Just Starting Out- Questions Regarding IC's, Waterproofing, Controllers, etc.

[Sparky95]

Hello All,

My name is Nick, I'm new and have a number questions before I start dropping money on the wrong things. I've tried to find what answers I could in things like the wiki, so hopefully I've narrowed it down somewhat.

Where this all started:
I saw a Twinkly rgb christmas tree and decided I must have one, but wondered if I could diy it for less money with more options. (Yes, you can.)

I was soon looking at an arduino, some 12V WS2811 bullet pixel strings, and even more questions than I started with. I'll take you through my thought process, and hopefully you can understand how and why I've come to the conclusions I have, and guide me. (Thanks in advance!)

Where it quickly ended up in my head:
Suddenly I found myself envisioning my big tree all lit up, as well as pixel icicle lights, a whole dazzling show in fact! (I know, I know. Don't get ahead of myself.) (I'll try.)

Roadblock #1:
Max number of pixels per string. My tree is big you see, big enough that I will easily exceed 1000 pixels. Here I was reading that an arduino mega is going to struggle driving more than 500... What to do?

Here's where things got complicated:
So an arduino can't drive more than 500 or so pixels at a decent frame rate, but WS2811s require tight timing that isn't easily achievable with a non-realtime controller such as a raspberry pi. Solution? 4 wire pixels with a separate clock line, which are less dependent on tight timing. Why are these WS2811 pixels so popular? (As I later found out the answer is Falcon Controllers and the like.)

Roadblock #2:
Which IC? There are so many to choose from, or so I thought. Ideal would be the APA107, but alas it isn't available in anything other than integrated smd leds. Next best option seems to be the P9813. Unfortunately the P9813 strings aren't very available, much less well waterproofed with black or green wire. So I find myself looking at buying them unwaterproofed, soldering on my own wires, and waterproofing them. With the waterproofing step I'm officially in over my head.

Revelation! (Doesn't solve everything like I initially thought, but hey.)
Enter the Falcon Controller. Suddenly I can drive up to 16000 WS2811 pixels at a decent refresh rate, albeit not all in one long strand. That's ok because voltage drop starts to be a pain after 1000 pixels anyway.

Roadblock #3:
Waterproofing. Apparently these bullet pixels aren't as waterproof as they say they are. Also they apparently shouldn't be supported by their strings like normal string lights.

Game Plan:
So here's what I've got so far.

Strands:
I'm going to get "IP67" 50 pixel 5V WS2811 strands with green wire.

At one end of each strand there will be a little waterproof aluminium box with the 3 strand wires along with four additional 14awg green wires going in one end of the box, and a 5 pin 17awg male pigtail coming out of the other. In the box will be a 48VDC to 5VDC @3A buck converter, some rf bypass capacitors, some fuses (1A for 48V, 4A for 5V), and a piece of coax to shield the data line as it passes through the box. All grounds are common. Each strand of 50 has a separate 5V supply this way. The four 14awg wires run the length of the strand and carry the 48V power. They also serve as the support wire of the strand, so that no tension is put on the wires entering the bullet.

At the other end of the strand will be a 5 pin 17awg female pigtail. Pins 1 and 2 are connected to +48V (2 of the 14awg wires) , pin 3 is data, and pins 4 and 5 are ground (the other two 14awg wires). The +5V at the end of each strand of 50 will connect to an rf bypass capacitor, and the other end of the capacitor will connect to ground.

This effectively creates plug and play 48VDC WS2811 strands, where max current for 1000 pixels is less than 10A. I plan to spread the wires apart at each pixel and dip it in a waterproofing sealant of some sort.

Controller:
I plan to use a Falcon F16-V3 and a desktop computer. Power for the F16-V3 will be supplied by a 48V to 12V buck converter. Each data line will branch out (along with a 10awg 48V power feeder) to its respective strand, using however many F-amps I need to get the data to the beginning of the strand. Each F-amp will be powered by a 48V to 12V buck converter. How many feet should I put between F-Amps, and how many feet between the F16-V3 and the first F-Amp? F16-V3 will be inside, strands will be up in a tree. Would running the data feeder through coax help it go farther?

