Powering a LED strip reliably

[MrFollies]

I'm putting together a LED sign board for display on the front dash of my car for school pickup.    You usually put a sign up there with the family name so you can swing by at pickup time and a teacher calls the family name.   It helps with traffic jams and not having parents get out of the car just makes things safer around young kids school zones at pickup.

I'm trying to get my son interested (more) in electronics, so I want to build a little Pi Pico or esp32 controlled digital sign board using a 8x32 (perhaps 2) WS2812B LED panels.
Powering it on with a current meter sees the board consume 2.4amps at white and the voltage drop to 3.4v (coming from a USB charger rated at 2.5amps).

Doing the math at 3 * 20ma * the number of LEDs tells me it could consume 15 amps at full glow.   Hmmm...   Maybe my math is wrong here, that seems a lot!

Since this is going to be in my car and plugged into the cigarette lighter, rated at 12v/10a I'll have enough power (especially since real world usage will be far less than the full glow as it's for text display).

So I'd like to design a small board to do the following.

• Take 12v input and regulate that down to 5v but be able to draw plenty of current for LEDs.
• Have some buffering in place so the board itself can handle draw fluctuations.
• Also have a smaller 5v reg to power the Pico/Esp board
• Not cause a fire in my car.

So number 3 is easy as I have plenty of 5v regs from pololu or ali express.
And for 1 I was thinking of just putting in a few larger electrolytic caps to take up the slack.    My hope is that by doing this, it means even if I display all white for a few seconds, it will load up the caps, rather than the input / reg circuit.   But I am a pretty new newb with this stuff so I have no idea if A) it would work like that, B) how to size the caps and C) if caps are the right choice or perhaps a super cap?

And then there is the question of a regulator.     Here I am really not sure.   Is my power figure inaccurate?    The strip seemed pretty bright at 3.4v . 2.4a,....

[Jwillis]

LEDs are current driven. 20mA is the maximum for standard LEDs, but they will operate at 10mA with no noticeable difference in brightness. They will also last much longer at a lower current. Different colours of LED have different forward voltages ranging from under 1.8V for red and as much as 5V for white. http://lednique.com/leds-and-colour/
So depending on the colour of the LED and the operating voltage, a current limiting resistor is required. https://www.digikey.ca/en/resources/conversion-calculators/conversion-calculator-led-series-resistor

As for those Matrix panels, I can't find much information on the amount of current they are supposed to use at 5V.

[Nominal Animal]

The typical max7219-clone driven single-color 8x8 LED modules you see at Banggood/eBay/Aliexpress, actually only light one 8-LED row at a time.  The current through each LED is controlled by a single 9.5k-47k resistor connected to MAX7219 ISET pin (pin 18, next to positive input pin), setting maximum peak current per LED to 100*VCC/R, with maximum limited by the driver IC itself to 40mA or less.

Thus, these will have a varying current draw even when displaying static LEDs, at up to 320mA per module, assuming the module does not try to overdrive the LEDs too much.  For the ubiquitous 4-in-one clone module, 32×8 matrix, powered from 5V, that corresponds to 1.5A or so (1.3A + losses) draw at 5V with all LEDs lit.  You'll want to use a switchmode supply that can provide at least 1.5A at 5V, preferably more; and preferably an automotive one considering the spiky 12V supply in a car.

LCSC and JLCPCB (for assembly at JLCPCB) currently sell the Hgsemi MAX7219 IC (non-English datasheet) for under 1.4€ apiece, so you could design your own board for your preferred 8×8 common cathode LED modules.  You only need the current-setting resistor and a couple of bypass capacitors per 8×8 module.  It is often easier to align the LED modules inside a frame, than use the PCBs to align the LED modules.

Putting a tinted/colored plastic sheet of the same color as the LEDs over the display will make it easier to read.  You can also consider getting these in the kit form for easier control of the Rset resistor (10k - 47k, lower resistance yields higher per-LED current).

For WS2812B, each LED includes their own current control, maxing out at about 60mA per LED, and generating 90mW of waste heat when run from 5V.  Indeed, lighting 256 such LEDs (a full 32×8 matrix) would consume over 15A.  You can reduce the waste heat by running them from a lower voltage, about 3.8V to 4.2V, reducing the waste heat generated.  The current draw is unaffected, but the total power draw is reduced.  I recommend using 74LVC1T45, 74LVC2T45, 74LVC8T245, 74LVC16T245, or TXU0104 for the level shifters (from 3.3V logic levels to 3.8V-5V you use for the LEDs).
Some kind of diffuser is often used that passes 80%-90% of the light through, but scatters it.

