LED Lighting (350 mA). Is there any standard for # of LEDs is a string?

[Zero999]

I was thinking use the TLV431 as a reference for an op-amp. If you used a quad, such as the LM324, it would actually work out cheaper than four TLV431s.
-------
That's odd. Does the label/manual say what frequency it's pulsating at? What does the waveform look like on an oscilloscope? I think it's likely you're mistaken. The symbol for steady DC is a solid line over a dashed line.
⎓

To your first point: that would defeat the purpose here. The idea is to have the controllers as POL regulators, one regulator per string of four LEDs. This allows me to place the strings in any physical configuration. Consider it as a "19 V, 350 mA, 4-Led module".
Imagine a "splitting box" with twelwe 19 V outputs powered from one laptop supply. The LED strings will do the rest themselves.

The voltage headroom on my design also allows me more freedom in cabling (losses).

Concerning the "pulsating DC": of course you're right, bad choice of words on my side, sorry. I interpreted it as "mediocrely regulated DC".

-----------------------

Something that might interest you (you seem to like playing with circuits).
The TLV431/Darlington power BJT is dangerous, which I only found out after having serious problems with my first circuit. An emitter follower (which this circuit basically is) can oscillate with significant amplitude at >1 MHz frequencies. In a discrete circuit it's usually not a problem, but add in a TLV431 which has a lot of gain, and it gets very hairy.
The TLV431/BD679 combination proved impossible to tame, which made me go back to the BD437 plus a relatively large base resistor. This is stable and works reliably. R1 as 1 kohm and R2 as 5.6 kohm works even better.
You live and learn


PS: (added later). Do you remember the "backlighting" thread where temperature stability was brought into the discussion?
Well, look at this:

If you're set on strings for four LEDs, then why use the TLV431? The LM317 would do and still have plenty of headroom for cable losses.

I'd say 19V is a bit on the high side, for strings of four + a linear regulator. Have you considered hacking the power supplies? It's highly likely changing a resistor will get the voltage down to between 14V and 15V, which still gives enough headroom for a linear regulator.

[mariush]

You could use a basic linear driver like  BCR421 which can do up to 350mA (absolute max rating somewhere around 500mA) using an external resistor (defaults to 10mA without one)
32 cents if you get 100, 20 cents if you buy 1000

Nice little device, but not very precise. With the LEDs being 1.50 Euro a pop (=6 Euro), I'd rather be certain of my LED current.

But Thanks.

The built in sense resistor of around 95 ohm (for the 10mA current limit) will be less precise, and will allow for +/- 10% as the datasheet says, so 9..11 mA.

You parallel an external resistor with that built in resistor, for example on Page 11 , Figure 7... it shows  2.1 ohm used for 350mA current.
So you could choose a good quality low ppm resistor to get let's say 320-340mA of current, then even with all the variation of the internal 95 ohm resistor and external resistor due to temperature change, the current in the series won't go over 350mA of current.

Basically... I'd look at a 2.2 ohm or a 2.4 ohm resistor ... for example https://www.digikey.com/en/products/detail/koa-speer-electronics-inc/SR733ATTE2R40F/9845654

Or you could even have two resistor footprints by the driver chip, and for example parallel two 4.7 ohm resistors (common value, mass produced, pennies, and less heat per resistor)
example : https://www.digikey.com/en/products/detail/vishay-dale/CRCW08054R70FKEAHP/2226710


Keep in mind that the TL431 will also drift with temperature, as will resistors and capacitors in same circuit, and if you have these on same PCB with the LEDs the pcb will probably be heated up to 60+ degrees Celsius
So account for that when you think other solution will be more precise at keeping current at 350mA.

[nctnico]

You can bin the leds so that the forward voltage of every series ends up very close to the forward voltage of other series that are in parallel. Super easy to make a test fixture of some sort to measure the forward voltage and brightness at various current levels.

That is way too much hassle. Typically you buy the LEDs from the manufacturer binned for color & brightness but you still have to expect differences large enough that make parallelling several strings problematic.

They do this parallel thing with led strips for monitor backlights and in other places and nobody complains about differences in brightness.

They likely have a way to deal with proper current sharing otherwise you will see difference in brightness in the panel. Been there, done that.

