high voltage
  • It’s a design from the mid 70s with everything that it implies -no thermal protection, not a current mode controller, crappy feedback sensitivity

It’s astonishing that there are so many nixie tube enthusiasts out there, yet there is no good solution for the most critical part of a nixie circuit: the high voltage supply. With that in mind, I decided to design my own PSU with one requierement in mind: build it like a tank.

The LT3757 flyback design

key components:

  • The LT3757 switching regulator controller from Analog / Linear Tech.
  • The DA2032 3A, 1:10 flyback transformer from Coilcraft.

The LT3757 is a good choice because it can specifically be used for high voltage, as demonstrated in the datasheet with a boost example to 300V. The downside is that it’s extremely expensive. At the time of writing this, I paid US$4.49 for one ($2.55 for 100); which is a ridiculous amount of money for a 3x3mm MSOP-10 package. $4.49 buys you a fairly nice middle of the range ARM micro-controller.

The DA2032 is a coupled inductor with a 1:10 turn ratio (a “flyback transformer” is a glorified coupled inductor – see below for further reading). The primary coil is rated for 3A. Because we need such a high boost ratio, we definitely need a high turn ratio on the transformer. On the wallet damage side of things, count US$4.50 for one of these.

We already have $9 worth of components and we haven’t even started building this converter! Then again, the goal of this prototype is not to be cheap, but to build a very reliable power supply capable of lasting years and years without failing. The only way to achieve this goal is to use quality components well within their electrical characteristics. Overstressing parts is the main reason why a circuit may work, but only for a short amount of time.

I simulated the design in LTSpice and everything looked good; so there was only one thing left to do: building it!

PCB and initial tests

PCB was manufactured by JLCPCB -once again they did a good job. It’s my first time prototyping with very tiny 0603 components so I ordered a stencil with it which made my life a lot easier and the soldering look very professional. Unfortunately the MOSFET I picked is a D2PAK (TO-263) package and not the DPAK (TO-252) footprint I used on the PCB. Oops. Well it still works but the drain of the MOSFET isn’t completely soldered on.


Benchmarking is done using dummy resistors as load, in this case I used 6x43K resistors in parallel which at 170V is 21.7mA. Since a typical IN-14 Nixie Tube uses about 2.5mA of current, this benchmark works as if you're powering over 8 tubes (21.7 / 2.5 = 8.68).

At first I measured an efficiency of about 70% at 12V. That's not horrible but it's still a substantial power loss. The power loss is mostly due to the crappy soldering of the MOSFET and the snubber behind the primary coil of the transformer.

The snubber is here to reduce voltage peaks seen by the MOSFET, which could typically end up damaging permanently the power supply. Because of this, Linear can get away with a 60V rated MOSFET switching the primary coil of their 300V boost example; at the expense of lost energy.

To improve the efficiency I removed the snubber and used a proper DPAK MOSFET, 100V rated.

I used a IRLR3110ZTRPBF from Infineon because it offers pretty good switching characteristics:

  • 14mΩ Rdson
  • 34nC total gate charge
  • Solid avalanche rating --that is very important as the switching MOSFET will see peak voltage above its max...
Read more »

  • 1 × LT3757 Power Management ICs / Switching Regulators and Controllers
  • 1 × DA2032 Inductors, Chokes, Coils and Magnetics / Current, Voltage and Power Transformers

  • Efficiency Measurements

    Tony12/30/2018 at 02:58 0 comments

    The test rig

    I measured the efficiency of the power supply using a rig made out of a female header and 8 30k resistors. That way I can measure loads at roughly 5mA intervals (172V / 30K = 5.7mA per resistor).

    It's not pretty but it sure does the job!

    In addition to that, I measured the efficiency using 3 different input voltage: 5, 9 and 12V.

    The results are in!

    The results at 5V are pretty impressive. Having over 80% efficiency on this ~34 boost ratio is an amazing result! Past 20mA it quickly drops though, showing the limits of the design at such low voltage. That being said, this makes a nixie clock driven by a single 5V wall wart completely possible! Very exciting!