JLCPCB currently sells Xinglight XL-5050RGBC-WS2812B for about $9.5 per 300 when assembled (C2843785), so designing your own WS2812B matrix modules as single-sided aluminium-core PCBs (also a special offer right now 5 pcs 100mm×100mm for just a couple of dollars) would be a possibility, with final assembly on top of an extruded aluminium heatsink.  Even if it draws only say 8A at 4V, that's still almost 32W of heat generated.  Note that for reproducing Western letters, an odd number of rows is better than an easier number of rows: 7 or 9 is much better than 8.  As typical letters are then 5 LEDs wide, you can count 5+1 (letter spacing) = 6 LEDs per letter.  Add occasional bypass capacitors between supply and ground for added stability.  You can chain up to 1024 LEDs in one data chain before you need to consider splitting them into multiple parallel chains.  I would consider creating 11×11-LED panels with vertical data chains (leftmost column first, then next column, all either bottom to top or top to bottom), as then each letter would be consecutive in memory even when displayed (similar to small 128×32 OLED displays), with slightly less than half the LED spacing on the sides, allowing continuous horizontal display by chaining modules.

All that said, it just might be more cost effective to use a small Linux SBC and an IPS TFT display instead; or perhaps a (hacked) digital photo frame.

An e-ink display would be even better, but large ones are pretty expensive.  BuyDisplay 4.2" (90mm×90mm, 85mm×61mm 400×300 active area) e-paper display modules cost about $23 apiece, and using multiple modules you'd have a small gap between modules, but would be daylight-readable.  Power would only be drawn when changing the displayed text, and it's black-and-white only.  Waveshare has 5.8" (152mm×52mm, 139mm×48mm 792×272 active area) e-paper modules for about $32 apiece, with cheapest BW 13.3" (286mm×212mm, 275mm×195mm 960×680 active area) SPI module costing about $150.

[JustMeHere]

Start here: https://github.com/2dom/PxMatrix

I'm using a 5V 8A power supply.  Not sure how much I need in the A dept, but when the panel is on white and at 90% intensity (don't go higher) there are 3 x ROWS x COLS LEDs drawing current. 

Note that your car will have 14 V when running.  It will drop down to 9 V while cranking.  Also, if you have a bad connection to your battery, you can see some massive voltage spikes (called a load dump.)  They are rare but can be large. 

App notes will help design the power supply:
https://www.ti.com/lit/an/snva780/snva780.pdf
https://www.ti.com/lit/an/snva681a/snva681a.pdf

And this datasheet is your application PCB:
https://www.ti.com/lit/ug/snva379e/snva379e.pdf

[MrFollies]

Thanks for the helpful information everyone.    I guess I was mainly worried about drawing too much current if the LEDs are turned on.
Maybe the easiest thing for me is to just put a 10a fuse on it and see what happens.

I actually don't expect for it to draw a lot...

[tooki]

Some of the libraries for driving addressable LEDs allow you to set a maximum current in software, and will simply dim the entire string to stay within the limit. So, for example, if you set all the LEDs to 100% white, it will dim it to a lower percentage, but if your display is mostly dark pixels, they’ll be left at full brightness.

[MrFollies]

Some of the libraries for driving addressable LEDs allow you to set a maximum current in software, and will simply dim the entire string to stay within the limit. So, for example, if you set all the LEDs to 100% white, it will dim it to a lower percentage, but if your display is mostly dark pixels, they’ll be left at full brightness.

Wow, that is actually a great feature.    I'll look it up.

[MrFollies]

A followup question.
I have a nice Pololu 15A 24->6v step down regulator.  DC24V150F6

While I am waiting for a 12->5v regulator to arrive, is there an easy way for me to drop the voltage down to an acceptable level for the WS2312B LEDs?
They can handle 5.3v.  So I was thinking I could use a (zener?) diode in the 6v positive output.     I know this is poor design, just wanted to get something going on the test bench.

Am I right in thinking that the power rating of the diode needs only to handle the dissipated power?    i.e. 0.7v over 15a is a bit over 10w?    So 2 zener diodes in parallel would give me a quick 5.3v?

Would Schottky diodes (3 in series to get 5.1v) be a better choice for this?

If I we're more confidant I'd consider hacking the pololu reg's voltage divider to get it down to 5v but I'd be asking the same question about power ratings and I'm not that confident identifying the divider circuit on the reg.

[Smokey]

Some numbers...

These LEDs are usually about 50mA max, so for example, assume about 5A for 100 leds at 5V.

....  So that's 25Watts.  But remember that switch mode converters are power converters so 25W from the 12V supply is only 2.08Amps.  Your power supply will only draw about 2A from 12V to make 5A at 5V.

