Page 1 of 3

Neon sign transformer power: a new look

Posted: Fri Aug 24, 2018 10:34 pm
by Rich Feldman
This is to re-open the topic of power output from neon sign transformers.

The most detail I've seen previously is Richard's old paper, neons.doc, linked in a HV FAQ thread called "#8 FAQ - Neon transformers - the facts". viewtopic.php?f=29&t=10333
One rule of thumb, from there, is that one might get 2/3 of nameplate OC voltage and 2/3 of nameplate SC current at the same time, into a resistive load chosen for maximum power. A more conservative expectation, consistent with a linear model, is 1/2 voltage and 1/2 current. In the case of a nominal 15 kV, 60 mA transformer, the maximum sustained power would be 400 W or 225 W.

As a first step toward data collection in my own lab, I just made a little battery-powered AC kilovoltmeter.

It's based on a voltage divider board from the flea market, whose resistors are mounted on PTFE-insulated eyelet terminals. Total resistance is about 90 megohms. Original design voltage is unknown. If we run it up to 18 kV, the meter current would be 200 microamperes and power dissipation would be 0.36 W per resistor. Max voltage per resistor would be 1800 volts, which I bet is OK even though the datasheet is missing.
Things I would do differently next time: Power switch less susceptible to accidental pushing ON, mostly soldered connections instead of clunky screw terminal strip, and a battery test button.

Of course for AC input, it indicates the average rectified voltage. This cheap DVM reacted to 120 Hz ripple voltage by aliasing, instead of rejecting it, so I added a filter capacitor. Rectifier needs to handle the divided voltage, not the full voltage. Forward drop of two diodes is an error term subtracting from full voltage. Maximum reverse leakage current in rectifier datasheet is 10 uA at 25 °C, which could cause substantial errors here, but is a super-conservative spec. I measured leakage on several specimens with 24 volts DC applied to the AC terminals. Chose the best: less than 40 nA in both directions (4 mV across a 100k resistor).

The first step for calibration was to measure the DVM response with respect to a 6-digit Agilent benchtop voltmeter. Between 0 and 15 V, the little meter reads about 0.6% low. Then I decided to calibrate for direct indication of RMS AC kilovoltage. A low voltage NST was dialed down to 950 V, as indicated by 1000 VAC range a handheld Fluke meter. Then the trimpot in picture was adjusted to make the divided voltage 956 mV (indicated as 0.95, occasionally 0.94). The little NST on Variac goes up to 4.4 kV AC, as indicated by new meter, but I have no independent measure of that.

Today I tried a DC test with an industrial hipot tester, found "in calibration" and preset to 2.13 kV. New meter in picture indicated 2.44, apparently 14% or 15% higher than the true DC voltage. It ought to read 11.1% high, the ratio of sine wave RMS to rectified average voltage. Good enough for now!

Re: Neon sign transformer power: a new look

Posted: Mon Aug 27, 2018 5:56 pm
by Rich Feldman
Just found another 15-year-old thread about NST characteristics, on this very forum. It contains lots of information and some arguing. viewtopic.php?t=4315 It even includes John DeArmond, "Neon John", whom I remember from pre-Internet Usenet days (1993-1994).

I've still not yet seen (except in my own data from last weekend) a printed I-V curve for a NST output, showing the alleged convexity to the right of simple linear model. Such as one passing through Richard's (2/3, 2/3) point. If there were any in Richard's neons.doc, they seem to have been lost in translation.

If the core and shunt materials had linear magnetization curves, without saturation, I think any consequent electrical parameters (magnetizing inductance, coupling coefficient, leakage inductance) would be independent of the operating point.

The output I/V curve would be linear, as shown in my sketch below, since the leakage inductance would have a constant reactance at 60 Hz. Or not, because its voltage drop is in quadrature with that in the resistive load? I'll let somebody else take a shot at answering that, by handwaving or formulas or simulation.

Last Saturday, I made plenty of measurements on a tiny NST/ballast that was convenient. Don't have time or photographs to present the results today. There _is_ time to get the experiment description out of the way:

1. Chose one small shunted HV transformer with no nameplate. Primary behaves like it's designed for 120 V, 60 Hz. Then secondary V_oc is about 4 kV, and secondary I_sc is about 7 mA.

2. Used it with a variac, some power resistors, one "neon" element, and three digital meters:
- Benchtop 6-digit multimeter with Agilent brand, for true RMS AC volt measurement at transformer primary.
- Handheld multimeter with Fluke brand, for true RMS AC mA measurement in series with transformer primary or secondary.
- Homemade AC kilovoltmeter presented in OP above, used exclusively for AC volt measurement at transformer secondary.