    9 and 12V are showing similar curves, and the power supply can output well over 40mA@172V using these. On very low current (~5mA) the 12V input shows a lackluster 71% efficiency but it rapidly increases to 85%+ efficiency as hte load increases.

  • Professionally Assembled boards are here!

    Tony12/29/2018 at 01:39 0 comments

    What you are seeing above is the first commercially produced nixie power supply; completely factory assembled!

    And my, it is a thing of beauty!

    Changes compared to the last prototype (Rev.B) are really kept to a minimum, but include:

    • Minor GND trace change that fixes a high current loop with the transformer.
    • Bypass of the gate driver with Vin.
    • ENIG (immersion gold) surface finish.

    PCB Assembly?

    Dealing with the PCB assembly process was mostly uneventful but a very good experience nonetheless. In addition to gerber files, you have to provide a complete BOM and a "centroid file": a CSV containing the location of the center for all components. Something which I had never done but turned out ok!

    During the whole process, the fab (PCBWay) was very professional and asked me to confirm both the orientation of the diode and the transformer. They didn't ask for the MSOP-10 component. Maybe my silk screen was not clear enough for these two?

    Since this is a fully automated PCB Assembly, the manufacturer had to pick ENIG (immersion gold) finish which is pretty much mandatory for 0.5mm pitch component in addition to looking very pretty.

    Another interesting change I had to make is to replace the pull-up resistor on the shutdown pin from 100k to 140k. The challenge of PCBAssembly is that the more components you have, the more expensive the assembly is. 140k is a value needed to set the frequency of the switching regulator, so by re-using the exact same component as a pull-up; I managed to reduce the BOM count by one line.

    In total, 10 boards fully assembled costed me US$286 including shipping. At $28.6 a piece, it is actually not outlandish considering the low volume and quality of the components used.

    What now?

    In the coming days I am going to sell these boards on my Tindie store after testing each of them individually and include proper efficiency testing for each. I'll make sure to report final efficiency numbers here on hackaday.

    If enough people are interested, I will then probably go for a bigger "production" run (~200 pieces) which should bring the costs down a lot.

  • Prototypes and revision

    Tony11/29/2018 at 01:02 0 comments

    The LM3488 experiment

    Since building the original prototype I've tried a few different things, for instance I tried a prototype using the LM3488 from Texas Instruments; a more or less direct competition of Linear's LT3757 (a bit simpler and cost roughly half!)

    I've sent a 10x10cm PCB with various configurations to see how it would work nicely:

    On the top right you can see here a flyback version of the PSU with a LM3488.

    On the bottom right above, this is a classic boost topology with a beefy power mosfet and power inductor. The bottom right DPAK package is not a transistor, but the switching diode!

    These are true prototypes where I even included some jumpers to select the switching frequency to check what kind of effect it would have on the overall power supply.

    Revision B

    In the end, I decided to stay with the LT3757 as it is the chip that gives me the most efficiency. It's probably linked with the fact that you can bypass the MOSFET gate drive regulator with Vin on Linear's chip, but that is pure speculation on my part.

    So I have sent a new revision of the board with Linear's chip:

    Please don't pay attention to the botched up job on the diode at the top, it was just an experiment I was running to see what is the impact of the dual diode vs a simple beefy SOD128 that is supposed to go on this footprint (a RFN1LAM6STR 25ns fast recovery diode)

    Compared the first revision I have added more comprehensive silk screen; a shutdown pin to turn off the regulator completely; and reorganized the layout so that the high voltage stays on the same side of the board.

    This is now very close to being a production unit!

View all 3 project logs

Enjoy this project?



Dan Julio wrote 11/29/2018 at 02:46 point

Nice job.  I think a few extra dollars spent on a reliable power supply would be a good investment if you're powering a few hundred dollar's worth of new Dalibor Farny nixie tubes!

  Are you sure? yes | no

Tony wrote 11/29/2018 at 06:23 point


  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates