Except, they flicker.
Here in the U.S. we are fortunate that our mains frequency is 60 Hz so the flicker is a bit less noticeable than it would be in the majority of the rest of the world where 50 Hz is used. For many inexpensive LED lightning strings, "full wave" bridge rectification - which would result in flickering at twice the mains frequency (e.g. 120 or 100 Hz for 60 or 50 Hz, respectively) - is NOT used, but rather half of the LEDs are on at any given instance: For some strings, it is every other LED, but others fully half of the string is lit at any given time and the lower flicker rate is more noticeable. I personally find this flicker to be rather annoying, but I can willfully ignore it, but some people find it very distracting and others say that it can cause eye strain and headaches over time.
In my experience, a lot of people people don't really notice the flicker unless it's pointed out - or perhaps they don't know how to describe it - but I find it to be somewhat distracting, so I was determined to find a "fix" for this - one that would not involve using DC, which often does not work! 1.
First, since this project involves high voltages, a few "weasel words":
This article describes the use of potentially dangerous voltages which could result in injury or death if safety measures are not taken. If you wish to experiment with what is described below, please be careful - and if you construct or modify any of the circuits on this page, please do so with safety in mind. As I cannot be responsible for others' actions, please do not attempt any of these unless you have experience in doing so and anything you do related to this page, you do at your own risk.
Additionally, the devices described may be suitable ONLY for powering some types of LED string lights and attempts to use them to power any other device could result in damage, injury or fire: If you are not thoroughly familiar with electronics, high voltage safety and related aspects, you should not attempt any of the projects below. Testing the suitability and safety of these devices for this - or any other purpose - is your responsibility.
The "audio amplifier" method
Using and abusing power transformers Here, we are using the transformer "backwards", applying a lower-voltage to the secondary to produce the high voltage (about 120 volts) needed to drive the LED lights. When using a power transformer this way, it's unlikely that it would be able to handle as much overall power nor will it produce the "expected" voltage when used this way - and there are several reasons for this. First, the primary will have been designed to handle slightly more power than will be output by the transformer - this, to overcome losses in the magnetics (eddy currents, imperfect coupling, etc.) There will also be higher instantaneous currents due to the fact that the reactance of the windings will cause the peak current to be offset from the peak voltage (e.g. power factor) meaning that more current must be conducted by the primary than one might think. Finally, to accommodate resistive and magnetic losses, the turns ratio of the primary and secondary will be offset to produce slightly higher voltage under no-load conditions so that at full power, the transformer's output will be close to the intended voltage. Because of the above, running a transformer "backwards" (e.g. power into the secondary) will likely result in reduced voltage and lower power-handling capability than it's rating when used in the way that it was intended. |
Having a decent assortment of AC power transformers kicking around I thought about using an inexpensive "Class D" audio amp module, fed with a sine wave, backfeeding the secondary winding of a transformer to get something resembling the U.S. 120 volt standard. With this method an easily-generated sine wave - or even a square wave - could be voltage-boosted and the frequency varied to drive an LED string at whatever frequency seemed high enough to render the flicker essentially invisible.
To do this test, I ordered a TPA3116 audio amplifier module from Amazon (the total cost: Less than US$10 for two of them)
and rummaged about in my transformer collection, finding a few
transformers that had 120 volt primaries and secondaries of around 6-9
volts. At this point the reader might ask "Aren't power transformers designed for 50 or 60 Hz?" - and the answer is yes,
but most power transformers will happily operate to at least ten times
their designed frequency - possibly even higher - with relatively little
impact on efficiency 2.
Powering the TPA3116 audio amplifier from 13 volts - which is about that which might be obtained from a "12 volt" LiFePO4 battery - I did some tests with several transformers, operating each at a different audio frequency - sourced by my trusty Wavetek 180 function generator - to determine which was the most efficient: This set-up is depicted schematically in Figure 1, below.
For this test, I used a string of 100 warm white LEDs with a stated rating of 7.2 watts, or 60 mA at 120 volts. As this lighting consists of a long string of series-connected LEDs, we can observe several things:
White LEDs, when driven to full brightness, typically need about 3 volts each, but they start to produce light (albeit very dim) at just above 2 volts.
Figure 1: The test set-up using a
Class-D audio amplifier and transformer to drive an
LED light string.
Click on the image for a larger version.- Via oscilloscope measurements, the individual LEDs just start to illuminate when the voltage across this 100 light string exceeds 70 volts.
