DIY multi LED grow light

bob0816 · Oct 1, 2020

printer said:
No, there is a voltage drop in the diode that needs to be overcome, then the diode starts conducting. The above graph using the Red curve is roughly 1.5V. The 1.5V has to be supplied before the current starts to flow, after that the additional voltage across the diode's internal resistance determines the current.

Aahh!
Short recap:
- V_f needs to be supplied so I can put current through the diode, that voltage does not produce light emission.
- to put a desired current through then, I take the internal resistance, calculated as the slope from the linear part of the diode-specific I_f/V_f graph, R_diode=V_f/I_f (R_diode is a fix material constant)
- the voltage needed for that/related to that current is calculated by rearranging the equation to V=R_diode*I, the overall V needed on that line is the sum of forward voltage to overcome diodes internal resistance V_f and V related to desired current
- wattage is I²*R (square of throughput current times internal resistance)
- since voltage is

3.1 V, 3.0 V, 3.05 V, 2.9 V, As I showed with the graph, there is a diode drop and a resistance drop in a Led (actually any diode). Just to make things simple let us assume the internal resistance of each of these Leds are the same and it is just the diode drop that is different. And let's say that once the diode drop voltage is supplied it only takes 0.1 V to get 20 mA through the Led. So a quick approximation, say the first diode needs 3.0 V to cover the diode voltage drop. That means 0.1V / 20 mA = 5 ohm. The math for the others is close enough, we will call all of them 5 ohms.

And say we turn up the power supply until it supplies 3.1 V. That means the first diode is passing 20 mA. The second Led has a diode drop of (3.0 - 0.1 V = 2.9 V) so it has 3.1 V - 2.9 V = 0.2 V across its 5 ohm resistance. 0.2 V / 5 ohms = 40 mA.

The third one, 3.1 V - 2.95 V = 0.15 V. 0.15 V / 5 ohms = 30 mA

The fourth, 3.1 V - 2.8 V = 0.3 V. 0.3 V / 5 ohms = 60 mA.

The power supply would be supplying 20 + 40 +30 + 60 = 150 mA.

So it is not good to put parallel Leds across a constant voltage power supply without balancing resistors inline with them. These resistors are much bigger than the internal resistance and it balances out the currents. It does come at a cost of added voltage and power wasted in them. But that is life.

Now what happens if you have a constant current power supply and you have it set for 80 mA? I will not do the math, just to say the 2.9 V and 3.0 V Led will get most of the current and the 3.05 V 3.1V Leds will get a fraction of the current and will barely turn on if at all.

very helpful, thanks a lot!

The different currents in a parallel wired CV circuit as calculated above (20 + 40 +30 + 60 = 150 mA) would be no problem, if the LEDs' optima would lie exactly or in tolerable range of these currents?

and http://lednique.com is a good source too!

bob0816 · Oct 1, 2020

Could an admin please exchange this part in post #18:

"λ, part number, V_f, I_f, I_opt, I_max
451 nm, 350 mA I_opt
WW, 700 mA I_opt
WW, 350 mA I_opt
640 nm, 350 mA I_opt
660nm, 350 mA I_opt
730nm, 350 mA I_opt
(will be edited, all work went down the drain when I hit some key by accident

)"

with the following and then delete this post, for a better overview:

former part choice:

λ ............. part number ............................... V_f ....... I_f ........... I_opt ....... I_max
451 nm . GD CSSPM1.14-UOVJ-W4-1...... 3.25 V . 100 mA .. 350 mA ... 1 A
WW ....... LUW CQAR NQNS-MCML-1....... 3.5 V .... 200 mA .. 700 mA .. 1.5 A
WW ....... GW CS8PM1.EM-LQLS-A333-1. 3.2 V .... 100 mA .. 350 mA .. 1 A
640 nm . GH CSSPM1.24-4T2U-1-0 .......... 2.6V .... 100 mA .. 350 mA .. 1 A
660 nm . GH CS8PM1.24-4T2U-1-0 .......... 2.6 V ... 100 mA .. 350 mA .. 1 A
730 nm . GF CSHPM2.24-1T3T-1-0 ........... 2.3 V ... 100 mA .. 350 mA .. 1 A
WW: warm white

with current knowledge I see, that the 700 mA WW (warm white), which I selected because of their higher PPF (photoactive photon flux) value of 228 lm in higher piece number over the 110 lm 350 mA, would have lead to wiring the 350 mA diodes in parallel of 2 in series with the 700 mA diodes, but they would have given out almost the same amount of light/ same flux value like the 350 mA diodes, when run on half optimum.

printer · Oct 1, 2020

bob0816 said:
Aahh!
Short recap:
- V_f needs to be supplied so I can put current through the diode, that voltage does not produce light emission.
- to put a desired current through then, I take the internal resistance, calculated as the slope from the linear part of the diode-specific I_f/V_f graph, R_diode=V_f/I_f (R_diode is a fix material constant)
- the voltage needed for that/related to that current is calculated by rearranging the equation to V=R_diode*I, the overall V needed on that line is the sum of forward voltage to overcome diodes internal resistance V_f and V related to desired current
- wattage is I²*R (square of throughput current times internal resistance)
- since voltage is

very helpful, thanks a lot!

The different currents in a parallel wired CV circuit as calculated above (20 + 40 +30 + 60 = 150 mA) would be no problem, if the LEDs' optima would lie exactly or in tolerable range of these currents?

and http://lednique.com is a good source too!

Glad it helped. Would it be acceptable in parallel? I don't think so without balancing resistors. You would have some leads putting out a lot of light and heating up while others don't do much. Just as an added bonus, the lower current through the leds the greater the efficiency, not a big jump but it is there. Also keeping the leds cool. But back to parallel leds. If I were to want to do it that way I think that since you have a lot of leds at hand you could set up a little test rig and sort them as to how much voltage drop per amount of current. Then put the ones with the same voltage in parallel.

Whether it is needed to do depends on how much spread you actually find you have. As an example I looked through resistor assortments years back. the resistors were 5% tolerance. But for the same batches of resistors they were roughly 1-2% around a mid value that fell within the 5% band. The takeaway is that a certain lot of leds made at the same time will have closer tolerances than a month apart.

As far as designing a board, I would look at the datasheet for the leds, stick a bunch in series parallel arrangement, have a power supply that limits in voltage and current, also with excess capacity in both. But I am not interested in making the best board I can. I just used a bunch of light bulbs and got on with my life. To each his own.

bob0816 · Oct 1, 2020

How to best test SMD LEDs? Could I put a test footprint on the board and lay them on top loosely or would the measurements differ when soldered?

printer · Oct 1, 2020

bob0816 said:
How to best test SMD LEDs? Could I put a test footprint on the board and lay them on top loosely or would the measurements differ when soldered?

That would work fine. Maybe a little pressure on the pads with the eraser part of a pencil.

bob0816 · Oct 1, 2020

Fantastic!
Or what function does a multimeter need to have to test them, is "diode test" just giving some electricity or is it also giving voltage / current values?

Two illustrative simple videos about current and voltage, and resistors

This video of a series about LED drivers shows the unsepperable connection between voltage and current the advantages and disadvantages of a CC+CV driver (MeanWell HLG-120H-54A /B)

bob0816 · Oct 1, 2020

Here is what started my interest in DIY LED grow lights from PCBs:
PC-case grow

looking forward to also pick up on the automation ideas in hydroponics and light cycles.
He has open sourced his .stl for a 3D printable design for self made carbon filter cases for PC-case grows.

PCB LED light repair / dirty reflow soldering

bob0816 · Oct 8, 2020

I have a first schematic scribble and some more questions:
The diodes in 5x3 segments are all 65 mA V_f, the ones in series are all 350 mA I_opt, but different V_f. GND1 and GND2 are either connected to a rocker switch that is connected fix to the drivers cathode or they will be fix poke-in ports where the cathode wire will be swapped.
- Can the resistors be placed 1/5row-segment, or does each row in a segment need it's own resistor?
- Can I leave out resistors for semgments equal in voltage drop (if V_drop is highest in segments compared) and adjust only the other segments to this drop?
Right now I'm confused how to fix the two-mode problem in a series circuit, this one is NOT finished as I didn't found a way to drive the first segment (on the left) with and without the marked LEDs AND the three other segments, all four in parallel.

And I'm unsure which configuration is supirior (saw both in different commercial boards):

I think the best would be the one on the right with a resistor to each line, but due to place and simplicity I tend to using only one per segment of five rows in parallel.
The configuration on the left would cause the electricity to run it's way (in greater proportion) over the parts with least resistance (due to manifacturing fluctuation), right?

Here my recent approach (NOT working)

bob0816 · Oct 12, 2020

I don't know whether the GND and VCC are selected right, they shall just symbolize positive and negative poles. Also, the last sentence from the post before is superfluous.

For now I have a series model calculated that satisfies almost all requirements that I applied. I considered the following in my design of the light (LEDs, PCB, driver):

- Cost efficiency [lm/€] [W/€], spare parts/excess value (min. PCB piece No.: 5)
- Energy efficiency [lm/W]
- Heat management / LED density
- Assembly/design (series/parallel): as simple as possible after all other requirements
-- driver choice
- Expandability
- Modulation capability: far red ON/OFF

I will post the finished design,design files and stats as soon as they are ready, but for now I need to put it on hold for a while. The light will be build hopefully until March.
It looks like it will be roughly 181 € for a 64 W light, applicable to 60x60 cm (2x2 ft), used as 48 W light in a 40x20 cm grow space.

bob0816 · Dec 3, 2020

FINAL VERSION

Schematic and 3D model of one 400x60 mm / 15.75x2.36 " strip:

Stats:
..................... 3 Strips parallel ... 5 strips parallel
If_opt .................. 58 W ....................... 97 W
If_1.5*opt .......... 88 W .................... 146 W (120 W max in my setup)
If_2*opt ........... 117 W ..................... 194 W (120 W max in my setup)
If_max .............. 269 W ..................... 448 W (not encouraged; shorter life time, needs active cooling, possible still)
(If_opt = 100% I from LED data sheets; 65 mA for LM301sH, 350 mA for LH351Hs; Vf=29.9 V;
For one strip: If_opt=5*65 mA=325 mA~350 mA -> If_1.5*opt=488 mA, If_2*opt=650 mA, If_max=1 A=300%)

Parts: (for 5 strips)
- 5x 400x60 mm PCB ($ 49 + $ 10 solder stencil + $ 17 shipping in my case: PCBway)
- HLG-120H-30A, Mean Well Driver
............. V Adj.: 27 - 33 V
............ Io Adj.: . 2 - 4 A
...... CC region: 15 - 30 V
- 300x LM301H, CRI 80, 3500 K, 38lm ~ 40lm @ If_opt and 25 °C (+10 replacements)
- 100x LM301H, CRI 80, 4000 K, 38lm ~ 40lm @ If_opt and 25 °C (+10 replacements)
- 26x LH351H Deep Red V2, 660 nm (+1 replacements)
- 10x LH351H Red, 643 nm (+1 replacements)
- 4x LH351H Far Red, 730 nm (+1 replacements)
(Wavelenght diagrams in the PDF attached)
- 15 Connectors
- 5 Jumpers
- cable
_______________________
$ 269.82 // 224,49 € (incl. replacement LEDs)

+ Conformal Coating (~ $ 18 // 15 €)
+ Aluminium Rails
+ Screws

__________________________
~ $ 287 // 240 € / 120 W (≤ 448 W [If_max] with different driver)

If anyone should be interested I will provide you with the gerber files for I can't upload them here, as well as the parts list as .xlsx spreadsheet for digikey cart upload. I will look for a way after I checked on them with the manufacturer or hold the PCBs in hand

bob0816 · Dec 4, 2020

Thoughts and Advice + datasheets

The indicated If_max values correspond to the maximum If_values given by the manufacturer. It is therefor possible to run the system at these currents. The higher A values dictate higher Vf values and therefore lead to higher wattage consumed. At the same time lowering the efficiency by going down in relative luminous flux due to gradual efficiency loss above 25 °C (77 F), putting out more light with a higher cost. Samsung says in the LH351H datasheet: "Unpredictable performance may be resulted by driving the product at above Max. If. But there will be no damage to the product."
If you have proper, active cooling and a look on the actual temperatures all over the boards, I see no problem in driving them at If_max if you restrict V and U (I) via driver, but I would highly recommend to drive the system on a lower current to be on the safe side, considering malfunction or destruction of one diode can easily result in blowout of first one and then multiple to all diodes when the given current is subsequently devided onto less and less diodes.
Conformal coating is highly needed to prelong durability.

bob0816 · Dec 4, 2020

For maximum juice, I'd suggest

HLG-480H-42A Mean Well Driver
............. V Adj.: 35.7 ~ 44.1 V
............ Io Adj.: . 5.7 ~ 11.4 A
...... CC region: 31 - 42 V

$ 132.28 / 111,68 €

7CardBud · Dec 4, 2020

I hate to be the bearer if bad news, but that design has numerous errors. Does your design software allow you to do a virtual run of the board to detect them?

bob0816 · Dec 5, 2020

Well, hit me right away, where do you see them? I'm here to learn

I'm using KiCad, but there was none showing up. I'm at amateur lvl, don't know my way around representing electrical properties in KiCad. For example how to let electricity flow through an LED footprint (contacts are seperate).
I see the problem with J1-driven (Far Red-) mode now, I wanted it to drive D22 but I'm bypassing it via D4

7CardBud · Dec 5, 2020

I'll shoot right from the hip, that first design is ready for the recycle bin. ;-)

It's never a full loss. Your always learning as you go and with repetition it will become much easier.

