# Is there any reason nowadays to use a neon indicator lamp instead of an LED?

I have a bunch of old neon indicator lamps recovered from old AC devices and used to indicate power status.

I wonder: is there any reason nowadays to use the neon lamps instead of LEDs? I see they are still widely available.

LEDs are cheaper even after considering the need for a diode in AC applications. They blink half as often in AC applications, but emit much more light per watt.

In what cases are neon indicator lamps still preferred?

• Electricians screwdrivers. – Andy aka Sep 17 '19 at 12:44
• I've seen them used as voltage regulators, though mostly in very old circuit designs. – Hearth Sep 17 '19 at 13:43
• Luminous efficacy isn't an unambiguous virtue. I have tons of electronic crap that annoys me with its bright LED lights that I don't want to see at night. – chrylis -cautiouslyoptimistic- Sep 18 '19 at 10:38
• Neon lamps are were also used as voltage limiters, surge arrestors, sawtooth oscillators and coarse voltage references... – rackandboneman Sep 18 '19 at 11:40
• @KalleMP Only if driven at redline. If derated, they can last a million hours. Often you derate them just becuase they're too bright at redline – Harper - Reinstate Monica Sep 18 '19 at 21:42

Neon bulbs use microamps of current when fed through a dropping resistor directly from the AC line. LEDs need 10× to 100× the current and can't be fed AC directly.

• @slebetman: The ~100 milliwatts we're talking about here rarely matters in line-powered equipment. – Dave Tweed Sep 17 '19 at 21:28
• Yes but that just means the stress on "microamps" also don't matter. Its main advantage is that it can feed directly off AC rather than efficiency - the main reason people don't use them these days is that equipment will invariably have a DC power rail anyway to power the CPU (heck, even a switch has a microcontroller these days) – slebetman Sep 17 '19 at 21:44
• @slebetman: The microamps is important, because it means that a modest 1/4 W resistor can produce the necessary voltage drop. An LED requires both a greater voltage drop AND more current, causing much more power to be wasted. – Dave Tweed Sep 17 '19 at 21:46
• Also note that that efficiency doesn't just come from exotic or expensive LEDs. I run a bathroom nightlight LED at around 0.5-1 milliamps, and it could be even lower but still usefully bright for illumination when the eyes are darkness acclimated. It would be overkill for an indicator light. It cost more than the red and yellow LEDs from the same shop, but it was still very cheap. – piojo Sep 18 '19 at 6:31
• @ThreePhaseEel funnily enough I was just complaining that 800uA was too bright through 3 parallel white LEDs (I couldn't see a much fainter spot from a different source) but my current source wouldn't give me less – Chris H Sep 18 '19 at 11:27

I think you'll find neons have more resistance to ESD, transient overvoltage, and high temperatures.

You can do many interesting things with neon lamp circuits, all of which are curiosities in 2019 (well maybe they're useful if you need EMP resistance, but I suspect we'll have bigger problems if that's a factor). Neons wear out, for one thing. From the above document:

When I were a lad, I designed a monitor system for Linotype machines (yes, that long ago) using a neon bulb oscillator triggering a then-newfangled C106 SCR (in that horrible package with the 0.05" nominal wide leads), to operate a stepper relay with precious metal contacts multiplexing thermocouples.

I don't see any real reason other than to save a few pennies to use neons as indicator lamps. An NE2H runs at around 2mA, which is sufficient to give more brightness from a decent LED (not all packaged indicators use decent LEDs, however). The neons will wear out if they're powered constantly (rated life is 12,000-25,000 hours which is less than 3 years 24/7).

• You may have missed the part in the book about millions of hours lifetime if they are intentionally underdriven. They may be underdriven anyway to put an indicator light at a sensible brightness. – Harper - Reinstate Monica Sep 18 '19 at 20:36
• @Harper The packaged indicators such as these have 25,000 hours for all neon lamps. I don't doubt you can get a feeble glow at 1/10 or 1/100 the current, but that's not the design decision that's typically made, and virtually all the neon indicators I own, from the WTCP solder station to the power bars, are flickering or dead. Never had an indicator LED fail in a commercial product. – Spehro Pefhany Sep 18 '19 at 20:53
• Wow, you really did not read your own reference. Page 16. Derating 1/4 is more than sufficient for human lifetimes. – Harper - Reinstate Monica Sep 18 '19 at 21:45
• Spehro, please explain the circuit above! And what was the Linotype thing you created? It was an actual lead-melting line-o'-type ejecting Linotype? Practically a mechanical supercomputer! Wow! Also where have I seen your name around the Internet? Were you on Usenet or something? I remember your name, though to an English speaker it's quite a distinctive one. I'm sure where you come from, it's like being called "John Smith"! – Greenaum Sep 19 '19 at 5:14
• @Greenaum Yes, Spehro Pefhany was on sci.electronics.* at least as far back as the '90s, I think. I remember seeing him on there back then. – Fleetie Sep 19 '19 at 7:07