3. Primary voltage was adjusted, using variac, to measured values closely approximating whole multiples of 10 V. All charts are based on the nominal step voltages, not the actual measured values.

4. Measured primary current, with secondary open and secondary shorted, from 10 V to 140 V. Both show incipient saturation at 140 V.

5. For the same set of primary voltages, measured secondary voltage loaded only by voltmeter (90-megohm), and secondary current burdened only by milliammeter (about 1 ohm).

6. For about six different resistive loads, and one luminous tube, primary voltage was set to 30, 60, 90, 120, and 140 V as measured. At each setting I recorded the voltage and current at secondary. That's more tedious than computing one from the other and the known resistance. It offers some protection against transcription blunders, because the computed V/I ratios should be constant in each sweep.

Pictures, charts, and conclusions to follow.

Re: Neon sign transformer power: a new look

Posted: Mon Aug 27, 2018 11:48 pm
by Rich Feldman
Just got clear on something that should've been obvious, and on which I speculated in prev. post.
The simple linear model of "AC voltage source with series inductance" yields an I-V curve that's one quadrant of a circle.
That's because the voltage drop due to leakage inductance is in quadrature with the voltage across a resistive load.

If we then include the in-phase voltage drop due to winding resistance, the curve from model is an intermediate shape
much like those from measurements.
Source impedances in this picture each have a magnitude of 500,000 ohms:
all real (500k resistor)
all imaginary (reactance of 1326 henries at 60 Hz)
a hypothetical mixture (100k + j*490k).

Minor effects from nonlinear magnetization are plainly evident in the real data.
But I bet linear LR models will be plenty useful for most predictions, such as what would happen if we put capacitors in the circuit.

Re: Neon sign transformer power: a new look

Posted: Tue Aug 28, 2018 5:18 am
by John Futter
so just to clear up the graphs
Blue = Real R load
Black dashes = immaginary reactance and grey line - mixture of both

Re: Neon sign transformer power: a new look

Posted: Thu Aug 30, 2018 2:14 am
by Rich Feldman
Yes, what John said. Sorry about posting a chart without labeling the curves.

This little transformer is responsible for first set of data from my lab. The device, and the method, are described a couple of posts ago.
Core laminations have no joints on the outside perimeter. Looks like the center leg is pressed into the outer section, a thing I've seen before in small motors but not previously in a transformer. Shunts have well-defined air gaps without need of paper spacers. Here sketched from a photograph.
nst_core.JPG (12.44 KiB) Viewed 7689 times
Here are the measured I-V curves for five different primary voltages and six different resistive loads, plus a luminous tube.
Black dashed lines connect the measured V,I points for each resistor configuration. Their straightness adds confidence that the measurements were transcribed properly. A red dashed line connects the measured V,I points for the luminous tube, which lit up fine even as primary voltage was variac'd up from 0 to 30 V.

In the next chart, the numbers are normalized to the measured V_OC and I_SC for each primary voltage. Colors are the same. On close inspection, the curve shapes are slightly sensitive to primary voltage. I bet that's from magnetic nonlinearity in the steel.
I added an iso-power line passing through V = 2/3, I = 2/3 point that Richard presented as a rule of thumb, long ago. So close to my "nominal primary voltage" measured curve, I suspect the experiment was somehow rigged. :-)
My intended application is on the lightly loaded, higher voltage side of the chart. With _this_ specimen, we can get 50% current at 80% voltage, or 35% current at 90% voltage.

Next report will include curves for simple linear models. That is, models with internal L and R values that don't depend on voltages or currents. Sorry about one hasty earlier post, where I said "linear" to mean straight-line I-V source characteristic.

Re: Neon sign transformer power: a new look

Posted: Sun Sep 09, 2018 5:30 am
by Rich Feldman
Here's a preview of a 100 mA AC/DC digital panel meter, put together much more quickly than the kilovoltmeter in OP. I hope it can serve on the high side of multi-kV circuits, like NST's and color TV sets, when put in a compact enclosure with an anti-corona shape. Other high side metering options include current transformers and optoisolators, I think neither trivial nor inexpensive for tens of mA and tens of kV.
A different digital panel meter of the same size, made for 0-900 mA DC, was considered and dismissed. Its built-in shunt resistor is less than 0.1 ohm. When fed with rectified AC milliamps, the reading wandered too much. it would need a very high-value capacitor to keep the ripple amplitude down.

The meter in picture is identical to that in OP. 0 to +30 volts DC, with automatic decimal point shift between 9.99 and 10.0.
AC or DC current goes through full-wave bridge rectifier to a 100 ohm burden resistor. 99.9 mA develops 9.99 V and is displayed as 9.99, with 1 watt dissipated in the resistor. A 220 uF capacitor is enough to take care of the ripple voltage. Meter indicates the average rectified current. For sinusoidal AC, we can get the RMS value by multiplying by 1.111.