- This string is actually three separate "sub-strings": Two of 33 LEDs and one of 34 LEDs. On each section of the string can be found a plastic module that typically consists of a resistor in series with the LEDs to limit current 3. Because LEDs are a type of diode 4 which light only when current is flowing in one direction - and it's likely that the "sub-strings" are not all be wired in the same polarity - it's likely that portions of the string will be lit at different times, depending on the instantaneous polarity of the AC waveform: See Footnote #1, below, for more detail.
The transformer used for most testing has a 120 volt primary and a 6.3 volt, 2 amp secondary which meant that it was rated for a bit more than 12 watts. Using this configuration, between 100Hz and 2000 Hz the Class-D audio amp pulled about 450mA at 12 volts - or around 5.4 watts to drive the LED string to a brightness that was subjectively equivalent to their being operated from 120 volts. Below 100 Hz, the current gradually increased without any change in brightness, reaching 625mA at 60 Hz indicating that it was actually more efficient to run the transformer at a frequency much higher than its designed 60Hz.
The variable frequency set-up also allowed me to subjectively determine above what frequency the LED string would appear to be "flicker-free". As expected, 50-60Hz was "obviously flickering" to my eyes, but this effect was significantly reduced by the time the frequency was increased to 100 Hz, becoming "just perceptible". As with many types of electronic displays - such as LED readouts - if one moved their eyes across the array of LEDs the flicker became quite apparent.
Increasing the frequency to 1 kHz makes the flicker itself imperceptible under normal circumstances and even though it can be perceived if one moves their eyes quickly across the LEDs, it is not distracting and above 2 kHz, even this effect disappears. With the simple set-up depicted in Figure 2, the driving frequency was audible because the laminations of the transformer would vibrate slightly: If you have sensitive hearing, you may want to consider this behavior if a constant tone is likely to be annoying.
It need not be a sine wave!
A bit more experimentation was done with this set-up with the function generator set to produce a square wave rather than a sine wave and the results were quite good: The LED brighness increased for a given voltage - but this was expected since, unlike a sine wave, a square wave will spend half of its time fully "on", illuminating the LED rather than this occurring only at the peaks of the sine wave. In other words, the "duty cycle" of the LED is higher. A similar effect can be caused by intentionally overdriving the amplifier with a sine wave, causing it to "clip" and in Figure 2, a bit of such clipping is visible. A means to generate a suitable square wave is described later in this article.
A few quick measurements indicated that while the LED string was noticeably brighter, its average current was still well within the "safe" range of typical small LEDs (e.g. well under 20 mA, average).
How many lights?
During this test I rummaged around, digging up several sets of LED lighting strings to see how many I could power with this simple set-up. The 100 LED string depicted in Figure 2 was no problem and a second, identical 100 LED string also seemed OK, but going beyond that seemed to cause the audio amplifier to start to "buckle".
This audio amplifier, based on the Texas Instruments TPA3116 chip is configured in "PBTL" mode meaning that although this is a "stereo" chip, the two channels are wired in parallel as "push-pull" which, in theory, could quadruple the output power, but with just a 12 volt supply, only 10-20 watts is possible, depending on the load impedance which, itself, is determined by the properties of the transformer that is used and how many lights it's trying to power.
One issue that arises with this simple circuit has to do with the fact that LED light strings conduct only at the peak voltage - and some LED strings are more sensitive to this peak voltage. This can result in some strings getting dimmer than others when the driving conditions change or when additional strings are added, loading down the output. This current "peak" during the tops of the waveform also has implications to the amplifier itself: At some point it will be unable to output the needed "peak" current and it will start to "buckle" (dropping in its ability to produce the peak power) due to this non-linear load even though it may not actually be at its maximum power ratings. Remember: Our LED lights are not linear loads like speakers and the amplifier will have a bit of difficulty driving them!
Something simpler?
This experimentation brought to mind that I could probably make something simpler: What about a single-ended FET switch driven by a square wave? Figure 3, below, shows a simplified circuit:
![]() |
| Figure 3: Single-ended inverter. The operating frequency is approx. 360 Hz with the values shown. |
For this, the same transformer with the 6.3 volt, 2 amp secondary was used as above and from my junk box, and for Q1, I grabbed an N-channel power FET with a 200+ volt rating and at least an 8-10 amp current rating, mounted to a suitable heat sink. The 1000uF capacitor (C4) is across the (nominally) 12 volt power supply to maintain low impedance for the circuit and the 10 ohm resistor with the 0.01uF capacitor form a "snubber" network that quenches some of the inductive "spike" that results from the FET turning off and the transformer's magnetic field collapsing, reducing it from well over 150 volts down to about 50 volts: I would not omit these components. Not shown is any sort of current limiting such as a fuse - and if you build something, I would definitely recommend it!