Have a look at this video by BigClive, the light effect (scroll towards end of video to see it) would require a significantly more complex circuit and/or obscure components if you would use LEDs and tried to get the same (random) light effect.

This is the circuit that is used:

I admit that this isn't a "killer application" of neon lamps but more a fun project.

Not an EE answer but aesthetics.

LEDs look uglier. Neon lights occupy more bands in the yellow orange spectrum while LEDs only have a few spikes. See this Q&A on spectrum of LED bulbs (hint: they're just blue LEDs with Phosphor coated lens).

Although more than just a simple lamp, The Nixie tube was a a popular display in the 50s which was used in various technical instruments like voltmeters and frequency counters. It's revival is in part due to its unique, vintage look. Plenty of examples of DIY hobbyists making their vintage displays.

Hellbus [CC BY-SA 3.0 (https://creativecommons.org/licenses/by-sa/3.0)]

• Red, yellow or orange LEDs aren't blue+phosphor (though a few greens are, like the lime green I've been using) – Chris H Sep 18 '19 at 11:30
• Neons look uglier when they inevitably begin to flicker and fail. ;-) – Spehro Pefhany Sep 18 '19 at 15:42
• @SpehroPefhany ugh, hate that neon flicker sooooo much LOL – Doktor J Sep 18 '19 at 21:01
• They used to (maybe still do) make neon bulbs with UV emission and phosphor inside the glass to turn that into red, green, blue and maybe other colors. The brightness was a bit disappointing, however, IIRC. – Spehro Pefhany Sep 18 '19 at 21:05
• @SpehroPefhany they're still used: My freezer (only a few years old) has a green neon power indicator. I have to turn the light off in the room or cup my hand over it to see whether it's on. The phosphors do age but they're not great even new. I think they're xenon rather than neon inside, due to the shortage of UV from neon – Chris H Sep 19 '19 at 8:38

If your equipment is mains voltage, no low-voltage DC available, a neon is a lot easier. As Dave Tweed said, for one thing the resistor can be smaller. Neons can also tolerate the odd voltage surge, and other out-of-spec conditions. An LED has a delicate and tiny structure necessary to it's function. A neon is just some gas and 2 bits of copper.

Things like kettles, radiators, etc, that are mains-only, typically use neons. If it's cheap garbage from Ebay / Amazon, they might (ab)use an LED cos it's cheaper and they don't care about reliability. But why do they always choose blue?

• Red, or green, will make (at least some) people think of a meaning (red for some error, green for ok (in the sense of "something was actually checked and found ok)). Blue doesn't have that problem. – Guntram Blohm Sep 18 '19 at 12:22
• Blue is cruise control for cool. – Harper - Reinstate Monica Sep 18 '19 at 20:13
• Blue is cruise control for insomnia and headaches from chromatic aberration. – R.. GitHub STOP HELPING ICE Sep 19 '19 at 0:26
• I think what it is, is blue LEDs were hideously expensive, 20 years ago, til that Japanese bloke discovered a way of making them cheaply. And got a Nobel Prize for it. Even when his process was implemented, they were expensive for a while. So blue LED = flashy! quality! much chrominance! Then the Chinese forgot this, and carried on using them slavishly anyway. Now it's a sign of cheap crap where the designers aren't fully aware of what they're doing. Like those UK plugs they make, with insulation sleeves on the earth pin. – Greenaum Sep 19 '19 at 4:50
• Incidentally, you can get cheap cigarette lighters with a LED torch in the base. Just an LED and a few tiny batteries, with a springy metal contact for the switch. Usually they're white. But some come with a blue LED, even though blue is the colour the human eye can see least. Green, or better, yellow, would be a better choice if you can't have white. But they use blue ones. A blue torch. What's the point? I suppose it's because they use the same higher voltage as white so can use the same battery arrangement. Even though you can't actually use them to see with. Oh, China! – Greenaum Sep 19 '19 at 4:54

Have you actually tried building mains neon vs LED indicators? I bet you haven't.

