LED Constant Current Source Scheme

[Robert Atkinson 2Participant @robertatkinson2]

Posted by Martin Kyte on 10/11/2020 20:19:06:

LM3404 is a good simple chip for a first design. 1 Amp

I was building little regulators initially, I tend to go for modules now as they have become more available. No point in reinventing the wheel when you don't have to. The scientists just want anything that works. The other issue is we have depreciated our PCB wet etch since we got the CNC PCB machine so I can't do Aluminium substrate board at the moment unless I send it out.

regards Martin

Hmm, costs 3 times as much as an LM317 and needs more supporting components.

Lots of ways to do this, LM317 is quick, easy, cheap and works. Not as efficent as PWM but not a electrically noisy either.

Horses for courses

Robert G8RPI

[Robert Atkinson 2Participant @robertatkinson2]

One other thought,

Don't use unfilterd PWM for a machine light – LEDs make excellent strobe lights.

Robert G8RPI.

[Gene PavlovskyParticipant @genepavlovsky]

Robert,

When we're making one-offs for personal use, 1 EUR vs 3 EUR part doesn't really make much difference. When you go into production, cutting cents is important. For hobby it doesn't matter that much. Switching converters naturally require more components, and produce more EMI as well. It should be carefully weighed whether one should be used. It might not even be more efficient, depending on the details.

If power supply voltage and the Vf of the LED chain are selected carefully, a linear regulator can be very efficient and a better solution. E.g. for my high-power ring LED lamp I'm using a 24V power supply "brick", and 8 CREE XT-E High efficacy LEDs. Those have max Vf=2.85, so the string of 8 has Vf<=22.8V. With ~1.2V to drop, and 700mA max current, the regulator would have to dissipate less than one watt, with about 16W going to the LEDs. Around 94% efficiency, I doubt a buck driver could do it any better.

I like the AMC7140 constant current regulator so far. One input capacitor, one resistor for setting the output current. PWM input for dimming (no need to use it if not needed). Dimming can also be done by using a pot in series with the current setting resistor. Low dropout (0.5V at 700 mA which is the max rated current). It's a Chinese-developed chip and has to be bought from China, they are less than a dollar per piece.

What do you mean by unfiltered PWM? I assume that with high enough PWM frequency (e.g. 20 KHz), there should be no flicker.

[Les Jones 1Participant @lesjones1]

I have used cheap step up switching voltage regulators (Based on the LM2577S) modified for constant current output. This is a link to the modification **LINK**
The same principle could be used with buck regulators The disadvantage of this idea is there is a 1.25 volt drop across the current sense resistor. This is not much of a problem when driving a series string of LEDs but for driving a single LED it wastes too much power.

Les.

[Gene PavlovskyParticipant @genepavlovsky]

Les,

Looking at Fig 35 in the datasheet for LM2596, I don't see a current sense resistor, only "programming" feedback resistors with rather large values. Doesn't seem like those would dissipate any appreciable power.

Edit: I think the current sense resistor voltage drop you mentioned only applies to the boost converter you were talking about. Might not be an issue with a buck converter.

Edited By Gene Pavlovsky on 10/11/2020 22:36:30

[Les Jones 1Participant @lesjones1]

Hi Gene,
The idea does not use any current sense input on the chip. It uses the feedback input for voltage regulation. In both cases (Buck & boost.) the adjustable voltage version would be used. The chip adjusts it's output attempting to maintain 1.25 volts on it's feedback input. So by choosing a resistor at the bottom end of the chain of LEDs that gives a voltage drop across that resistor of 1.25 volts with the desired LED current and connecting that point to the feedback input means that the chip adjusts it's output voltage to a value that gives a current through the LED chain that produces a drop of 1.25 volts across the resistor at the bottom of the LED chain.

Les.

[Anonymous]

There are plenty of switchmode ICs specifically designed for driving strings of LEDs with constant current from suppliers such as Analog Devices. These devices incorporate PWM control of brightness. It seems sensible to use a device intended for the purpose rather than force a device intended for voltage control to act as a current source.

Andrew

[Gene PavlovskyParticipant @genepavlovsky]

Les, right, I understand what you're saying.

Andrew, I fully agree. It seems reasonable to use a purpose-designed chip. Besides LM3404, mentioned earlier, here are some more options available from Mouser at reasonable price:

IS31LT3953
MAX20053
AL8843Q

The latter two can even be used as high-side drivers, so the LEDs negative pole doesn't have to go back to the IC.

Sometimes, though, it would be nice to pick up a (cheap) ready-made module from China, which could be used as is. So far I haven't found a PWM-dimmable DC-DC LED driver module like this. Perhaps I'm not looking for correct term. Searching for dimmable LED driver produces tons of typical devices that operate from the mains voltage.