[Zero999]

A few transistors can be used to make a switching regulator with current limiting, which can be used to drive a string of LEDs. The zener diode can be removed, or you might want to keep it and increase it to a higher voltage, such as 15V,
https://www.romanblack.com/smps/a04.htm

[Benta]

Keep in mind that the TL431 will also drift with temperature, as will resistors and capacitors in same circuit, and if you have these on same PCB with the LEDs the pcb will probably be heated up to 60+ degrees Celsius
So account for that when you think other solution will be more precise at keeping current at 350mA.

I think you missed my replly #19 and the LED current plot.
I'm really not worried about TLV431 temperature drift. Even if if was 10 times worse, it's still good.

[Kjelt]

Is there even some kind of standard?

AFAIK only difference is between LV (upto I believe 50VDC but could be a bit higher) and HV. HV you need more isolation larger distance between traces etc.
And there is a difference in driver being mains isolated or not. Latter needs extra safety.

LEDs strings in parallel don't work well. You'll see differences in brightness. I've seen current mismatches up to 25%. For the best results, you need a current source per string (even if it is a simple dropping resistor).

I disagree, I have seen many ledlight pcbs from A brands and they use 3 to 4 parallel strings even without resistor. I can take pictures for you, from pro lighting fixtures for parking garages costing €150 a piece to kitchen lighting. The trick is to use A brand leds from the same bin. For our hobby use buy a reel and use those they are also from the same bin.
The delta in forward Voltage drop is less than 0,1%. But when one led dies or is damaged then you have to replace the pcb in total ofcourse (average 25000 hours).

[Benta]

The delta in forward Voltage drop is less than 0,1%.

Source/documentation?

[nctnico]

LEDs strings in parallel don't work well. You'll see differences in brightness. I've seen current mismatches up to 25%. For the best results, you need a current source per string (even if it is a simple dropping resistor).

I disagree, I have seen many ledlight pcbs from A brands and they use 3 to 4 parallel strings even without resistor. I can take pictures for you, from pro lighting fixtures for parking garages costing €150 a piece to kitchen lighting.

That may be, but my experience is different (also with monitor backlights). When looking at white LED datasheets from Osram, they specify a variation in forward voltage of about 1% within a bin. It is possible that the lamps you have seen, have the PCB traces act as a dropping resistor just in case. LEDs from the same batch may have a much closer match but I doubt a manufacturer is going to bin LEDs within a 0.1% match.

[Benta]

Thanks to a lot of helpful inputs and feedback here, I think I've now arrived at my final circuit. Some adjustments/tuning were indicated.
For a 19 V input, I've settled on a 5-LED string. This leads to a power dissipation in the current regulator circuit of around 1.5...2 W, which is acceptable to me.
LED current is set at 340...350 mA (pretty standard for power LEDs).
Neither voltage nor temperature stability are an issue, I attach plots of the LED current for anyone interested.
Cost is around 60 cents for the the controller plus 7.50 Euro for the LEDs. This is not a cheap China product, but uses OSRAM quality LEDs.
Feel free to copy/modify the design as you like. (KiCAD design file (.zip) is also attached).

Cheers.

[mariush]

I don't know where you get that 1.5 euro per led from ... I can find Osram Oslon 150 degree leds at under $0.5 in quantity (600+)

You have the CM series with CRI 90+ , the EM one with CRI 80+ and the PM with CRI 70+

The better CM ones are 1$ each  for 10 on Digikey, 0.7$ for 100 : https://www.digikey.com/en/products/detail/ams-osram-usa-inc/GW-CSHPM1-CM-KSKU-XX58-1-350-R18/6599016
Same price for the EM ones (CRI 80+) : https://www.digikey.com/en/products/detail/ams-osram-usa-inc/GW-CSHPM1-EM-LPLR-XX58-1-350-R18/6599021

Sure add VAT and shipping but we're still at much less than 1.5 euro per led.

Anyway, unless those osram are better binned or your need those lenses on the leds ... these Samsung leds seem to be just as good for much lower price, at 15 cents if you buy 100 : https://www.digikey.com/en/products/detail/samsung-semiconductor-inc/SCS9WT93HPL2WLS03F/6569400
Still CRI 90, but it's 5.9v -ish forward voltage, 150mA current ...

[Kjelt]

The delta in forward Voltage drop is less than 0,1%.

Source/documentation?

Testreports of reels of leds.
With the 3.xx V types you will have max ten mV's differences on leds from the same reel ( ofcourse when in series with the same current).
If you get different reels , different production dates your mileage may vary.

[Kjelt]

That may be, but my experience is different (also with monitor backlights). When looking at white LED datasheets from Osram, they specify a variation in forward voltage of about 1% within a bin.