Here, try a project. Sometimes, a leg of split-phase or 3-phase AC power will drop out at the poletop. Either neutral gets lost, causing L1 to rise and L2 drop, or a leg is lost, causing L1 to be backfed through 240V appliances. Make a very useful gadget that grabs L1, L2 and N, and visually indicates whether L1-N and L2-N are very coarsely the same voltage, and L1-L2 is coarsely 2x the others. The user switches off breakers and watches to see if that affects brightness one to the other. Having 3-4 lights and having the user watch for differences in brightness is fine. Durability a million hours. Needs to be simple and very low parasitic load, energy budget 720mW (3ma). Go.

However, your point is fair that 99% of electronics innovation is done on the low-voltage side of the wall wart. Neon has no place there; it would need an active circuit driving it, and that compares disfavorably with LED.

Neon and LED are different indicator types for different regimes of electricity. When you argue against neon, you're arguing against mains voltage as a supply source. As a big fan of houses having low voltage DC systems so they can happily exist in power-out situations without generators or inverters -- I can't really disagree.

• The problem with low-voltage DC systems is the voltage drop. You're using much more current, MUCH more! So the voltage drops get huge, and you'd have to run cable like the stuff on a car battery, if you were gonna start powering all electronics from it. High voltage is much better for distribution. It's like a microcosm of the electrical grid. The longer the distance, the higher the voltage they use. It's the same at home, a scale of dozens of metres means some high voltage, low current, system wins. It's best to convert to DC at point of use. And pretty cheap too. – Greenaum Sep 19 '19 at 5:01
• We don't have split-phase AC in Europe (in fact, I have 3-phase lines directly to my energy meter), so the issue you mention was completely unknown to me. – FarO Sep 19 '19 at 8:23
• @FarO then do it 3-phase, it's just one more phase. European houses also lose phases and neutral. Bonus points for getting the 400V lights to be the same brightness as 230V. And have more energy budget; 2000mw. Visible indicator (differential brightness is fine) for a lost phase or neutral, given that 3-phase appliances will backfeed the dead phase. – Harper - Reinstate Monica Sep 19 '19 at 16:04
• @Greenaum yes, you have to be selective about what you're powering, but lighting and electroncs loads are no problem. For others it is a matter of competent design. HVAC: passive solar design, water heater: gas or solar, etc. LVDC as a unit replacement for buffoonishly designed all-electric homes won't work, obviously. – Harper - Reinstate Monica Sep 19 '19 at 16:18

The need for neons can be related to the industry. For example, equipment that wears out in nuclear facilities must typically be replaced with identical gear, because the cost of qualifying new equipment to operate in the system is so great that it makes no sense to use anything but the old design. If the old design had neons, the replacement typically will too.

The need for neons can also be cosmetic. I build nixie tube clocks for friends and use neons for the colon separating hours and minutes. The color closely matches the color from the nixie filaments, and LEDs just wouldn't look as good.

They also have very low current draw and so are sometimes used as indicators in older designs in emergency power or ride-through applications.

• less relevant to your answer itself, but relevant to your anecdote: there are red-orange LEDs out there that very closely match the color of neon/nixie tubes; since neons have spectral peaks around 600nm, 625nm, and 650nm, try looking specifically for an LED with a wavelength of 600-650nm; places like Mouser let you filter by wavelength(s). – Doktor J Sep 18 '19 at 21:13
• That's helpful but I bet they don't have the same flicker! Neons just have an indefinable unique quality. Nothing else is quite the same. And they're nice, people like them. Maybe it's just nostalgia. Actually on a nixie clock it probably is! – Greenaum Sep 19 '19 at 4:56
• @Doktor Thanks for the heads-up. – Bort Sep 19 '19 at 12:15
• At that Mouser link, there are some red/yellow bi-color LEDs that might better emulate the spectral emission of neon lamps. Putting the LEDs on half-rectified DC, perhaps with a small and not-wholly-adequate smoothing capacitor, could emulate the flicker too. Using $V_{ripple} = \frac{I_{load}}{\mathit{f} \times C}$ if we accept ~1V ripple and our input frequency is 60Hz, and the LEDs use 20mA, $1V = \frac{0.02A}{60Hz \times C}$, $1 \times 60 = \frac{0.02}{C}$, $60 \times C = 0.02$, $C = \frac{0.02}{60} = 0.0003F$ or 300uF; so a 200-300uF cap would give you some good soft flicker. – Doktor J Sep 24 '19 at 18:22

The current requirements of a Neon are low enough that they can generally be driven directly from the mains with a simple and cheap series resistor.

