DIY Drivers On The Cheap

You can drive LEDs or COBs with a constant voltage supply using a very simple constant current circuit on the output. This won’t be as efficient as buying a constant current driver, but it can be VERY cheap and easy. The simplest such circuit consists of one resistor!

Say you have a constant voltage power supply that puts out 12 volts. You have some LEDs that run at 2.5 volts at 700 ma. If you connect four in series, that will total 10 volts. So your series resistor would have to drop 2 volts at 700 ma. Ohm’s law says E / I = R, so 2 volts / 0.7 amp = 2.86 ohm. You can’t get a 2.86 ohm resistor, but Digikey has 3 ohm 2 watt resistors for 29 cents.

So lets see how the circuit works with a 3 ohm resistor. Let’s assume the current stays the same, for now. At 700 ma a 3 ohm resistor would drop E = I x R volts, so it will drop 2.1 volts. The LEDs will get 9.9 volts at 700 ma. Pretty close! Actually, the LED current will drop a bit so this is just approximate. Notice that I pick the next higher resistance you can get, so the current would not go up.

The resistor power dissipation would be P = E x I so that is 0.7 x 2.1 = 1.47 watts. This is why I specified a 2 watt resistor. If you use a lower rated one, it would burn up.

Now lets think about efficiency: If you are burning up 2.1 volts out of 12 volts total in your resistor, the efficiency of this circuit is 1 - (2.1 / 12) = 0.825 or 82.5%. But you also have to count the efficiency of your 12 volt power supply. A very good switching supply might have 95% efficiency, so the combined efficiency is 0.95 x 0.825 = 78.3%. You see why people pay a lot of money for a Mean Well constant supply with 95% efficiency!

Because of the low efficiency when you burn up a lot of power in your resistor, you want to run as many LEDs or COBs as you can with the fixed voltage. But you always need at least 10% of your power burned up in the resistor for the sake of stability. If you operate with too little power in the resistor, this circuit can be very unstable, because LEDs or COBs can change their current draw a lot with a small voltage or temperature change.

Don’t forget to use a good heat sink. Power LEDs or COBs MUST have a good heat sink or they will get too hot and burn up. It’s a good idea to measure the heat sink temperature when you first turn it on, and for the next couple of hours. You want to see where it stabilizes. 85 C is probably too hot!

@dbrn32 @BIGE @Bogleg @MattyBear @Venomous @nostril @Myfriendis410


So here’s a way to get a better regulated DIY constant current circuit: We can use a linear regulator called an LM317T. Digikey has them for 58 cents each. This regulator is designed to put out a constant voltage but we can use it as a current regulator by the addition of a single resistor. It can pass 1.5 amps and have an input of 40 volts and an output of 37 volts, with a total power dissipation of 20 watts. Do NOT ever try to pass more than 1.5 amps through this part or it will burn up. A safe design margin would be 1.2 amps, but that will require a good heat sink.

Here is the circuit:

Say you have a constant voltage power supply that puts out 48 volts and a COB that wants 36 volts at 1 amp. You connect the positive terminal of the COB to the power supply + output. Then you connect the negative terminal of the COB to the LM317 input terminal. You connect the part of the regulator circuit marked OUTPUT to the power supply - output. The LM317 has to deliver 1 amp, so we can calculate R = 1.2 / 1 amp. So R is 1.2 ohms. (This is because the LM317 always wants 1.2 volts between OUT and ADJ.) This resistor will dissipate 1.2 volts x 1 amp, so it has to be able to handle 1.2 watts. If we look at Digikey, they have a 1.2 ohm 2 watt resistor for 29 cents.

Now, how much power is the LM317 dissipating? It is dropping 12 volts at 1 amp, so that is 12 x 1 = 12 watts. Reasonable for a TO220 package, but it does have to be mounted on a heat sink. Fortunately, TO220 packages have a hole you can use to screw it down to the heat sink, but beware: The LM317 connects the output pin internally to the mounting tab! This is for better heat dissipation, but it means your heat sink will be electrically hot (in this case at +12 volts) unless you use a silicone pad or mica wafer with heat transfer goo under the part and a special insulating washer for the screw. That is highly recommended.

Notice I put the COB first in the series. That was so the LM317 only sees an input of (48-36) = 12 volts above minus, which may be grounded. The LM317 can only take an input voltage of 40 volts, but in this circuit it only sees 12 volts! You can use this trick to go much higher total voltage. For example, a 170 volt power supply with 150 volts of COBs would only look like 20 volts to the LM317! 20 volts at 1 amp would be 20 watts, which is the permissible power dissipation of the LM317.

Now obviously, the efficiency in the first case would be 1 - (12/48) = 75%. Not great, but if you had power supply you could adjust to put out 38 volts to drive a 36 volt COB, it would be 1 - (36/38) = 94.7%! Never design the LM317 to drop less than 2 volts. It gets very weird if the out to adj voltage is less than 1.2 volts.

