Tuesday, December 31, 2013

A simple inverter for driving neons/nixies using Radio Shack parts

I was visiting a friend of mine a few weeks to who was helping me repair the Model H horn on my old Atwater Kent 20C receiver.  This radio and horn lives in my office at work and one day, I heard a thud as the horn portion spontaneously broke off from the base and landed on the carpeted floor.  Inspection showed that the original cast aluminum part had a lot of voids and impurities and it finally broke - after nearly 90 years!  To fix it, we drilled out the aluminum and I fitted a steel sleeve that he'd machined and used metal epoxy to secure it - a very strong and (pretty much) invisible fix!

But I digress...

After the repair we wandered into his ham shack to talk for a while.  He is a collector of old, unbuilt kits and has a soft spot for the Radio Shack "P-Box" kits from the late 60's and early 70's.  For a list of those kits - along with much of the documentation, look here:


In particular he was interested in replicating the "Goofy Light" kit - see item 28-130 on the above link.

This kit is pretty simple:  A one-transistor oscillator along with a transformer produces 100-ish volts that power a series of neon lights.  Depending on how they are wired, they will produce a "chase" sequence or just blink randomly.

To replicate this identically would have required getting two parts that would likely be difficult to find:  A 2SB54 PNP germanium transistor and 1k-200k audio transformer.  Being practical, there was really no need to use a PNP germanium transistor in this:  A 2N3904 or similar NPN silicon would be just fine - but the audio transformer was another matter!

The actual impedance of the transformer wasn't as important as the turns ratio - and for the 1k : 200k transformer that would be the square root of the impedance ratios, as in:

sqrt(200/1) - 14.142 : 1 turns ratio.

Being that the turns ratio is one of the factors that determines the voltage transformation, it was fairly important that we find something that was sort of close.

If the waveform had been sinusoidal, a (theoretical) 6 volt input would produce about 85 volts of AC on the output.  Fortunately (!) for us, it's not quite that simple.  While the transistor/circuit losses would likely reduce the drive level to 2-3 volts or so, the waveform was likely to be anything but sinusoidal - more likely, it would be rather "spikey" as the transistor snapped on - then off again and it would be this "ugly" waveform that would likely have spikes and ringing that would produce voltages far in excess of that determined by just the turns ratio!

In searching the online Mouser-Key catalogs we found a number of possible candidates for substitution, including the Mouser 42TM-114RC which was a 20 ohm to 4.6k (15.2 : 1 ratio)  transformer with center taps on both the primary and secondary and it was readily available for just $2.74.  While this transformer would probably work just fine (the absolute impedance isn't terribly important in this application) he was interested in what might be on-hand locally.

At some point he'd been to Radio Shack and picked up a couple of their 273-1380 8 ohm to 1k (center-tapped) audio transformers and we decided to see if we could make that work.

The obvious difference - aside from the absolute impedance values and the lower turns ratio (this transformer had an 11.1 : 1 ratio) was that it did not have a tapped "primary" - and it was this tap that had provided the feedback path in the original P-Box circuit.  It did have a tapped secondary and I wondered if I could make that work so I threw together the following circuit using flying leads on the workbench.

A word of warning:
  • The voltages that can be generated by this circuit could potentially be lethal, so be very careful.  (You are unlikely to get more than a "tickle" or a slight bite, but be aware!)
  • It is possible that you'll blow up a transistor or two if you experiment with parts values and output loading.  You may also ruin a transformer, so pay attention to anything that is getting hot!

For this first version, I'd omitted Cfilt and Ca, using a 10k resistor for Ra and a 2N3904 for Q1.

Figure 1.
A simple circuit using a transformer and transistor to produce fairly high voltages.
The noted colors are for the Radio Shack 273-1380 transformer and indicate phasing:  If the phasing is
incorrect, the circuit may not oscillate!
(V+ is connected at the junction of the Red and Blue wires.)

Amazingly, it worked the first time, lighting the NE-2 type lamp attached to it - this, with a power supply of just 7 volts.  While the circuit worked, the current consumption was a bit higher than we would have liked - around 50-60mA at 7 volts - not terribly efficient, but then again, it was not at all bad for such a simple circuit on the first try!

Changing R1 to 47k, the circuit still worked fine and pulled only about 15-20 mA and the output voltage was a bit lower - still able to (just barely) light the neon at 7 volts, but brightly illuminating it at 12 volts.

After I got home from his place I decided to experiment with the circuit a bit more and in so-doing I put the 'scope on the output lead using the version without any capacitors and Ra = 10k and with the power supply at 6 volts, getting the waveform below:

Figure 2:
An example of the waveform being output from the transformer.

As you can see, the waveform is very "spiky" (to be expected) with the peak part being mostly at a negative voltage (e.g. below zero, the dashed line in the middle.)  With just a 6 volt DC supply the total waveform was about 226 volts peak-to-peak (more or less) with the oscillation frequency being about 2 kHz and the current consumption being about 50mA.

For the heck of it I decided to reconfigure the circuit as follows:

Figure 3:
 A reworked version of the same circuit as in Figure 1.  It produces a waveform that is essentially
an upside-down version of that in Figure 2, but at lower voltage since we have but half of the
secondary winding to produce an output voltage.
(V+ is connected at the Red/Green wire connection)

The result of this was a circuit that produced a lower peak-to-peak voltage than that in Figure 2 because only half of the secondary was referenced to ground via the center-tap and the power supply.  The waveform also looked similar to the one in Figure 2,  but upside-down - that is, the spike was positive-going.  It also drew a bit less current - around 35 mA.  (Cfilt and Ca were omitted and Ra was still 10k).