If you try to drive a LED in a similar way you have at least two problems.

• The reverse breakdown voltage is nowhere near high enough to withstand mains voltages, so you end up having to add another diode (potentially another LED) in inverse-parallel.
• Substantially more current is needed, say 10ma or so which is enough to make a resistive dropper consume unconfortbally large amounts of power. You can solve this by using a a capacitive dropper instead but capacitive droppers have their own issues which require extra components to mitigate.

So neons tend to win out in "pure mains" applications, things like indicators on sockets, plugs, power strips, wall switches, surge protectors, heaters with simple electromechanical thermostats and so-on.

If a device is intelligent enough to have a microcontroller on it then LEDs generally win out.

• As a point of clarification, a device "intelligent enough to have a microcontroller on it" already necessarily has a low-voltage DC power rail somewhere, making it trivial to power a LED (even if the µc uses something weird like +15VDC, dropping that with a resistor is trivial) – Doktor J Sep 18 '19 at 21:15
• rainierwolfcastle.jpg – Greenaum Sep 19 '19 at 5:07

Most regular LEDs have about 2 Vdc drop in forward mode and need about 1 mA to light up. although 10 mA is recommended with standard LEDs.

In addition, they can only withstand about 2 Vdc in reverse mode, which means that they need a rectifier diode in series to protect them when the ac voltage changes polarity.

In this configuration, the LED lights up only half of the time, which may reduce its brightness; this could be good or bad depending on the application.

Neon lamps use a 100 K to 250 K, 1/4 W resistor (see http://www.farnell.com/datasheets/57560.pdf).

Neon lamps are not that bright and in certain applications may not stand out when lit, especially in highly lit areas.

With time, neon lamps develop flickering and electrode polarization (only one of the electrodes glows).

Neon lamps are available in various base sizes (see https://www.bulbtown.com/neon_lamps_and_light_bulbs_s/909.htm). Some even come with an internal resistor sized for a specific operating voltage.

Several answers have already discussed the technical reasons, but they don't discuss the applications where those reasons would apply.

A lot of times, you'll see Neon indicators in industrial-grade machinery, for three reasons, two already mentioned:

1. Cheaper / easier to drive directly from main voltage (this is important as main voltage rises, like 480V 3-phase systems);
2. When under-driven, lifetime is extended substantially;
3. Simpler installation and circuitry (you can run neons directly off 480V / 277V AC with proper resistors);

As a result, industrial applications using these higher-voltages (note that ANSI still considers 480V a "low-voltage") can save power and cost with appropriately sized neon's. (Sure, it's minimal savings, but typically you need a reason to change something, not keep it the same.)

There are some other curious aspects of industrial systems that generally hold true as well:

• A lot of the logic is 24V DC or 24V AC using relays instead of transistors, easy to drop neons right in;
• The logic often activates raw 480V AC (or 277V AC) equipment -- again, easy to drop a neon in;
• Safety is always a critical issue, because these are often large, dangerous pieces of equipment, so if the bulb dies there should be a possible backup -- once again trivial with neons;
• Industrial equipment is generally expected to take a less-skilled person to service it than smaller electronics (like computers, etc.) -- neons have very common hot-swap plugs on them (LED's are often soldered);
• Industrial equipment may not even have a DC line! -- this means you have to rectify the AC for every LED;

When these aspects are considered, it should make it easier to see why a good chunk of (especially older) industrial-grade equipment would prefer neons to LED's. Especially when considering that if a neon "dies", it's often just extremely dim (vs. completely off as an LED typically is) so it still has minimal functionality.

As an example and anecdote, when I was building the glass manufacturing machines, one feature our machines had was that when plugged in, there was an indicator system that told you if the 480V outlet you plugged into was wired properly. It was extremely cheap to build with Neon's (a few lights and resistors), but would have been slightly more costly (and complex) to build with LED's. (Sure, considering the cost of the machine it was barely noticeable so who cares, but we had no reason to use LED's there and this system has been used by the company for decades, so why spend extra time and money designing a new one?)

Source: I used to maintain and build industrial machinery for a plastic injection molding facility, and a glass machine manufacturing facility.

This is a very niche use, but some neon lamps can be used as logic elements due to the voltage difference in the striking vs maintaining voltages.