To drive the FET I used the ubiquitous 555 timer and the diagram in Figure 3 shows a variation from the typical 2-resistor circuit by adding a capacitor in series with pin 7 (R2) and this arrangement and resistance ratio causes the 555 timer to produce a proper, 50% duty cycle square wave: Using the "standard" 2-resistor arrangement 6 would result in a duty cycle of significantly greater than 50% and would likely reduce efficiency and increase heating. I have little doubt that the a bit of tweaking of the duty cycle could be done to optimize efficiency and specific loads, but the above does a pretty decent job overall.
For frequency control, capacitor C1 could be tweaked: Using the values shown for resistors R1-R3 the oscillation frequency is around 360 Hz and if you want this frequency to remain fairly stable, it should be a plastic capacitor (polyester, mylar, etc.) rather than any sort of ceramic capacitor which could vary in value more than 2:1 over a wide temperature range. If you wish to select a different frequency, the frequency is "more or less" inversely proportional to the value of C1 - see the article linked in the footnote for more information.
![]() |
| Figure 4: Testing on the workbench. On the scope in the background can be seen the square-ish wave with spikes, suppressed significantly by the snubber network. Click on the image for a larger version. |
The measured current into the LED string at the mains voltage output of the transformer was about 30 milliamps peak - or about 15 milliamps, average - well within the ratings of the LEDs and the power supply consumption was a bit over 600 milliamps - more than with the audio amplifier, but commensurate with the fact that the LED string was being driven with more current. Adding a second, identical 100 light string reduced the brightness of the first only slightly while not quite doubling the current, but tacking a 500 light string increased the current to about 2 amps - the likely safe rating of the transformer - and dimmed the 100 light string significantly: In actual use, both instances would be usably bright, but it showed the limits of this very simple circuit.
Because of its simplicity, the brightness of the LEDs, the voltage across them, the current through them and the power consumption of the circuit at DC will increase quite a bit with a higher supply voltage: To provide any degree of regulation, more complexity would be required, but for the purposes of providing illumination without annoying flicker, its probably "good enough".
As a sanity check, I used a variation of the above circuit using an external function generator rather than the 555 timer so that I could vary the frequency. With the transformer, the efficiency remained about the same from 100 Hz to 2 kHz indicating that it is remarkably insensitive to frequency variation.
If you build something like this, safety is paramount and anyone using it should be protected from high voltage and with proper fusing on the DC side. Additionally, it should be very clearly labeled as to its purpose with warnings to not use it for anything else. If you aren't thoroughly familiar with electronics and voltage safety, you should not try it at all!
![]() |
| Figure 5: The 350 watt modified sine wave (MSW) inverter before modification. It's nothing special - a cheap inverter. Click on the image for a larger version. |
As it happens I had an old 350 watt DC-AC inverter kicking around 7. I'd bought this perhaps 20 years or so ago at Harbor Fright. Some more modern inverters are computer-controlled, but not this one as evidenced by the fact that its output frequency was around 59.5 Hz as measured with my Fluke 87 voltmeter and slowly drifting upwards: If it had a microcontroller, it would have been more or less "spot on" 60 Hz and would not have drifted 7.
Pulling it apart I spotted a "daughter" board that looked the the "brains" of the inverter and on it I could see that one of the chips was a 556 dual timer - two 555 timers in a single package. Using my oscilloscope and probing the output pins (pins 5 and 9) I saw a 119 Hz waveform on pin 9 - twice the 59.5 output frequency and sure enough, there was also a 4013 dual flip-flop on this same board being used as a divide-by-two to get an exact 50% duty-cycle square wave. A bit of further sleuthing (e.g. varying the DC supply voltage) revealed that the other half of the 556 timer (it's output on pin 5) varied in duty cycle along with that of the "modified sine wave" output of the inverter.