[MrFollies]

Thanks smokey...   But that's not what I am asking.

I am asking if I drop the supplied 6v voltage to 5.3 with some zener (or other) diodes (forward bias), is 10w (10.5 to be exact) enough to cover about 15A?   (I have 256 leds - 8x32 board).
In other words...   For a zener diode with forward bias, dissipating 0.7v of power should be around 10w when the current draw is about 15A.

Or does the power calculation for the diode rating need to take into consideration the entire voltage flowing through the diode (i.e. 6v = 90W @ 15A)

[Smokey]

256 LEDs * 0.05A max per LED = 12.8A.  At 5V that's 64W.

The WS2812B have their own internal PWM driver.
The blue LED needs about 3.7V, so technically that's the minimum voltage you can send them and still have them work. 

So for maximum efficiency you want a 12V to 3.9V switching power supply (which you can probably accomplish with a 5V regulator that has some adjustment range).  At 3.9V and 12.8A that's essentially 50W.  So 50W / 12V = 4.16A total from your 12V supply in your car....

But... You will run into voltage drop along the WS2812B string with 256 LEDs, so you will probably need to split that 12.8A up over 2 or 3 different power supplies and power points on the LED string.

.... and no, dropping from 6V to 5V (delta of 1V) with a diode at 13A is probably not a good idea.  P=IV, so at 13A and 1V, that's 13W continuous in the diode(s).  They will get pretty hot.  Just get 2 10A 12V->5V regulators.  They sell them for car LED installs for like $30.

[Buriedcode]

A zener is a shunt regulator - and a poor one at that. It would be very inefficient to drop voltage using a zener. 

If I read your posts correctly.. you already have a 24V to 6V buck/step down that is capable of the current you *think* you require. You want to use this for the LED's so ~5.2V.  Dissipating 15A * 0.8 = 12W.
Yes 3 Schottky diodes in series, each on a heatsink would mean that each diode dissipates less power than a single diode - and as this is a one off cost isn't really an issue, isn't a bad idea.

But why would you need 15A?  If you're making a "sign" then surely you'll never have all the LED's on at the same time?  You're accounting for worst case - all LED's on, at full brightness, all white. Unless you're making a general purpose outdoor light I don't see why you would do this (it would be really inefficient for start, both in terms of power and the fact you're got RGB, individually addressable LED's jsut doing white).

Displaying text will reduce the current requirement considerably, unless you also have the background at some bright colour, which would make it harder to read.  You can halve the current further simply by having the maximum value of each LED's dies colour component 127 rather than 255.

First decide what you're going to display, then let that determine your worst case current requirement, then that determines *how* you power the LEDs.

[tooki]

Some numbers...

These LEDs are usually about 50mA max, so for example, assume about 5A for 100 leds at 5V.

....  So that's 25Watts.  But remember that switch mode converters are power converters so 25W from the 12V supply is only 2.08Amps.  Your power supply will only draw about 2A from 12V to make 5A at 5V.

They’re 60mA per LED.

[tooki]

While I am waiting for a 12->5v regulator to arrive, is there an easy way for me to drop the voltage down to an acceptable level for the WS2312B LEDs?
They can handle 5.3v.

5.3V is what the datasheet lists as the absolute maximum value. Normally, this is where I’d remind you that “absolute maximum values” are NOT your design values — a common beginner mistake is to assume the absolute maximum values are the numbers you should design around. They aren’t (they’re simply the values beyond which the device may be permanently damaged; it may stop working properly before that!), and a proper datasheet will list the design values in the electrical characteristics.

Now, because the WS2812B datasheet is awful, and later on lists a supply voltage of 4.5-5.5V, I think it’s fair to assume they actually can handle 5.3V, though I see no advantage in pushing them to their limits. I’d just get a 5V buck converter, or one whose voltage you can adjust to 4-5.0V.

[Smokey]

Some numbers...

These LEDs are usually about 50mA max, so for example, assume about 5A for 100 leds at 5V.

....  So that's 25Watts.  But remember that switch mode converters are power converters so 25W from the 12V supply is only 2.08Amps.  Your power supply will only draw about 2A from 12V to make 5A at 5V.

They’re 60mA per LED.

..maybe.. it would be great if they put that in the data sheet huh

[tooki]

..maybe.. it would be great if they put that in the data sheet huh

Yeah, it would. But I'll refer to my second reply above:

...the WS2812B datasheet is awful...

