I would like to drive 6 Nixie-Tubes from a standard USB-Port (500mA / 2.5W), which is somehow hard to do because \$ 190 \text V \cdot 3 \text{mA} \cdot 6 = 3.5 \text W\$.

If I could reduce the source voltage to 120 V the power consumption would be about 2.2 W, which could probably be doable with a flyback-converter (~90% efficiency). I would then use short 190 V impulses to ignite the Nixies and use 120 V for most of the time.

This schematic shows the idea:


simulate this circuit – Schematic created using CircuitLab

V1 would be the flyback-converter and I1 would be some kind of µC controlled current sink. The Nixies would then be driven with about 5 times I1.

This looks about right in simulation, see figure below, but I'm not sure how well this works with real hardware. DC sweep of voltage vs. current

  • Is this generally a very stupid thing to do?
  • I would have to use one (non-existent) chip with 7 matched transistors with \$V_{CE} \gt 80 V\$. What kind of transistor could I use instead, so that the currents vary no more than, say, 10%?
  • Transistor choice is generally not trivial because of the relatively high \$V_{CE}\$ drop vs. small currents. Are there any other parameters to watch out for?
  • Temperature is obviously a problem (\$ 10 \text{V} \cdot 3 \text{mA} = 30 \text{mW} \$), so I would connect them to the same heat-sink
  • Is there anything else (i.e. voltage ripple from the flyback) to consider?
  • \$\begingroup\$ Why not just use current-limiting resistors? \$\endgroup\$ Aug 31, 2016 at 13:34
  • \$\begingroup\$ You mean a JFET current regulator or current-limiting diode? I haven't quite found any that match my requirements (the high voltage drop). Also I would like to change the brightness without using PWM. It's more like a challenge to see if it's doable, really ;) \$\endgroup\$
    – michi7x7
    Aug 31, 2016 at 13:39
  • \$\begingroup\$ There is no saving here- the current mirror acts as a linear regulator- to get 3mA at the nixie element you need 3mA @ 190V, plus 20% for the control current = 4.1W total. \$\endgroup\$ Aug 31, 2016 at 13:40
  • 1
    \$\begingroup\$ Plus you will have to manually choose the transistors so that they have matching beta, and tightly thermally couple them. \$\endgroup\$ Aug 31, 2016 at 14:03
  • \$\begingroup\$ @SpehroPefhany: Well, as stated, I would like to reduce the 190V to 120V (flyback converter) for most of the time. \$\endgroup\$
    – michi7x7
    Aug 31, 2016 at 14:18

1 Answer 1


Okay, so you want to dynamically adjust the 190V converter down to 120V after switching.

You don't really need closely matched transistors for this application because you have relatively (and appropriately) large emitter ballast resistors. You can use MPSA92 transistors which are cheap and very widely available.

For example, suppose one transistor has a Vbe of 0.6V and the other has a Vbe of 0.65V (because it's not as hot, say) then the difference will only be 0.05/3V which is less than 2%. Temperature variation will probably be more important than unit-to-unit variation but I doubt it's a problem for visual brightness. You could minimize it by putting the pairs together face-to-face but again I doubt it's worth it. There is also an SMT version available MMBT.. and mounting them close to each other doesn't hurt.

You can use MPSA42 NPN transistors to drive the current mirrors. Again, cheap and widely available. 30mW (check your numbers) is not very much dissipation at all. Below 200mW for TO-92 or about 120mW for SOT-23 you're likely fine in any reasonable environment.

The 'rating' of 625mW on the TO-92 transistors is a bit of an exaggeration but 1/3 of that is not unreasonable. Note that if the output somehow gets shorted the transistors will see 560mW with the 190V supply, which will make them very hot fairly quickly, but should not kill them instantly if TO-92.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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