Looking for advice for an LED flag project

[Inglorious89]

Hey y'all, I'm currently making an LED American flag (the Betty Ross version with 13 stars) and I just got done with a bunch of testing and number crunching and have some current values I'd like to get checked out before I go ahead with buying the LEDs I need as they're going to run me a few hundred.

Here's a photo I made with the design and all of the current values I found and calcuated: http://i.imgur.com/LPHwy1f.png

I'm planning on using this power supply: https://www.amazon.com/Dunlop-ECB00...d-3035-4a40-b691-0eefb1a18396&pf_rd_i=desktop

And here are the LEDs I'm planning on using:
Blue
https://www.digikey.com/product-det...-BGF-CS0U0352/C4SMC-BGF-CS0U0352CT-ND/6138519
Red
https://www.digikey.com/product-det...-RJF-CT0W0BB2/C5SMF-RJF-CT0W0BB2CT-ND/6138570
White
https://www.digikey.com/product-detail/en/cree-inc/C513A-WSS-CY0Z0231/C513A-WSS-CY0Z0231-ND/5120069

My basic question is this: are the numbers I got for the amount of current each group needs and the total current correct, and with the design I have will the LEDs light up and be about the same brightness? Any other suggestions or advice would also be nice.

Thanks!

 

[Chemelec]

18 Volts, But I Don't see Where it gives a CURRENT RATING?

 

[Inglorious89]

18 Volts, But I Don't see Where it gives a CURRENT RATING?

The power supply I linked? If that's what you're referring to then the page lists 18V 1A.

 

[Audioguru]

LEDs are not light bulbs with a filament. They are a semiconductor and have a range of forward voltage.
For example the "3.1V" blue LEDs might have some with 2.8V and others with 3.5V. Or they all might be 2.8V or 3.5V. Then you must reduce how many LEDs are in series and increase the resistors values.
Or you can buy millions of LEDs and test them all. You might find some that are 3.1V.

The "18V" cheap Chinese power supply also has a range of voltage. Some might be 16V and others might be 20V.
You must calculate for the ranges.

.

 

[Chemelec]

Your Calculations don't appear Correct.

1) It appears those LED are all rated at 20 mA.
So you need to determine How Many of each, you can put in a Series String so they add up to Somewhat Less than your 18 Volts.
Note Different colors also have Different Voltages.

2) Than you need to Add a Current Limit Resistor to Each String, so Each String Draws about 20 mA.

3) Than you can add up all the Strings to Determine Total Current Required.
.

 

[Inglorious89]

LEDs are not light bulbs with a filament. They are a semiconductor and have a range of forward voltage.
For example the "3.1V" blue LEDs might have some with 2.8V and others with 3.5V. Or they all might be 2.8V or 3.5V. Then you must reduce how many LEDs are in series and increase the resistors values.
Or you can buy millions of LEDs and test them all. You might find some that are 3.1V.

The "18V" cheap Chinese power supply also has a range of voltage. Some might be 16V and others might be 20V.
You must calculate for the ranges.

.

So the goal of reducing the amount of LEDs per string is to reduce the chance that the LEDs will have wildly varying brightness levels due to the different Vf values, correct? If so, then how many LEDs would you recommend (should I cut my current values in half, for example)?

 

[Inglorious89]

Your Calculations don't appear Correct.

1) It appears those LED are all rated at 20 mA.
So you need to determine How Many of each, you can put in a Series String so they add up to Somewhat Less than your 18 Volts.
Note Different colors also have Different Voltages.

2) Than you need to Add a Current Limit Resistor to Each String, so Each String Draws about 20 mA.

3) Than you can add up all the Strings to Determine Total Current Required.
.

Is 20mA the recommended current or just the minimum? My logic for having such varied resistor ratings is to make sure each LED string has the same brightness and to make sure the brightness isn't blinding. And how much less for the power suppy, 1-2 less?

 

[KJ6EAD]

The LEDs you've selected are potted in tinted transparent epoxy. Undiffused LEDs are not pleasing to view directly at close range.

The red LED has a peak wavelength of 621 nm. That's an almost orange "red". 660 nm is a nice cherry red.

Your stars represented with two LEDs and spaced unevenly will not look good. Breaking free from a fixed grid may help as may using a single, larger LED. There are plastic lenses and diffusers available that may give a better appearance.

It is very unlikely that all three LEDs will have an equal apparent luminance when driven at equal currents. If you want them as bright as possible, but equal, you'll have to reduce the current for two colors to match the dimmest. In a panel indicator design I did recently, it was necessary to add a 10k resistor to one and 2k to another to match the intensities!

Consider purchasing small quantities of candidate LEDs for proof of concept before committing hundreds of dollars.

 

[Inglorious89]

The LEDs you've selected are potted in tinted transparent epoxy. Undiffused LEDs are not pleasing to view directly at close range.