Interestingly, the datasheet for the WS2812B-2020 (the smaller 2020 size) fixed some of the issues in the original datasheet, such as listing 12V in the prose (!), and it does indeed list current: 12mA per channel (so 36mA total). Note that this chip version is different, using a 2kHz PWM frequency and reducing the current for those smaller LEDs, so its datasheet doesn't directly apply to the original 5050-size model.

[Smokey]

..maybe.. it would be great if they put that in the data sheet huh

Yeah, it would. But I'll refer to my second reply above:

...the WS2812B datasheet is awful...

Interestingly, the datasheet for the WS2812B-2020 (the smaller 2020 size) fixed some of the issues in the original datasheet, such as listing 12V in the prose (!), and it does indeed list current: 12mA per channel (so 36mA total). Note that this chip version is different, using a 2kHz PWM frequency and reducing the current for those smaller LEDs, so its datasheet doesn't directly apply to the original 5050-size model.

https://www.digikey.nl/htmldatasheets/production/2906294/0/0/1/ws2813b-rgb-led.html
5050 size WS2813 lists 18mA each so 54mA full on. 

https://www.lcsc.com/datasheet/lcsc_datasheet_1809261040_Worldsemi-WS2813B-B_C114592.pdf
And that datasheet lists 16mA

Fun huh!

... and yes I know WS2813 != WS2812

[Nominal Animal]

They’re 60mA per LED.

..maybe.. it would be great if they put that in the data sheet huh

It is in the WorldSemi WS2812 datasheet (20mA per color component, under LED characteristic parameter).

The WorldSemi WS2812B datasheet lists 16.5mA per color component (thus 50mA total; under Typical Application Circuit), when using either of the recommended circuit schemes (shown for 5V and 12V supplies).

[tooki]

The WorldSemi WS2812B datasheet lists 16.5mA per color component (thus 50mA total; under Typical Application Circuit), when using either of the recommended circuit schemes (shown for 5V and 12V supplies).

Link? No WS2812B datasheet I have found contains any such specification under the example circuits. I've never found a datasheet showing the 12V configuration at all.

This is the datasheet one finds all over the place: https://www.alldatasheet.com/datasheet-pdf/pdf/1179113/WORLDSEMI/WS2812B.html

[MrFollies]

.... and no, dropping from 6V to 5V (delta of 1V) with a diode at 13A is probably not a good idea.  P=IV, so at 13A and 1V, that's 13W continuous in the diode(s).  They will get pretty hot.  Just get 2 10A 12V->5V regulators.  They sell them for car LED installs for like $30.

Like I said, it was just to get going with what I had on hand for testing...    For the end application I'd find the right regulation.
Each LED board (8x32) has 3 plugs so the idea of having a common rail and perhaps using 3x5A regulators sounds like a really good idea to me - thanks!

Edit: Hang on a sec.    What happens when 3x5A regs all have slightly different voltage output values.   If they are just connected in parallel could they interfere with each outer?

[MrFollies]

A zener is a shunt regulator - and a poor one at that. It would be very inefficient to drop voltage using a zener. 

If I read your posts correctly.. you already have a 24V to 6V buck/step down that is capable of the current you *think* you require. You want to use this for the LED's so ~5.2V.  Dissipating 15A * 0.8 = 12W.
Yes 3 Schottky diodes in series, each on a heatsink would mean that each diode dissipates less power than a single diode - and as this is a one off cost isn't really an issue, isn't a bad idea.

Thanks.    I know it's not terribly efficient - but I guess I was interested in if my assumptions about the power being handled by the diode were correct.    Thanks for replying.

But why would you need 15A?  If you're making a "sign" then surely you'll never have all the LED's on at the same time?  You're accounting for worst case - all LED's on, at full brightness, all white. Unless you're making a general purpose outdoor light I don't see why you would do this (it would be really inefficient for start, both in terms of power and the fact you're got RGB, individually addressable LED's jsut doing white).

Displaying text will reduce the current requirement considerably, unless you also have the background at some bright colour, which would make it harder to read.  You can halve the current further simply by having the maximum value of each LED's dies colour component 127 rather than 255.

You're 100% correct of course.    But since I'm going to be writing code, I wanted something in place to make sure if my code were wrong and it went all white, I did not destroy the LEDs.    Of all the components, they are the hardest to get hold of (7-14 days from China).

But that makes me think of the power supply.      What happens to a power supply when the output load draws way too much amps.     Does it damage the reg?     Do some types of regs handle this better than others?

[Nominal Animal]

I may have looked at the wrong datasheet by accident, because I cannot find it anymore.

I recommend looking up the datasheet at LCSC, because they tend to include the current draw.  It looks like WorldSemi has provided many different versions of the datasheet to different vendors.  (WS2812B-V5, WS2812B-XF02, et cetera.)