![]() |
| Figure 6: Parts changed on the "daughter" board to make it operate properly at nearly 10x the original frequency - See text. Click on the image for a larger version. |
On a "modified sine wave" inverter, it's output is NOT really a sine wave at all - and it's not just a square wave output (e.g. the AC voltage swinging positive to negative) either, but there is a "dead time" between the swing from negative to positive when the output voltage is zero, and another of the same duration when it swings from positive to negative. For a modified sine wave inverter the goal is for the "average" voltage to remain the same over time so that the power being pulled by the load (which would ideally be resistive) would be constant, but if the power source - a 12 volt supply in this case - were to increase, the peak voltage would go up as well. In order to maintain the same "average" power, as the supply (e.g. battery) voltage increases, the "dead time" also increases so that the voltage is output for less time - and vice-versa when the supply voltage decreases. This is the "PWM control" property mentioned above.
For more about a "modified sine wave" and power inverters in general, see this Wikipedia article - link.
For a resistive load, the result is that the power being delivered by the inverter is, on average, constant with this method and in general, it works pretty well. With an LED load, this will result in a variation in the duty cycle of the waveform which can cause varying brightness, but for a given load, the LED brightness says fairly constant as the DC input voltage changes.
Additionally, on the inverter pictures in Figure 7, I added some radio frequency interference (RFI) filtering to prevent it from producing interference that might impede the use of HF and VHF radios in the immediate vicinity: These sorts of modifications are similar to those documented in a previous blog entry - "RF Quieting a Samlex 150 watt Sine Wave Inverter" - link. The only changes that I had to make for this modified inverter was to use 0.0047uF Class Y capacitors instead of 0.1uF Class X capacitors across the AC output: With the higher operating frequency - and especially the square-ish wave output, a 0.1uF common-mode capacitor, during initial testing, caused a rather significant amount of extra current draw from the DC supply when unloaded because the high-frequency components of that waveform were being conducted through it: The 0.0047uF capacitor across the output - along with the bifilar inductor and the other class-Y capacitors to ground along with filtering on the DC input of this inverter was sufficient to prevent any noticeable "QRM" (interference) when operating HF and VHF on a nearby antenna.
The advantage of modifying this commercial inverter is that it was already in a suitable case and that it was suitable for driving as many LED lights as I was likely to ever need! While I'm skeptical that this inverter could really have handled 350 watts continuously without overheating - and I also suspect that raising its operating frequency has contributed to this "de-rating" as well - I don't doubt that it could handle at least 200 watts of LED lights: At one public service event - a foot race in a remote mountain location that went overnight - a lot of LED strings were connected to it and it was loaded to at least 80 watts and run for over 14 hours with no problems at all. Since then, it has also been used to power LED lighting while camping on several occasions.
Once again, unless you are thoroughly familiar with electronics and high-voltage safety, I would recommend that you DO NOT attempt this sort of modification.
Final comments
I've used the modified inverter several times and unless you "jiggle" your eyes, the flicker is imperceptible - and I suspect that this would be true for the other methods described here as well.
As was done with the modified inverter, I would strongly recommend that it be very clearly marked as depicted in Figure 7 so that no-one is likely to "accidentally" plug anything other than LED string lighting into it. This method should work with practically every simple series-string LED light, but there are a few types that should not be used with this:
- Any LED lighting with an animation or "chase" controller. These often have a small box that controls the lighting to make it flash in a pattern or sequence. Many of these devices have a simple controller on them that may be powered by a capacitive dropper which would be damaged by anything other than standard mains current. Additionally, some of these lighting controllers use the mains frequency to "clock" the sequence: That type of lights - if it works at all - would operate at many times the original animation speed.
- Anything with a "wall wart" or voltage converter. These types would already be powered from DC and should not flicker unless they are "animated" (e.g. "chase" or patterns) in some way. The voltage converter may be damaged if powered from this.
- Any sort of programmable or "smart" lighting. Again, this type of lighting would be powered from DC and would not benefit from a higher mains frequency, and its power converter might be damaged in doing so.
- There are commercial, 400 Hz inverters available that are intended for aeronautical purposes, and these may occasionally show up on the surplus market. It is possible that these - if they output the correct voltage - might be usable for driving LED string lights such that they are essentially "flicker-free".
As stated in the "Weasel words" above, you do any of the above at your own risk!
* * * * * * *
Footnotes:
- Some of the Diode/LED sections are actually wired in opposite
polarities which means that during each cycle of the AC waveform it may
well be that one of the three sections of LEDs illuminates at an
instant when the other two are off - and vice-versa: This is one of the
reasons why one can't necessarily resort to powering the LED string
with DC rather than AC to get rid of the flicker altogether - and to do so would probably mean that only a portion of an LED string would light. Additionally, at these elevated voltage, DC becomes quite hazardous as switches, fuses and loose connections are likely to arc dangerously as there is no "zero crossing" as found in an AC waveform for such an arc to self-extinguish.