You always keep a safe margin with your specs or you get many returns.
So lets say it is 0,05% in reality that is ten mV's difference in Vfd.
If it is evenly distributed you have no problem. If you are really worried measure each led and put it in different new bins with tighter spec. But as I said even A brands don't bother.

It is possible that the lamps you have seen, have the PCB traces act as a dropping resistor just in case. LEDs from the same batch may have a much closer match but I doubt a manufacturer is going to bin LEDs within a 0.1% match.

That could be but the traces are pretty wide.  I will do a measurement on them again, measuring each led for Vfd and the voltage between the leds (trace resistance).

[Kjelt]

I can't find this back in the datasheets. Most manufacturers do Lm and CRI binning.
 Only a few manufacturers do Vfd bins it seems.

[nctnico]

With very large batches you can do your own verification and gamble it but if the parts start to vary wider for some reason, you have no claim at all.

A fellow engineer once told me their company needed 1% capacitors for a product. Because the 1% capacitors where expensive, they bought 5% capacitors and tested them theirselves. The out of spec capacitors where used elsewhere so nothing got lost. After some time they started to receive batches of capacitors where all capacitors where outside 1%. Turns out the manufacturer had started checking the capacitors themselves to sell as 1% tolerance.

Closer to home: I have been bitten as well by components that got worse. At some point a batch of a products started to show strange behaviour. Turned out the decoupling capacitors where not as good as they used to be. Same part number and still in spec but simply made worse.

[Zero999]

Here's what I was talking about earlier. It requires just under 1V of headroom.

It depends on your requirements. Does it actually matter if the current falls a bit below the set point? You'll probably not notice if it's 300mA, rather than 350mA.

I suggest you measure the voltage of the power supply and the forward voltage of the LED strings at different currents.

[deadlylover]

Which LEDs are you using? Were they the OSLON square on eBay binned below the BBL? (like this 3500k) Those are an excellent choice.

I've always wanted to try them but I have been a Nichia fan lately. (it helps that Cutter Electronics in Australia is kinda local to me, so I just use their strips for DIY projects)

A while back I upgraded my room/desk lighting using the Nichia Optisolis 757 series. Museum grade lighting because I'm a bit of a colour pencil hobbyist too. 

[Benta]

Here's what I was talking about earlier. It requires just under 1V of headroom.

OK, so right now I have three resistors,* a BD437 and a TLV431. (*) Ok one is actually three to hit the value and spread power.
And now I need the same PLUS an LM358 (plus an additional dropper resistor for the reference). Not really a simplification to my mind.

[Zero999]

Here's what I was talking about earlier. It requires just under 1V of headroom.

OK, so right now I have three resistors,* a BD437 and a TLV431. (*) Ok one is actually three to hit the value and spread power.
And now I need the same PLUS an LM358 (plus an additional dropper resistor for the reference). Not really a simplification to my mind.

It wasn't supposed to be simpler, but more efficient. Now the transistor can be cheaper and doesn't even need a heatsink. A cheaper TL431 can also be used if the values of R2 & R2 are changed to 160k & 12k.

It depends on how many strings of LEDs you want per luminaire. If you want to use two or more strings, then my circuit requires fewer components. For another string all that's needed is an extra transistor and current sense resistor, then there's the LM324, if you need three or four strings.

[Benta]

It wasn't supposed to be simpler, but more efficient. Now the transistor can be cheaper and doesn't even need a heatsink. A cheaper TL431 can also be used if the values of R2 & R2 are changed to 160k & 12k.

It depends on how many strings of LEDs you want per luminaire. If you want to use two or more strings, then my circuit requires fewer components. For another string all that's needed is an extra transistor and current sense resistor, then there's the LM324, if you need three or four strings.

In a perfect world, yes. But (just as an example) the OSLON voltage binning (K2...M2) spans 2.75...3.2 V. So the BC337 will either run nice and cool or be in flames, depending on LED bin availability.

Second, and I've said this before: the idea is to have the 4/5/6-LED strings running autonomously at 19 V for free placement.

[Zero999]

It wasn't supposed to be simpler, but more efficient. Now the transistor can be cheaper and doesn't even need a heatsink. A cheaper TL431 can also be used if the values of R2 & R2 are changed to 160k & 12k.

It depends on how many strings of LEDs you want per luminaire. If you want to use two or more strings, then my circuit requires fewer components. For another string all that's needed is an extra transistor and current sense resistor, then there's the LM324, if you need three or four strings.