[edit] Fusors and other sparky HV circuits are notorious for destroying series milliammeters. To this circuit I'd add a Transient Voltage Suppressor diode (between 10V and 25V) in parallel with the R and C. I think the C itself affords some protection. [/edit]

Re: Neon sign transformer power: a new look

Posted: Sun Sep 09, 2018 9:11 pm
by Richard Hull
Note some of these little meters use 3.3v MPUs and will not indicate a voltage below that value. They are mostly created for monitoring 12 and 24 volt storage battery systems. The hamfests and catalogs are flooded with these little chicom items ranging in price from $1.95 on sale to over $7.00 for the same little item. These little two wire wonders have no need to read a voltage below 3- 5 volts.

I notice three wires to your device. You might have one that has a true zero to 30 volt range. Two wires for a 9v battery to power the item up and a third wire for the voltage to be measured input.

Of course the free Harbor Freight digital meters could be worked the same way, but don't have the bright glowing display of an LED.

Richard Hull

Re: Neon sign transformer power: a new look

Posted: Mon Sep 10, 2018 2:58 am
by Rich Feldman
Yes, my tiny LED panel meters are from ebay for $2 or $3 each. Could have got some about the same for $7 at the flea market.
The voltmeters are all 0 to +30 with three wires; negative power and negative V_in share a wire.
The ammeters (in various ranges) have built-in shunts and four wires (two thin and two thick). Haven't looked to see how the two pairs are connected internally.

I just got a new cheapie DMM, like the ones from Harbor Freight. Figured on using it with the voltage divider in OP, and no rectifier, to conveniently switch between correctly scaled AC and DC kilovolt ranges. Then discovered that the meter, and others like it, are poorly suited for that plan.

We all know that most "nice" DMM's have standard 10 megohm input impedance.
Some of us have noticed that the "cheap" DMM's are often only 1 megohm. Caveat emptor, if you plan on using one with a standard HV divider probe!
Those input impedance are only for DC, and I learned new things today about AC voltage mode on various meters.
The "nice" ones are AC coupled, and don't respond to DC voltage of either polarity. Input resistance is infinite (viewed with a DC ohmeter).
The "cheap" ones, in AC V mode, look like a 1/2 megohm resistance in one direction and infinite in the other direction. When we apply +10 volts DC, the meter reads about 21 volts AC. When we apply -10 volts DC, the meter reads zero.
I think the presence of a transistor socket and hFE mode generally goes with meters in the second category.

Here are some meters and some of their properties. Not shown in picture are fancy Agilent benchtop meter and one Micronta analog V-O-M.
Turns out the fat red meter, from an estate sale, is UL listed and might be made in USA. Henceforth it will get more respect in my lab.

Re: Neon sign transformer power: a new look

Posted: Sat Sep 22, 2018 11:33 pm
by Rich Feldman
Haven't had much time for the next step: floatable digital milliammeter. As mentioned before, general purpose $3 digital multimeters are very limited for measuring AC. If we need an external rectifier, might as well opt for a whole package that's much smaller and more fun.

It took just a few minutes to build and test the whole circuit in breadboard form. Then hours, so far, putting it into a corona-resistant metal enclosure. The box has room for all the parts. So far, all that's installed is the DPM and a set of home-made receptacles for banana plugs.
Without banana plugs, the electrical subassembly can be lifted right out.

Re: Neon sign transformer power: a new look

Posted: Sun Nov 18, 2018 4:21 am
by Rich Feldman
The Altoids-box mA meter has been assembled, calibrated, and put into service.
Tonight I measured the first of my 15060 NST's, with primary voltages of 60, 120, and 140, and loads of short, luminous tube, 1R, 2R, 4R, 8R, and open circuit. Next time out, will crank it up to 277 (the nominal primary voltage), or 290 or even 320 as discussed in another thread.
The I-V chart is unremarkable, except for showing an apparent blunder in transcribing a reading at 60V input and 8R load.
Here's a close-up of the new instrument. The power switch is in middle position, to read the battery voltage when loaded by our 100 ohm current sense resistor.
A full scale reading on the digital panel meter would be 300 mA, with 9 watts in the resistor. In case that's ever wanted, briefly, the R is as big and as thermally isolated as I could easily manage, short of thermal attachment to the case and/or air holes. Next time, get a panel meter with a more appropriate native voltage range, like 2V or 0.2 V as Richard recommends in fusor FAQs. Then current sense R value and power dissipation are much smaller, but anti-ripple capacitor value (for AC use) needs to be much larger.

It would have been much easier to put a small analog mA meter in a box, with a similar bridge rectifier. But that wouldn't be so easy to read precisely from a safe distance.