- While intended to operate at 50 or 60 Hz, power transformers will typically operate fairly well at many times this frequency - even well into the kHz range. "Slightly" higher frequencies (e.g. 100-300 Hz) may well result in higher efficiency for the transformer than the design frequency owing to lower losses in the transformer cores, but as one goes still-higher, the losses of the core can start to mount - not to mention the increased losses of the windings themselves due to "skin effect" as well as inter-winding capacitance at the very high end of the audio range. In other words, power transformers can often be used in lieu of "real" audio transformers in vacuum tube circuits if you don't care too much about very low (<50 Hz) and very high (>5 kHz) audio frequencies - and in this application, where we are driving LEDs at mid-audio frequencies, they will work fine.
- The value of this resistor varies depending on the number of lights in the string, the mains voltage and how much current each LED can handle. For the 100 light LED string depicted above, this "module" contained a 1/2 watt, 2.4k resistor. As each string contains about 33 lights - each of which drops 3 volts, that means that when fully lit there is about 99 volts across all of the LEDs. As a 120 volt sine wave peaks to about 155 volts, that means that during the peak of the AC cycle, 56 volts is being dropped across the 2.4k resistor and Ohm's law tells us that this equates to about 23 milliamps - well within the current tolerance of small LEDs like this, which can easily handle 20-30 milliamps: During this peak, the instantaneous power across the resistor is about 1.3 watts - well above the 1/2 watt rating of this resistor. Considering that the LEDs are lit only during half the cycle - and only when the voltage is swinging above 99 volts - the average LED current is actually far less than 23 milliamps and the average power dissipation of the resistor will be less than 1/2 watt.
- The module with the resistor will sometimes contain a diode to prevent reverse voltage from being applied to the LEDs. In practice, this is a good idea since modern nitride LEDs (e.g. white, blue, green) will degrade very rapidly if reverse voltage is applied across them. It would seem that the makers of LED strings have determined that, with a bunch of LEDs in series, that a diode - which would cost a few pennies - is not really necessary for a reasonable lifetime of the the light string and it is often omitted on long strings with dozens of LEDs. On shorter strings - say, 10-15 lights - I've noticed that the diode is usually present as the "reverse" voltage on each LED would be much higher without it, along with a larger resistor capable of dropping more voltage and dissipating more power.
- Be careful when powering any mains-powered device from something other than a pure sine wave - or any waveform at other than a standard mains frequency. While "resistive" devices (heaters, incandescent lamps) are forgiving in terms of how they are powered, other devices may malfunction or even be destroyed by something other than a sine wave. In particular, devices using a "capacitive dropper" (e.g. a capacitor in series with the mains voltage to limit current) would be immediately damaged by non-sinusoidal waveforms - or even a sine wave at other than a standard mains frequency. For example, if I had connected my "Kill-A-Watt" power monitor to any of the devices described here, it would likely be damaged immediately by a non-sinusoidal wave, not to mention too-high a frequency as either of these would permit far too much current to flow via its "capacitive dropper" circuit.
- A diode or two is often used along with resistors to coax the 555 timer configured as an oscillator to produce a square wave, but this shows a means by which the addition of an extra resistor can do this. Among the places where this is described is in this article in Electronic Design Magazine from March 20, 2022, linked here.
- For the record, this inverter, purchased at Harbor Freight in the late 1990s or early 2000s - before this company seemed to standardize on their own "house brand" names. It bears the "Power On Board" brand - and is apparently made by "Vector" and is part number "VEC032POB". These units can sometimes be found on EvilBay, but note that there are likely other devices out there that could be similarly-modified.
- It is often the case that microcontrollers' clock frequencies are controlled by a quartz crystal and as such, would be accurate to within a few hundred parts-per-million at worst. In this case, an observable frequency shift - from a cold start and through a few minutes of warm-up - meant that we are not likely dealing with a crystal-controlled microcontroller. If this inverter had used a microcontroller to set the frequency it would be unlikely that it could be "hacked" to operate at a frequency other than that for which it was designed.
* * * * * * *
This article stolen from ka7oei.com
[END]






