The red LED has a peak wavelength of 621 nm. That's an almost orange "red". 660 nm is a nice cherry red.

Your stars represented with two LEDs and spaced unevenly will not look good. Breaking free from a fixed grid may help as may using a single, larger LED. There are plastic lenses and diffusers available that may give a better appearance.

It is very unlikely that all three LEDs will have an equal apparent luminance when driven at equal currents. If you want them as bright as possible, but equal, you'll have to reduce the current for two colors to match the dimmest. In a panel indicator design I did recently, it was necessary to add a 10k resistor to one and 2k to another to match the intensities!

Consider purchasing small quantities of candidate LEDs for proof of concept before committing hundreds of dollars.

I probably should've mentioned this in the first post, but I actually did buy 10 of each of the LEDs I linked in the first post and all of my numbers I got from testing with them. Reason why my resistor values are so varied is because those are the values where the LEDs appear to be the same brightness while also not being blindingly bright.

And as for the star LEDs, I'm thinking designing the star are around the stars themselves (i.e. place the stars first then fill the rest of the grid with an appropriate amount of blue LEDs) would fit best with your suggestion.

But I am curious: would my design work on purely an electrical level (as in every LED would light)? Would it just look average to bad or is there some fundamental flaw that I'm not understanding that would cause a section of the LEDs to go out?

 

[Chemelec]

QUOTE: Would it just look average to bad or is there some fundamental flaw that I'm not understanding that would cause a section of the LEDs to go out?

Possibly you supplied too much current.

 

[KJ6EAD]

I…I actually did buy 10 of each of the LEDs I linked in the first post and all of my numbers I got from testing with them…

But I am curious: would my design work on purely an electrical level (as in every LED would light)? Would it just look average to bad or is there some fundamental flaw that I'm not understanding that would cause a section of the LEDs to go out?

I'm pleased to know that you did some testing and brightness tuning but in your effort to have as many LEDs as possible per string, you've gone too far. For good immunity to voltage variation, the resistor portion of each string should comprise approximately 15-20℅ of the voltage drop. For 18V that leaves 14.4V for LEDs. Taking your blue LEDs as an example and even accounting for the V-I curve in the datasheet which puts the drop per LED in the 2.6-2.7V range for your low current, you miss the mark by a couple of Volts.

As you can see, this raises the question "How well-regulated is the power supply?" If it's a tightly regulated switch-mode device, you can reduce the resistor margin to 10℅ and safely get more LEDs per string. I usually choose a 24V supply for this type of project. It's the highest readily available voltage that stays below the 30V hazard level. With a 24V switch-mode supply, you can run 8 blue LEDs per string with 10℅ voltage in the resistor.

There is a resource section associated with this site dedicated to LEDs that you may want to peruse if you haven't already.

https://www.electronicspoint.com/resources/got-a-question-about-driving-leds.5/

 

[Audioguru]

The spec's for the Cree LEDs you selected do not say their minimum voltage. If you design for 20mA with "3.2V" LEDs and do not have enough voltage for the resistor then if the LEDs are actually only 2.8V then they will soon burn out.
If the LEDs are actually 3.8V then they will not light.
You must design for the range of voltage from minimum to maximum.
Here are the datasheet spec's for another good LED:

 

[Audioguru]

Hee, hee. Post a video of your LED flag on fire.
I made a night light with 36 powerful LEDs a guy sent to me. Boy oh boy did it get HOT! Of course the LEDs were running at about 75% of their maximum allowed current (VERY bright at night) and the case had many holes for cooling but there were too many LEDs in the small space.

 

[Inglorious89]

I'm pleased to know that you did some testing and brightness tuning but in your effort to have as many LEDs as possible per string, you've gone too far. For good immunity to voltage variation, the resistor portion of each string should comprise approximately 15-20℅ of the voltage drop. For 18V that leaves 14.4V for LEDs. Taking your blue LEDs as an example and even accounting for the V-I curve in the datasheet which puts the drop per LED in the 2.6-2.7V range for your low current, you miss the mark by a couple of Volts.

As you can see, this raises the question "How well-regulated is the power supply?" If it's a tightly regulated switch-mode device, you can reduce the resistor margin to 10℅ and safely get more LEDs per string. I usually choose a 24V supply for this type of project. It's the highest readily available voltage that stays below the 30V hazard level. With a 24V switch-mode supply, you can run 8 blue LEDs per string with 10℅ voltage in the resistor.

So if I'm understanding correctly, at least 10-20% (only 10% if the supply is of a certain type) of the power supply voltage needs to be across the resistor, correct? And this is to account for varying Vf values, which ties into what Audioguru has been telling me.