You want to design your circuit first. From the looks it was 20 Whites, 1 Far Red and 2 Deep Red.
Looks like you knew the parts you wanted to use and then kinda winged the circuit as you went.

Next you want to get an idea of your board layout for the parts and then how to power them.
Instead of lots of long traces it's better to lay out power and ground planes that can power your circuit.
You can then tap into them anywhere you need. They also run cleaner, long curvy traces are more prone to noise from high frequency
switching power supplies, EMI...etc.

Once you get the basics down you can make any size board you like. Long boards with series or parallel circuits or maybe square boards with series parallel circuits.

I wish this site had a marketplace cause something like this would make an awesome group buy. The price of making pcbs drops in price per unit
pretty fast as the numbers go up.

Like 10 pcbs $100
100 pcbs $300
1000 pcbs $800

I dunno if people making deals through unsolicited PM's is frowned upon here. I would kick in some Paypal for a dozen or so pcbs.

A generalized layout using power and ground planes.

bob0816 · Dec 5, 2020

Thanks man, really appreciate the time and effort.

7CardBud said:
Looks like you knew the parts you wanted to use and then kinda winged the circuit as you went.

Exactly how it went

had the parts and the grow space dimensions and then went back and forth adjusting to a 5 pcs -PCBs light due to minimum order at the PCB manufacturer, fizzling with different constellations (lowest PCB pc dimension costs, series/parallel = driver, two channel, most LEDs possible).

Saw that approach already at the Ivisible Sun R strips but couldn't make much of it at that time, was thinking about resistors (->Invisible Sun R strips) when now I think it's just bridges. Read about long loopy traces and frequencies but it didn't sink deep enough, now it did.
Putting power and ground along the long sides didn't occured to me whatsoever, thanks for the tip and the picture!

I once already thought it would be impossible to run the system two-channeld, but somehow discarded that insight..
I think I got it now, will try to rewire it with as little changes to the positions of the LEDs, power on one long side and
to ground lines, bridged with the jumper on the opposite.

Thanks for the chart, wondered how much the traces could take, thought other designers worked with the same parts(=specs)
on PCB, too, so nothing horrible will happen, go as big as possible (was > 600 mA under max current) . Minimum trace width was
related to the connected-to pad then (diameter of pad=diameter of trace), no trace was smaller than the majority of traces.

Read about Delta Electronics drivers in another thread, do you have some thoughts on them?
I just learned about arrow, can you compare it to digikey?

bob0816 · Dec 6, 2020

7CardBud said:
Looks like you knew the parts you wanted to use and then kinda winged the circuit as you went.

The number and positions of the Far Reds and their two-mode function dictate a lot of the wiring setup. I have the new wiring on paper, will do the new PCB layout soon.

bob0816 · Dec 18, 2020

Version 1.1

(the GND pinns are connecting the heat pads from the 3535 LEDs to a copper plane)

(added boxes to mark the strip's position in the grow light (Far Reds on 2nd and 4th))

I will redesign the in-block (3030) connections and copper fills for more efficient heat dissipation and try same with the 3535 (more difficult due to three pads, two connections, one heat dissipation pad). Also I will try to find a way to wire the two 3535 in the middle in a way that gives a plane space for a logo/is more symmetric. I hope to be finished and able to send the draft to the manufacturer this weekend.

cobshopgrow · Dec 18, 2020

great to see one fiddling on a own PCB.
what PCB material you plan to use? if FR4 i would reconsider the heat management, using the backside, by vias.
do all your 3535 leds have the heatsinkpad seperated from the anode? that will be the case most of the time, but there are 3535 leds out where its a problem having the sink and gnd connected.

bob0816 · Dec 18, 2020

Nice you can find joy in the project.
It will be Aluminium, Metal Core PCB (MCPCB).
All 3535 are Samsung LH351H, the pad designs are in the data sheets in post #31, (lh351h_red_rev1-2_8-20-18-pdf, page 8 e.g.).
I don't know whether we mean the same with "ground", I still have to figure out a lot about electronics, especially the terms and when to use them correctly so everybody feal free to correct me when I'm wrong. In my setup the heat pads are connected to a plain copper plane to spread the heat wider, I just marked them GND in KiCad so I can hook them up to the copper fill (in), it's not connected to the negative pole/anode.

DIY multi LED grow light

Member

Member

Well-Known Member

Member

Well-Known Member

Member

Member

Member

Attachments

Member

Member

Attachments

Member

Attachments

Member

Well-Known Member

Member

Well-Known Member

Member

Member

Member

Well-Known Member

Member