Power Supply:
120VAC to 48VDC 30A Meanwell SMPS, fed to three 10A individually switched, fused, and metered branch circuits. More supplies if needed. All grounds common system wide, being sure to avoid ground loops of course.

Questions/Notes: (This is the last of them, for now :P)

Regarding waterproofing, I read an article on the wiki here detailing the options for sealing pixels such as clear epoxy, plasti-dip, and liquid tape. It was an older article, so do you have any updates regarding how they've faired over time?

In general, have I missed anything major? Anything I should know, flaws in plan, etc? I live in Bay City, MI so it gets cold and wet if that helps. I plan on making two prototype strands and hanging them in my room to start. I'll probably place them outside to see how they handle being wet too.

Are the 4 wire types capable of more pixels per string and faster data rates worth exploring?

Thank you for reading my small novel! Any questions, comments, suggestions are much appreciated! Also if this belongs elsewhere on the forum please let me know.

[MartinMueller2003]

Hello All,

My name is Nick, I'm new and have a number questions before I start dropping money on the wrong things. I've tried to find what answers I could in things like the wiki, so hopefully I've narrowed it down somewhat.

Where this all started:
I saw a Twinkly rgb christmas tree and decided I must have one, but wondered if I could diy it for less money with more options. (Yes, you can.)

I was soon looking at an arduino, some 12V WS2811 bullet pixel strings, and even more questions than I started with. I'll take you through my thought process, and hopefully you can understand how and why I've come to the conclusions I have, and guide me. (Thanks in advance!)

Where it quickly ended up in my head:
Suddenly I found myself envisioning my big tree all lit up, as well as pixel icicle lights, a whole dazzling show in fact! (I know, I know. Don't get ahead of myself.) (I'll try.)

Welcome Nick.

Arduino is a series of devices. The ESP8266 series of devices have proven to be very capable WS2812 controllers. The Arduino IDE supports the ESP series of controllers. Many people are moving to the WeMos D1 Mini (an ESP8266 platform). Works great for over 600 pixels on a single string.

Roadblock #1:
Max number of pixels per string. My tree is big you see, big enough that I will easily exceed 1000 pixels. Here I was reading that an arduino mega is going to struggle driving more than 500... What to do?

Here's where things got complicated:
So an arduino can't drive more than 500 or so pixels at a decent frame rate, but WS2811s require tight timing that isn't easily achievable with a non-realtime controller such as a raspberry pi. Solution? 4 wire pixels with a separate clock line, which are less dependent on tight timing. Why are these WS2811 pixels so popular? (As I later found out the answer is Falcon Controllers and the like.)

700 - 800 pixels is max single string length for a 25ms refresh rate.
For 5v pixels you will want to inject power at least every 50 pixels in a string.
50 pixels draws 3A. This means your buck converters would be running at 100% load. Not a good design. You should leave 20% buffer. That means your buck converters should not be driving 50 pixels.

Roadblock #2:
Which IC? There are so many to choose from, or so I thought. Ideal would be the APA107, but alas it isn't available in anything other than integrated smd leds. Next best option seems to be the P9813. Unfortunately the P9813 strings aren't very available, much less well waterproofed with black or green wire. So I find myself looking at buying them unwaterproofed, soldering on my own wires, and waterproofing them. With the waterproofing step I'm officially in over my head.

Revelation! (Doesn't solve everything like I initially thought, but hey.)
Enter the Falcon Controller. Suddenly I can drive up to 16000 WS2811 pixels at a decent refresh rate, albeit not all in one long strand. That's ok because voltage drop starts to be a pain after 1000 pixels anyway.

The big driver here has been cost. The WS2811s are not the best choice but they are the lower cost option. Voltage drops gets to be a problem after 15 pixels. I also use buck converters but I power inject every 25 pixels. Lots of available overhead.

Roadblock #3:
Waterproofing. Apparently these bullet pixels aren't as waterproof as they say they are. Also they apparently shouldn't be supported by their strings like normal string lights.