With the 24V power supply, if 2.4V of that supply is across the resistor, that leaves enough voltage (21.6V) for about 8 LEDs if they're all getting ~2.65V. Would I have to change my resistors with this design or is it OK for me to leave them as is? And I'm assuming I can repeat the same design for the white and red LEDs, correct?

 

[Audioguru]

You did not calculate how much voltage is needed if the blue and white LEDs voltage is the 3.8V maximum. Then the LEDs need 8 x 3.8V= 30.4V plus maybe 3 or 4V for the resistor. You must calculate both minimum and maximum or buy thousands of LEDs and test them all and hope you get some the same.

 

[Chemelec]

Chances are if each color comes from One Batch, they are usually fairly Close to the same specs.
I Don't mean: Red, Whites and Blues are the Same Batch.

BUT, All Reds from one Production Batch.
All Whites from Another Production Batch.
All Blues from Another Production Batch.

And If you buy a certain color at Different Times, It is Not Likely to be the Same Batch.

 

[Inglorious89]

You did not calculate how much voltage is needed if the blue and white LEDs voltage is the 3.8V maximum. Then the LEDs need 8 x 3.8V= 30.4V plus maybe 3 or 4V for the resistor. You must calculate both minimum and maximum or buy thousands of LEDs and test them all and hope you get some the same.

So, as an example, if I use a 24V power supply then 3.5V of that needs to be across the resistor, and in that case I have about 20.5V to work with. If I'm preparing for a worse-case scenario where most if not all of the LEDs in the string have a Vf of 3.8V, then I can only have 5 to 6 LEDs in that string.

 

[KJ6EAD]

So if I'm understanding correctly, at least 10-20% (only 10% if the supply is of a certain type) of the power supply voltage needs to be across the resistor, correct? And this is to account for varying Vf values, which ties into what Audioguru has been telling me.

With the 24V power supply, if 2.4V of that supply is across the resistor, that leaves enough voltage (21.6V) for about 8 LEDs if they're all getting ~2.65V. Would I have to change my resistors with this design or is it OK for me to leave them as is? And I'm assuming I can repeat the same design for the white and red LEDs, correct?

Correct on both counts. You will need to calculate new resistor values but the task should be trivial since you have accurate, experimentally derived current and Vf values. Audioguru is painting a bleak picture of extremes and has not accounted for the lower than typical currents you're running. I looked at the V-I curve for your blue LEDs and extrapolated the Vf as 2.6-2.7V since the given graph cuts off at 5mA. Chemelec is also correct that you're unlikely to see wide variation within a batch.

If you want to entertain the extremes with some math, start with the specified variance of the power supply, take the upper and lower values you get from this. Apply these supply voltages to the upper and lower limits of Vf you can expect at your current then include the resistance and you'll quickly discover how the 10-15℅ resistance voltage drop protects a string of LEDs from their manufacturing variances and power supply variance.

 

[Inglorious89]

Correct on both counts. You will need to calculate new resistor values but the task should be trivial since you have accurate, experimentally derived current and Vf values. Audioguru is painting a bleak picture of extremes and has not accounted for the lower than typical currents you're running. I looked at the V-I curve for your blue LEDs and extrapolated the Vf as 2.6-2.7V since the given graph cuts off at 5mA. Chemelec is also correct that you're unlikely to see wide variation within a batch.

If you want to entertain the extremes with some math, start with the specified variance of the power supply, take the upper and lower values you get from this. Apply these supply voltages to the upper and lower limits of Vf you can expect at your current then include the resistance and you'll quickly discover how the 10-15℅ resistance voltage drop protects a string of LEDs from their manufacturing variances and power supply variance.

Alright, I think I'm understanding a lot of what I'm being told here. I read the LED guide you posted earlier and it seems to be confirming all of my thoughts, but it did raise one question: how do I determine what current value I want if I'm connecting a bunch of LEDs in series (and this is referring directly to section 2 of that guide)? If the current determines the brightness, how can I know how much current will result in how much brightness (without testing the whole circuit)?

And as if right now I *think* I have a good idea of what I need to do to get my flag going, but I was going to run over the numbers with a few co-workers tomorrow to see what they say. If everything seems to be groovy, then thanks for the help!

 

[(*steve*)]

The best solution is to have a constant current source powering each chain. But even this doesn't guarantee the same brightness.

Given that you're buying these from a reliable source, you're likely to get genuine parts presumably from a single batch. Thus you're less likely to get large differences in brightness or Vf within the items you buy.

If there are any that are particularly bright or dim, you may be able to swap then around to make it look better or swap them with some spares.

I would probably go for undiffued LEDs and mount then behind a diffusing panel. Third would even out brightness differences considerably and allow you to use far cheaper LEDs. For the stars, I would probably mount them through the diffusion panel so they appear as bright points of light. If you do this, you may want a different LED current for these.