Again, pay attention to heat sinks. Now you have to mount the COB on a heat sink and you have to mount the LM317 on it as well. Cobs send a bit less than half their wattage into the heat sink, so a 36 watt COB is going to send 18 watts and the LM317 is going to send all 12 watts. So the total is 40 watts. An Intel processor is 60 watts, so a PC heat sink with fan would certainly do in this case. Or something bigger without a fan.

@Covertgrower @Daddy @Onlythebest79 @MAXHeadRoom


Thanks for posting that @1BigFella I’ve book marked this so I can find it again later.

1 Like

I have a circuit that goes up to about 2 amps and drives COBs right off the AC line without a power supply, but it takes a lot more parts, Like 11. It is very specific for certain COBs that total about 133-160 volts at somewhere under 1.4 amps. The closer to 160 volts on the COBs, the more efficient it is, since 120 VAC gives you about 172 volts DC when you use a full wave bridge and filter capacitor.

If anybody has real electronics skills and can assemble circuits on a perf board, I can post it.


I don’t need board info, yet. But I’d like to reserve the right to ask for it later lol. This is pretty cool! I have some leds I may try this with for a light to go over clone dome.

1 Like

LM317 is especially good for little supplemental strips or LEDs when you already have a 12 volt power supply. If you have a 12 volt 3 amp supply, you can run 3 of them in parallel. Each 1 amp strip has it’s own LM317 so the usual series/parallel rules don’t apply.

I have the basic electronic understanding. I’d love to build my own lights. Post away the schematic please.
The schematic would be very much appreciated by myself and I’m sure others.


It’ll take me a couple of days. I need to draw a clean schematic and upload it, then make a Bill of Materials so you get the right parts. Patience please.


@1BigFella I’d actually be looking to run something like 4 mono’s around 700ma or so. They’re xm-l2’s, 3 volts nominal max current is 3a. I have quite a few so could do more or less if it makes circuit less complex or whatever. But looking to do something around 10 Watts.

You could hook multiple LM317s together in parallel to handle 1.5 amps each. So two of them could be connected with their own resistors to get 3 amps. There’s nothing wrong with that. There is no sort of feedback or crosstalk between them. Each would just supply 1.5 amps into the common output line. You can get up to any current you want by using multiple regulators and resistors.

So would you want to connect a series chain of those xm-12s to a higher voltage power supply? I think you can get 18 volt 3 amp laptop power supplies pretty cheap. You could use three LM317s to deliver 700 ma into 4 chains of 5 of those LEDs each. Or a 12 volt supply could run chains of 3 LEDs. At 700 ma per chain, you can run as many from one power supply as it has output amps.


I have a whole box of random power supplies in the garage, I’d have to go through it but I’m sure there a few that would work. I have some salvaged heatsinks I could probably get up to 5 maybe 6 mono’s on if we don’t push them too hard. I’ll have to find some time to go dig through it.

Please don’t attempt this unless you know what you are doing with electronics.
The first time you plug it in, stand WAY back and have a fire extinguisher handy in case you made a mistake.

Okay, here it is:

Sorry about the weird formatting. The website has a mind of it’s own.

Bill Of Materials: COB Driver (Digikey part numbers and prices)

Fuse:SlowBlow 4 amp__________________________507-1531-ND_______0.36
Bridge:3 amp 200 PIV___________________________KBPC102-ND______1.17
NTC:Negative Temperature Coefficient Thermister____495-6652-ND_______1.13
C1:270 uf 200 volt____________________________493-5455-ND_______2.63
L1:The COB string (2 Vero29C or 3 Vero29B in series)
Q1:90 watt NPN 10A 140V
__ FJA4310OTU-ND___1.91
Q2:____2N3904 NPN______________________________2N3904FS-ND______0.29
R1:____0.62 ohm 2 watt___________________________P0.62W-2BK-ND_____1.24
R2:____62 ohm 1/4 watt___________________________62QBK-ND__________0.10
R3:____470 ohm 1/4 watt___________________________CF14JT470RCT-ND__0.10

6 VDC is just any 6 to 12 volt wall wart DC power supply

COBs are not specified because you have to pick your color temperature, and decide if you want two of the C series or three of the B series. Three of the Bs are more efficient. Veros are in the $21 to $27 range each at Digikey. Feel free to buy them elsewhere, if you can find a better deal.

VERY IMPORTANT: Use a series string of COBs or diodes with a forward voltage of 130 to 165 volts at the chosen current. If higher, the LEDs may not light up. If lower, you blow up Q1. I

R2 sets the current output of the LM317 to 19 ma. This feeds into the base of Q1 to turn it on.
R1 is the current sense resister. 0.62 ohm gives you a voltage of about 0.76 volts with 1.23 amps flowing through the COBs. This voltage turns on Q2 which supplies negative feedback to the base of Q1 by soaking up the 19 ma, thus turning it off.

You can change the COB current by changing the sense resister. Divide 0.73 volts by the resistance to calculate the new current.