[Robert Atkinson 2Participant @robertatkinson2]

Posted by Gene Pavlovsky on 10/11/2020 22:01:01:

Robert,

When we're making one-offs for personal use, 1 EUR vs 3 EUR part doesn't really make much difference. When you go into production, cutting cents is important. For hobby it doesn't matter that much. Switching converters naturally require more components, and produce more EMI as well. It should be carefully weighed whether one should be used. It might not even be more efficient, depending on the details.

If power supply voltage and the Vf of the LED chain are selected carefully, a linear regulator can be very efficient and a better solution. E.g. for my high-power ring LED lamp I'm using a 24V power supply "brick", and 8 CREE XT-E High efficacy LEDs. Those have max Vf=2.85, so the string of 8 has Vf<=22.8V. With ~1.2V to drop, and 700mA max current, the regulator would have to dissipate less than one watt, with about 16W going to the LEDs. Around 94% efficiency, I doubt a buck driver could do it any better.

I like the AMC7140 constant current regulator so far. One input capacitor, one resistor for setting the output current. PWM input for dimming (no need to use it if not needed). Dimming can also be done by using a pot in series with the current setting resistor. Low dropout (0.5V at 700 mA which is the max rated current). It's a Chinese-developed chip and has to be bought from China, they are less than a dollar per piece.

What do you mean by unfiltered PWM? I assume that with high enough PWM frequency (e.g. 20 KHz), there should be no flicker.LM317

That is all true, The problem is thread drift. The original post belittled a pefectly valid approach – the LM317.

Why use a specialist chip for a hobby project (unless it's to learn about the chip)?

By unfiltered PWM I mean ut that, Switched DC without an inductor (or posibly resistor and capacitor) in series so the current through the LED is continious. LEDs will happly strobe at tens of kiloherz and this bings the risk of motion freeze with a machne light. I investigated white LEDs for use s a strobe for a day job project. I was worried the persistence of the phosphor used would limit their usefulnes but on chrt G8RPI.king with a high speed detector and 'scope found that they actually produce much shorter pulses than a standard xenon flash tube. You would have to add a quench circuit to the flash tube to even get close.

Robert G8RPI

[Michael GilliganParticipant @michaelgilligan61133]

Posted by Robert Atkinson 2 on 11/11/2020 10:33:39:

Posted by Gene Pavlovsky on 10/11/2020 22:01:01:

That is all true, The problem is thread drift. The original post belittled a pefectly valid approach – the LM317

.

Robert … As the originator of this thread, I must object to that misrepresentation [my emboldening]

…. celebrated would be a much more appropriate word.

MichaelG.

[Gene PavlovskyParticipant @genepavlovsky]

The original post by Michael regarding the LM317-based CC regulator looked positive to me.

I agree the LM317 approach is valid, also valid in some circumstances are plain current limiting resistors and specialist chips. A hobbyist is more likely to have both the resistors and LM317's already in their stash of parts, specialist chips would have to be ordered on purpose. In the hobbyist case using these specialist chips might bring no tangible improvements (maybe save you a tiny bit on the electricity bill). LM317, though, can drive high-power LEDs, which is very wasteful and impractical with current limiting resistors.

I do like playing with LEDs so I do order specialist LEDs, might as well order a specialist chip to see what works best, and learn new things. The specialist driver chips I mentioned above (e.g. AL8843QSP-13, MAX20053) accept a low-frequency (<=500, 1000Hz recommended) PWM input for dimming, but this just sets the output current. The chips themselves switch at a high frequencies (1, 2.1Mhz), I doubt there would be any flickering in this case.

There might be flickering in my AMC7140 circuit, although I am not sure at 20 KHz there could be any strobing with machine tools running at <=2000 RPM (33.3 Hz)? Besides using as high frequency as practical, what do you suggest to reduce the probability of flickering in my case?

[Martin KyteParticipant @martinkyte99762]

"The chips themselves switch at a high frequencies (1, 2.1Mhz), I doubt there would be any flickering in this case."

That's the max regulating frequency. Essentially the chip switches on at the frequency set by the design and stays until the current rises to the sense voltage. Current is supplied to the load during the off period from the stored energy in the inductor.

PWM at high frequencies and adding an output capacitor is your best option.

Personally I have never seen strobing effects in my workshop , indeed any workshop from any light source.

regards Martin

[Robert Atkinson 2Participant @robertatkinson2]

Posted by Michael Gilligan on 11/11/2020 10:43:13:

Posted by Robert Atkinson 2 on 11/11/2020 10:33:39:

Posted by Gene Pavlovsky on 10/11/2020 22:01:01:

That is all true, The problem is thread drift. The original post belittled a pefectly valid approach – the LM317

.

Robert … As the originator of this thread, I must object to that misrepresentation [my emboldening]

…. celebrated would be a much more appropriate word.

MichaelG.

My appolgies, you are correct, it was oher responses that questoned it. I relied on memory rather than checking.

Robert G8RPI.

[Michael GilliganParticipant @michaelgilligan61133]

Thanks for that, Robert

MichaelG.

[Author]

Posts