More experimentation with the circuit in Figure 1:

Because I was interested in the higher voltage I rewired the circuit back to the configuration in Figure 1 and did more testing, this time adding the following circuit to the output:

Figure 4:

This simple circuit converter that will handily convert the negative-going  portion of the waveform to positive and then add it to the positive-going part, which meant that with the circuit in Figure 1 operating from 6 volts you could get an unloaded voltage of over 200 volts DC.

For this testing I added Ca, using a 0.1uF capacitor and when I did this the quiescent current of the circuit dropped from 50-60mA to around 15-20mA when operating from 6 volts.  The output waveform looked about the same as that in Figure 2 but the oscillation frequency was now closer to 1 kHz.  When using the circuit in Figure 4 the voltage was slightly lower - but this was probably due to the high voltage spike occurring less often and keeping the capacitor (C2) charged in spite of the 10 Megohm load of the voltmeter.

For this circuit, C1 and C2 should be at least 0.1uF and rated for 250 volts or more while D1 and D2 should be high voltage diodes such as the 1N4004 or 1N4007 - or, better yet, a high-speed, high-voltage switching diode such as the RGP15G.

The actual values of C1/C2 depend on your load and how much ripple you can tolerate.  At 1-2 kHz, C1/C2 = 0.1uF will produce a fairly clean supply if you are only pulling a few hundred microamps.  For testing I used 0.22 uF for C1 and 0.47 uF for C2.  If you want a "cleaner" supply (i.e. less ripple) than the value of C2 can be increased further.

I decided to do a bit more testing with this circuit at different loads and supply voltages and came up with the following results, measuring the voltage across C2.  (Ra = 10k, Ca = 0.1uF):

6 Volts:
14k load - 52 volts output (3.7mA, approx. 190 mW)
100k load - 102 Volts output (1.02 mA, approx. 104 mW)
10 Meg load - 220 Volts output (22uA, approx. 5 mW)

10 Volts:
14k load - 63 Volts output (4.5mA, approx. 283 mW)
100k load - 155 Volts output (1.55 mA, approx. 240 mW)
10 Meg load - 325 Volts output (32.5 uA, approx. 10.6 mW)

15 Volts:
14 k load - 72 Volts output (5.1 mA, approx. 370 mW)
100k load - 234 Volts output (2.34 mA, approx. 548 mW)
10 Meg load - 460 Volts output (46 uA, approx. 21 mW)

  • The values in parentheses indicate the current flowing through the resistor being tested and the total power being dissipated by it.
  • A 14k load was chosen since this was the first resistor that I'd grabbed while the 10 Meg load was that of the meter that I was using to measure the voltage.

Using a high-current transistor:

For the heck of it I changed Q1 from a 2N3904 - a rather generic transistor - to a KSD5041, a specialized, high-current transistor designed specifically for photoflash use:  As compared to the 2N3904's 600 mA capability, the KSD5041 can handle about 5 amps!

As expected, the circuit drew more current when unloaded - around 70 mA or so, but I got much more voltage on the output when operating it from 6 volts:

14k load - 65 volts output (4.6 mA, approx. 297 mW)
100k load - 140 volts output (1.4 mA, approx. 196 mW)
10 Meg load - 500 Volts output (50 uA, approx. 25 mW)

In putting the 'scope on the output the waveform looked much like that in Figure 2, but the spike was much "sharper" and taller - no doubt due to the transistor turning on more firmly and allowing a higher magnetic flux to build up in the transformer.  In looking at the voltage across the collector of Q1, I noted that the waveform peaked up to about 55 volts - somewhat above the 40 volt rating of the KSD5041 transistor!

When I raised the power supply voltage to 10 volts I measured over 650 volts on the output before it started to sag and was accompanied by a dramatic surge in power supply current.  While the circuit still worked, the current was still high when I returned the voltage back to 6 volts and upon pulling the transistor and testing it, its current gain was very low, indicating that it had been damaged - unsurprising since I had already been seeing 55 volts on its collector when I'd been running at just 6 volts!

What's Cfilt for?

You'll notice "Cfilt" on the schematic diagrams.  If you are operating this circuit from a battery with very short leads, you can probably leave this capacitor off since the battery itself - and the short wires - will have a fairly low impedance, an important property for this circuit to function well.

If you are running this from a power supply - particularly one that is shared with other circuits - Cfilt should be used.  A suggested value for this capacitor is 47-220 uF and "low ESR" capacitors are recommended for this.  A word of warning, however:  Even with a good quality filter capacitor this circuit is likely to put noise on the power supply!

Additional comments:

You may substitute a PNP transistor (such as a 2N3906) for Q1 if  you reverse the power supply voltage,  If you do so you will also get a voltage waveform that is an upside-down version of the one in Figure 2.

If you need very low current at a higher voltage you can use just a simple, series diode and filter capacitor, taking advantage of the high voltage "spike" that is produced.

What is this circuit good for?

This circuit can be used for several things:
  • High voltage supply for Neon indicators.
  • It can also be used as the high voltage source for Nixie tubes - provided that both the power dissipation of the transformer/transistor and voltage under load are taken into consideration.
  • Powering of electroluminescent strips including the so-called "EL Wire" - although the cheap inverters sold for that purpose will likely work better!
  • The generation of a plate supply for low-power vacuum tube projects
If you want even more voltage there are several options you can increase the voltage by adding more capacitor/diode stages.  For more information on high voltage multipliers, look at the following web page:

http://en.wikipedia.org/wiki/Voltage_multiplier  (link)

In theory, it should be possible to get thousands of volts from this circuit but remember that as you go up in voltage, the amount of current that you can pull will go down!

Final words:

This circuit is not particularly efficient but with a bit more work and complexity, it could probably be made to be a bit better.  Its biggest advantage is that it uses parts that you are likely to find at your local Radio Shack and in your junk parts pile!


This page stolen from ka7oei.blogspot.com