Bullet nodes are SUPER SENSITIVE to tension on the wires. Look up Boscoyo bullet mounting strips to see a good option for protecting and aligning your pixels. There are MANY other options for holding your pixels. The wires running between the pixels ARE NOT ON THAT LIST.

Game Plan:
So here's what I've got so far.

Strands:
I'm going to get "IP67" 50 pixel 5V WS2811 strands with green wire.

At one end of each strand there will be a little waterproof aluminium box with the 3 strand wires along with four additional 14awg green wires going in one end of the box, and a 5 pin 17awg male pigtail coming out of the other. In the box will be a 48VDC to 5VDC @3A buck converter, some rf bypass capacitors, some fuses (1A for 48V, 4A for 5V), and a piece of coax to shield the data line as it passes through the box. All grounds are common. Each strand of 50 has a separate 5V supply this way. The four 14awg wires run the length of the strand and carry the 48V power. They also serve as the support wire of the strand, so that no tension is put on the wires entering the bullet.

At the other end of the strand will be a 5 pin 17awg female pigtail. Pins 1 and 2 are connected to +48V (2 of the 14awg wires) , pin 3 is data, and pins 4 and 5 are ground (the other two 14awg wires). The +5V at the end of each strand of 50 will connect to an rf bypass capacitor, and the other end of the capacitor will connect to ground.

This effectively creates plug and play 48VDC WS2811 strands, where max current for 1000 pixels is less than 10A. I plan to spread the wires apart at each pixel and dip it in a waterproofing sealant of some sort.

When you start talking about 10s of thousands of pixels, and divide by 25 you will get an interesting number for the quantity of those boxes. I use 3d printed boxes to provide the same functionality,

FYI: As for eliminating ground loops, that is VERY difficult to accomplish. There is a general guideline that makes it hard. Data and V- must run together from one pixel to the next. That means that tying grounds together elsewhere DOES NOT MEET THIS REQUIREMENT. Think of it as "Data and V- must be in the same cable and should be twisted around each other and both wires must be connected at both ends to the pixels.

Controller:
I plan to use a Falcon F16-V3 and a desktop computer. Power for the F16-V3 will be supplied by a 48V to 12V buck converter. Each data line will branch out (along with a 10awg 48V power feeder) to its respective strand, using however many F-amps I need to get the data to the beginning of the strand. Each F-amp will be powered by a 48V to 12V buck converter. How many feet should I put between F-Amps, and how many feet between the F16-V3 and the first F-Amp? F16-V3 will be inside, strands will be up in a tree. Would running the data feeder through coax help it go farther?

Power Supply:
120VAC to 48VDC 30A Meanwell SMPS, fed to three 10A individually switched, fused, and metered branch circuits. More supplies if needed. All grounds common system wide, being sure to avoid ground loops of course.

Questions/Notes: (This is the last of them, for now :P)

Regarding waterproofing, I read an article on the wiki here detailing the options for sealing pixels such as clear epoxy, plasti-dip, and liquid tape. It was an older article, so do you have any updates regarding how they've faired over time?

You might want to look into the F48. It gives you an easy way to run data injection points far away from the controller.

As for data driver devices (F-Amp is only one of the types available). There is a lot of controversy and each installation has its own special conditions. The following numbers are based on empirical evidence and are not definitive values.

Cheap 24 AWG wire with no twists: 10' - 15' between regeneration points.
Good 24 AWG wire with twists: 15' - 20' between regeneration points.
CAT5 24AWG solid wire: 30" or more (that is as far as I measured).
CAT6 22 AWG solid wire: 45' or more (that is as far as I measured).

In general, have I missed anything major? Anything I should know, flaws in plan, etc? I live in Bay City, MI so it gets cold and wet if that helps. I plan on making two prototype strands and hanging them in my room to start. I'll probably place them outside to see how they handle being wet too.

Are the 4 wire types capable of more pixels per string and faster data rates worth exploring?

Thank you for reading my small novel! Any questions, comments, suggestions are much appreciated! Also if this belongs elsewhere on the forum please let me know.