You can read about somebody using this, rather than a microcontroller, to control a clock:

Is there a reason? Sure. Suppose you're making a relay-based computer, with no semiconductors, and wanted memory, say PROM. Each bit is a dip switch, or a little jumper wire. But how do you connect those to data lines? You need a voltage controlled switch, like say a relay... or a gas discharge tube, and neons definitely fit the bill, although normally you'd be driving them pretty hard that way (10-20mA) even if for short pulses. So, neons don't quite fit the bill, you need gas discharge tubes if you care about longevity, but I'll share a secret: you can use a conducting GDT to trigger a neon bulb, just for the visuals. The current needed to trigger a subsequent relay will be carried by the GDT, while the neon bulb looks pretty.

With diodes or neons or GDTs, you can drive the memory in a cube matrix form, with just one relay per each line/plane per dimension, e.g. a 16x8 matrix can select a byte from a kilobit (128 byte) array. The address decoder tree for such an array takes ~48 relays. Without GDTs, you'd also need a thousand coupler relays in addition to the address decoder tree, and other circuitry such as bus driver relays remains the same.

The high-ish voltage needed to strike GDTs that are cheaply available (~90V) is not as nice to deal with as the low (5-48V) voltages needed to trigger bus coupling relays, but a GDT takes way less board area and is cheap vs. a relay - if you find a good deal on auction sites.

It's hard to buy large (>1000) lots of fast relays for $0.10 each (I need them fast - 1ms or faster when new) - you find one deal or two and then it's crickets for 6-18 months. Low voltage GDTs for$0.10 each show up in reasonably sized lots about every 2 months, although the recent downturn has somehow doubled the supply. You'll get the couple thousands pieces you need much sooner than you otherwise would. That's still about 2/5ths compared to distributor pricing in >1ku quantities, so if it's a hobby product that you're putting together yourself, you have plenty of time to trade for dollars, since assembling the PC boards with hundreds of parts per board seems to take forever. But even primary market distributor prices for GDT's aren't crazy - you can get one for under \$0.50 in 1k quantities.

They may weigh about as much as the smallest of mass-manufactured SMT low voltage relays because of the ceramic envelope, but you can pack them 5x as densely, and they are easier to solder, too.

Disclaimer: There are many ways to design things that work well, and the above isn't by any means a golden goose, nor even a wooden goose, and it only works well enough anyway.

• Quite a theoretical answer, since is anyone making a relay-based computer, with no semiconductors, in 202x? if you have examples please post them, because I'm interested! – FarO Jan 26 at 9:59
• @FarO Well, I am building one. No point in writing too much about it before it's all done, but I can share what I have learned thus far. As for progress: I got about 8k relays sitting in a cabinet to my left, and about 2k already soldered onto several 9U Eurocards (PCBs, no hand wiring). I have probably got 2-3k too many, after I figured out how to replace relays with GDTs, so I may be able to make much more robust execution units vs. the original design, so in the end it's still a 10k relay machine, now with 8k GDTs (1kbyte addressable space, double the initial design, 20% RAM, 80% PROM). – Kuba hasn't forgotten Monica Jan 26 at 19:39
• My point of pride is the ~5W per Hz of clock performance figure (<10W/Hz in the most conservative estimates based on measurements of subcircuits). It's a dynamic design, 0W used with lights off and clock stopped. I don't think there are/were any stored program relay computers of some practical use, ever, that came within an order of magnitude of that. It's easy to toss a lot of relays at a problem and need a diesel genset installed outside your house just to run it. That was definitely my anti-goal :) Very preliminarily, the major source of acoustic noise is the case fan so far (2k relays in). – Kuba hasn't forgotten Monica Jan 26 at 19:50

I remember using neon indicator lights as crude, inaccurate, but simple voltage references.

The reality is that in today's electronics there are not the voltages present for neon operation. And there are for LEDs.

• I agree with you for today's consumer electronics. Industrial electronics often have to sense and/or be powered by a wide range of standard voltages, so voltages that will light up neons in these applications are more accessible. – Bort Sep 18 '19 at 14:15
• A mobile phone, maybe, yes. But in the home there's always mains voltage. Actually I'd quite like a phone with neon indicators on it. You could do it with a boost convertor. I wonder if we could troll Apple into adding them to Iphones? Hipsters love all that retro shit. I mean, so do I, but for better, deeper reasons! And I started it before they did! – Greenaum Sep 19 '19 at 5:05