In a perfect world, yes. But (just as an example) the OSLON voltage binning (K2...M2) spans 2.75...3.2 V. So the BC337 will either run nice and cool or be in flames, depending on LED bin availability.

Second, and I've said this before: the idea is to have the 4/5/6-LED strings running autonomously at 19 V for free placement.

You have many LEDs in series, so the variation in forward voltage drops will cancel each other out. You won't have six 2.75V LEDs, or 3.2V LEDs, but an even distribution of forward voltages, giving a total forward voltage of about 3V.

As I said before, you should get hold of the real components and measure the voltages at different currents. It's probably also a good idea to breadboard it before finalising the circuit.

I strongly recommend running the LEDs at less than their ratings. It doesn't reduce the brightness much, but greatly increases the efficiency and lifetime.

[T3sl4co1l]

I don't mind running at ratings, personally; as long as they're good parts, that's what really matters.  That, and thermal management.  Which, it sounds like they're the pricey kind, so they ought to perform well (high lm/W).

Last pricey ones I got were, let's see... Cree XTEAW-00-0000-00000HGE3.  Far smaller than the old gull-wing-cylinder style they replaced, but their efficiency is so much greater, they still run cooler.  It's... a little insane.  And quite wonderful.

So, if it's already designed to run an acceptable temperature with older chips, or, like, resistors (i.e., assuming 100% applied power goes to heat -- in reality it's more like 50% now!), they'll run quite cool.  If it's new design, and you're accounting for dissipated power only, or don't have much opportunity to dissipate heat in the first place (compact, enclosed?), yeah, derating may be desirable.

Tim

[Zero999]

I don't mind running at ratings, personally; as long as they're good parts, that's what really matters.  That, and thermal management.  Which, it sounds like they're the pricey kind, so they ought to perform well (high lm/W).

Last pricey ones I got were, let's see... Cree XTEAW-00-0000-00000HGE3.  Far smaller than the old gull-wing-cylinder style they replaced, but their efficiency is so much greater, they still run cooler.  It's... a little insane.  And quite wonderful.

So, if it's already designed to run an acceptable temperature with older chips, or, like, resistors (i.e., assuming 100% applied power goes to heat -- in reality it's more like 50% now!), they'll run quite cool.  If it's new design, and you're accounting for dissipated power only, or don't have much opportunity to dissipate heat in the first place (compact, enclosed?), yeah, derating may be desirable.

Tim

I'm just going from my experience of commercial LED bulbs, which generally run the LEDs at their ratings and invariably don't last as long as they should. Then there's the Dubai lamp which is more efficient and lasts for much longer too.

If budget isn't an issue, the LT3080 can be used as a current sink with a low drop-out voltage.

[nctnico]

I don't mind running at ratings, personally; as long as they're good parts, that's what really matters.  That, and thermal management.

Thermal management is easy to underestimate. Even 60mW LEDs need like 4x4mm pads in order to have sufficient heatsinking in real life applications. Otherwise they will (not can, but will!) fail. A LED is a semiconductor so their operational temperature is rather low compared to passives like a resistor. It is quite similar to power transistors. Once you've got them mounted onto a heatsink and apply derating for elevated environment temperatures, you often realise you can only dissipate 25% of what the transistor is rated for.

[Benta]

You have many LEDs in series, so the variation in forward voltage drops will cancel each other out. You won't have six 2.75V LEDs, or 3.2V LEDs, but an even distribution of forward voltages, giving a total forward voltage of about 3V.

I have an engineering rule that I always stick to: never do a basic design based on probabilities, typical values or wishful thinking.
You can always downgrade afterwards based on experience.

And thinking that a reel of parts exhibits a nice Gaussian distribution is wishful thinking. You never know if a really big customer has asked the manufacturer to select parts with a certain V_F. This happens much more often than you think (don't ask me how I know, I've been in the semiconductor industry for 40+ years). The rest is sold on the open market, but the Gaussian distribution is gone, although all parts confirm to specs.

[Benta]

I don't mind running at ratings, personally; as long as they're good parts, that's what really matters.  That, and thermal management.

Thermal management is easy to underestimate. Even 60mW LEDs need like 4x4mm pads in order to have sufficient heatsinking in real life applications.

Very true, but in this case quite easy: total P_D is 19 x 0.35 A = 6.65 W. Not insurmountable. The idea is to mount the lights/LEDs/regulator on an aluminium baseplate anyway. Shape TBD.