Anyway, even though WorldSemi does not allow direct download for their datasheets, the specifications for the various chip variants is listed here.  In short:
WS2812A = 3×16mA = 48mA
WS2812B = 3×12mA = 36mA (all WS2812B-... variants)
WS2812C = 3×5mA = 15mA (all WS2812C-... variants)
WS2812D = 3×12mA = 36mA
WS2812E = 3×12mA = 36mA
WS2812S = 3×16mA = 48mA
Although that page states 3.3V - 5.5V, their datasheets vary a bit in that, with the shared common range being 3.7V - 5.3V (according to the datasheets at LCSC).

[Nominal Animal]

Because the current limiting in the WS2812 chips is linear, using supply voltage closer to the lower limit will mean less heat generated in the LED matrix at the same light output.  It is the heat that will fry your LEDs.  For example, if you used 4V instead of 5V supply, the power dissipated would drop by 20%.

Just out of interest, I looked at TI Webench for high-efficiency converters with DC input between 8V and 24V, output 4V at 4A, <10mV ripple, using integrated inductor modules.  One of the circuits it suggested is based on TLVM13640, which Mouser sells for 6.59€ in singles.  The other components are two 47µF X7R 6.3V 10% caps, two 10µF X7R 50V 20% caps, one 1µF X7R 10V 10% cap, one 8.2pF C0G 50V (0.25pF ≃ 3% tolerance) cap, a 14 kΩ resistor, and a pair of resistors of about 40kΩ total as a voltage divider for 1V output; 28.7kΩ and 10kΩ for 3.9V, 30kΩ and 10kΩ for 4.0V, and so on.  It should exceed 90% efficiency for 0.5A-4A current; about 95% at 1.5A-2.5A.  I like these modules, because having the inductor integrated it is easy to keep the loop area small even on a 2-sided board.  Do look at section 11 Layout in the TLVM13640 datasheet, and remember that some voltage will be dropped by the cables and traces between the supply and the LEDs.

Thermal dissipation is key here.  At 4V 4A = 16W output, Webench promises something like 93% efficiency, which means input power is 16W/0.93 = 17.2W, and the difference, 1.2W, is dissipated as heat.  Based on the datasheet, on a four-layer 2oz 75mm × 75mm board, the junction-to-ambient thermal resistance is 33.1°C/W, so the temperature increase would be about 40°C.  Designing some kind of a heatsink sounds like a good idea to me.  Maybe use a big Al heatsink with a Ø 8mm hole for the IC on the same side as the IC, with the rest of the components on the other side of a two-sided PCB?  Dunno, me a hobbyist.

[Smokey]

.... and no, dropping from 6V to 5V (delta of 1V) with a diode at 13A is probably not a good idea.  P=IV, so at 13A and 1V, that's 13W continuous in the diode(s).  They will get pretty hot.  Just get 2 10A 12V->5V regulators.  They sell them for car LED installs for like $30.

Like I said, it was just to get going with what I had on hand for testing...    For the end application I'd find the right regulation.
Each LED board (8x32) has 3 plugs so the idea of having a common rail and perhaps using 3x5A regulators sounds like a really good idea to me - thanks!

Edit: Hang on a sec.    What happens when 3x5A regs all have slightly different voltage output values.   If they are just connected in parallel could they interfere with each outer?

No, not in parallel.  You break the positive supply line of the string of LEDs after some number and you power each section from its own power supply.  Think about it.  If you have 15A of current running through a string that is all going through the trace on that flat flex PCB the LEDs are attached to.  What gauge wire would you want for 15A and how much copper is actually in that flat flex trace?  You will get killer voltage drop and then it will melt from dissipation.

Go check out the WLED resources.  They have help pages describing exactly how to wire all this stuff up.  If you are planning on writing your own software and haven't seen WLED, you should check out that too.

[MrFollies]

No, not in parallel.  You break the positive supply line of the string of LEDs after some number and you power each section from its own power supply.  Think about it.  If you have 15A of current running through a string that is all going through the trace on that flat flex PCB the LEDs are attached to.  What gauge wire would you want for 15A and how much copper is actually in that flat flex trace?  You will get killer voltage drop and then it will melt from dissipation.

Go check out the WLED resources.  They have help pages describing exactly how to wire all this stuff up.  If you are planning on writing your own software and haven't seen WLED, you should check out that too.

Yeah that makes a lot of sense.     The problem is that breaking the VCC track (looks like it's a filled plane) on the flexible tracks might be difficult.    I will check it out.