These are resistors Digikey has in stock:

0.5 ohm 1.46 amp
0.51 ohm
0.56 ohm
0.62 ohm
0.68 ohm
0.75 ohm
0.82 ohm
0.91 ohm
1.00 ohm 0.73 amp

Get 2 watt or higher.

Be aware that Q1 and any LM317T connect electrically to their mounting screws! You don’t need a heat sink mount for the LM317 but you MUST use a heat sink with Q1 and it must be insulated. That means you need to get a TO3P silicone pad or mica wafer with heat transfer goo and an insulated washer to mount it on the heat sink. Once you have it mounted you can use an ohm meter to be sure none of the leads are in contact with the heat sink.

Not bad for $9.51 in parts. Obviously you need about a 3" x 4" piece of perfboard to build it on, some solder, some 18 gauge stranded wire. some heat shrink tubing for the Q1 connections, and a 6 foot extension cord for the power cord. I just strip a little insulation off about 6 inches from the female outlet end to connect the board, and then I can plug the wall wart into the female end of the extension cord. My camera is screwed up. I should have a new one coming on Friday and I’ll post a pic of my board.

@dbrn32 @Greasemonkey @Covertgrower


Pretty sweet! Thanks for taking the time to put all that together.

I think this would be a cool project for someone that has the knowledge or training. Especially since a lot of us would already have the majority of the general components.

I would have to get a lot of parts. But this is pretty sweet. Thanks for posting that @1BigFella

The parts are really no problem, if you use Digikey. Just go to their web site and put in all those part numbers. They do charge a few dollars for shipping but nothing unreasonable. No minimum order. I have ordered a single part and they just ship it out the same day. They do offer quantity discounts, so if you wanted to build 10 of these some parts would be cheaper.

They also have a very intuitive search engine. Like put in “resistor”, and pick “thru hole parts” and they show you 20 billion resistors. Then you pick the resistance you want and the wattage and do select “active” and “in stock” to see what they actually have ready to ship. You can sort by any field (like price). Easy as pie.

1 Like

I suspect a circuit can be even simpler: Nothing says you can’t run LM317s in parallel, input and adj pins connected together but each with their own current limit resistor. This could get rid of Q1 and Q2, their resistors, and the little power supply.

In fact, TI even shows a circuit with three of them in parallel this way. The LM317T is limited to 20 watts of power dissipation, so some attention has to be paid to the input versus output voltage and currents. For example, you could run three of the Vero29B COBs with two LM317T regulators. Each regulator would supply 900 ma to the COB string, so it would run at (52 x 3) = 156 volts and 1800 ma. With 172 volts input from the AC front end, the regulators would have to drop 16 volts. At (16 v x 0.9 amp) = 14.4 watts each: Success!

You could also run four of the Vero29D COBs at 2100 ma and stay at about 18 watts of dissipation in each LM317, using just two of the LM317s.

At just 58 cents for a LM317, we may have hit on a super cheap driver circuit. I’m going to try building one.

I see that Digikey is not stocking the VERO29 COBs now. Got them from RapidLED.


Keep us posted on that build! Thanks for doing this @1BigFella

Shouldn’t take long. Ordered last night and both RapidLED and Digikey say they have shipped!


@1BigFella I’m working with driverless COBs, so your design might be the way to go. Especially if I can use a cob that puts out decent light. The cheap Chinese smart chip ones don’t. I’m going to save this whole post. It’s nice to have an engineer around.

Software engineer, actually, but most of my software worked at the chip level so I learned some EE on the way. Now retired, but I worked at it for about 35 years. Before that I worked in medical research, and before that I got a degree in Biology.

The inspiration for my driver designs is that you can select COBs (or lots of LEDs) with a forward voltage in the 133 v to 156 volt range. 120 VAC full-wave rectified and filtered gives you a voltage of 170 VDC. (It’s because 120 VAC is actually 120 volts RMS, so equivalent to 120 VDC. But the peaks are ± 170 volts!)

The AC line charges up the filter capacitor on it’s peaks, and then it discharges through the COBs until the next charging peak comes. The main “engineering” part is to get the capacitor value right so it doesn’t discharged down to the voltage level the COBs blink off at the end of the discharge cycle. That is determined by the COB current and forward voltage.

Anyway, you don’t want the COBs to pulsate, so we need a device to soak up that extra voltage every cycle. I started with a big honking transistor, but with reasonable limits I can use LM317s. In my latest design with three Vero29B’s they add up to 156 volts so the LM317s have to soak up 14 volts and pass 900 ma each (I use two of them). That’s well below the 20 watt power limit of the part.

And the good news is that they only have to soak up that 14 volts for about 1 millisecond out of every 8.3 millisecond cycle! The rest of the cycle the voltage is dropping to about 2 volts, so the driver efficiency is higher than 14/170 = 92%.

Anyway, let me build it and we shall see. But the key is to get COBs with the right forward voltage so it doesn’t need any transformers or a switching power supply. Vero29 B and D series just happen to have that. Others may too, but I have not checked them out.