The pixel types with 4 wires typically are able to support longer strings at the same refresh rate. It is not the pixel type that controls string length, it is the data rate. The WS2811 pixels have an 800000 bit rate. This translates to ~700 pixels at 25ms refresh rate. In essence the WS2811 based strings have an infinite string length. BUT the refresh rate on a 3000 pixel string would be so slow that it is not useful for blinky flashy.

The WS2801 can support up to 1000000 bits/second and would therefore support up to 820 pixels in a string at 25ms. I do not know the data rates for the other pixel types.

Hope this helps
Martin

[Sparky95]

Arduino is a series of devices. The ESP8266 series of devices have proven to be very capable WS2812 controllers. The Arduino IDE supports the ESP series of controllers. Many people are moving to the WeMos D1 Mini (an ESP8266 platform). Works great for over 600 pixels on a single string.

My apologies, I should have been more specific. I was looking at the Arduino Mega 2560, but will look into the IDE series you mentioned.

700 - 800 pixels is max single string length for a 25ms refresh rate.
For 5v pixels you will want to inject power at least every 50 pixels in a string.
50 pixels draws 3A. This means your buck converters would be running at 100% load. Not a good design. You should leave 20% buffer. That means your buck converters should not be driving 50 pixels.

50 pixels draws 3A at full brightness/full white. I find full brightness to be too much, and they're not going to be full white most of the time. So it won't be running at 3A much, if ever. It will likely hover between 1.5 and 2A. Well within spec, especially with a heatsink on the IC. I may add a second set of wires spanning the set to feed 5V from both ends. I've ordered a couple of the buck converters I plan to use and am going to torture test them to determine their safe operating range too.

Bullet nodes are SUPER SENSITIVE to tension on the wires. Look up Boscoyo bullet mounting strips to see a good option for protecting and aligning your pixels. There are MANY other options for holding your pixels. The wires running between the pixels ARE NOT ON THAT LIST.

As I said, I'll be running four 14awg wires along each strand, and the pixel strand will be anchored to it to prevent any tension being applied to the wires entering the bullet. If need be I can add support cable to the bundle.

When you start talking about 10s of thousands of pixels, and divide by 25 you will get an interesting number for the quantity of those boxes. I use 3d printed boxes to provide the same functionality.

I agree, however I'm using aluminum boxes for two reasons. One is heatsinking for the buck converter, the other is for rf shielding. It's a Faraday cage. Good news is that quantity reduces price somewhat.

FYI: As for eliminating ground loops, that is VERY difficult to accomplish. There is a general guideline that makes it hard. Data and V- must run together from one pixel to the next. That means that tying grounds together elsewhere DOES NOT MEET THIS REQUIREMENT. Think of it as "Data and V- must be in the same cable and should be twisted around each other and both wires must be connected at both ends to the pixels.

I should have said" eliminate ground loops as much as possible". Data and ground will be fed from one pixel to the next, but ground will also be connected at one or both ends. If one proves sufficient then it'll eliminate that ground loop at least. If not chokes could fashioned to eliminate the ground loop as far as the data is concerned.

You might want to look into the F48. It gives you an easy way to run data injection points far away from the controller.

I'm avoiding the F48 because it locks me into 3 wire pixels.

As for data driver devices (F-Amp is only one of the types available). There is a lot of controversy and each installation has its own special conditions. The following numbers are based on empirical evidence and are not definitive values.

Cheap 24 AWG wire with no twists: 10' - 15' between regeneration points.
Good 24 AWG wire with twists: 15' - 20' between regeneration points.
CAT5 24AWG solid wire: 30" or more (that is as far as I measured).
CAT6 22 AWG solid wire: 45' or more (that is as far as I measured).



The pixel types with 4 wires typically are able to support longer strings at the same refresh rate. It is not the pixel type that controls string length, it is the data rate. The WS2811 pixels have an 800000 bit rate. This translates to ~700 pixels at 25ms refresh rate. In essence the WS2811 based strings have an infinite string length. BUT the refresh rate on a 3000 pixel string would be so slow that it is not useful for blinky flashy.

The WS2801 can support up to 1000000 bits/second and would therefore support up to 820 pixels in a string at 25ms. I do not know the data rates for the other pixel types.

Hope this helps
Martin

Thank You. That definitely helps quite a bit. I'm not planning on too much fast movement, so I think the F16-V3s specified 24fps at 1024 pixels should be sufficient. The P9813 has a much faster data rate than the WS2801, so in theory it is the superior choice.

[MartinMueller2003]

24fps is well below real motion rates. It is about a 40ms refresh rate. To me, that is a visible issue on my mega tree and arches. FYI 25ms refresh rate is 40fps. That is 10 fps faster than the eye can easily track.

[ezellner]

I run an all 5 V setup. Although somewhat different than what you are proposing (for example, no buck converters - all driven from 5 V Meanwell power supplies), it might be helpful to see a working installation. I am not that far from you and if you're willing to take the drive, I can walk you through what I have done which may help seeing it live. PM me if interested. I'm retired so am available almost any time.
A couple of separate additions to previous comments. Make sure you specify pixel wire gauge at 18 AWG or better. If not, you'll get 24 AWG and you won't be happy. I would not consider waterproofing my own pixels. The IP67 pixels I have purchased are holding up quite well after three seasons with no signs of deterioration. All my power wiring is 14 AWG and signal is 18 AWG stranded (no shielding). I run between 50 and 70 pixels per power feed and intensity runs at 50% which is still very bright.

[Sparky95]

24fps is well below real motion rates. It is about a 40ms refresh rate. To me, that is a visible issue on my mega tree and arches. FYI 25ms refresh rate is 40fps. That is 10 fps faster than the eye can easily track.

Duly noted, thank you. I am quite sensitive to flicker and movement (ac led strand flicker gives me a headache), so I'd likely have a problem with it. The Falcon states that it can do 25ms with 680 pixels or less iirc. So if I go with 600 I should be just fine.

I run an all 5 V setup. Although somewhat different than what you are proposing (for example, no buck converters - all driven from 5 V Meanwell power supplies), it might be helpful to see a working installation. I am not that far from you and if you're willing to take the drive, I can walk you through what I have done which may help seeing it live. PM me if interested. I'm retired so am available almost any time.
A couple of separate additions to previous comments. Make sure you specify pixel wire gauge at 18 AWG or better. If not, you'll get 24 AWG and you won't be happy. I would not consider waterproofing my own pixels. The IP67 pixels I have purchased are holding up quite well after three seasons with no signs of deterioration. All my power wiring is 14 AWG and signal is 18 AWG stranded (no shielding). I run between 50 and 70 pixels per power feed and intensity runs at 50% which is still very bright.

I considered running it all right at 5V, but for 1000 pixels that's 60A. I'd either have to string it up with 2 AWG or I'd have to have 40-60 separate power feeds running up the tree. Since I could very well end up with 3000 pixels, I'd end up running welding cable from the power supply. Welding cable is expensive, and voltage drop is still a concern. The beauty of running 48V is that I can lose 12V due to voltage drop at the end of the line, and the remaining 36V will still have plenty of power for the buck converter to continue spitting out 5V. I also save more on wire than the buck converters cost. Add to that the weight savings (I am stringing these up in a tree after all) and its almost a no brainer for me. I'd go 72V or even higher if buck converters were cheap and readily available for it. My main requirement is that I be able to string them end to end with no additional power injection, up to the number of pixels a port can theoretically control (1000 or 20 strands). The buck converters satisfy that.

The pixel strands I was looking at use 22 AWG, as I was unable to find 5V strands with 18 AWG green wire. That said, 22 AWG should be sufficient for a 50 pixel strand if fed from both ends.

I wasn't considering waterproofing non-waterproof pixels, but I had read that the IP67 pixels were notoriously unreliable, and far from actually being IP67. They suggested additional sealing of the pixel. I believe it was on here somewhere, but it was a couple years old too. Have manufacturers stepped up their quality since then? Or are you ordering from somewhere specific?

My power wiring will be 10 AWG on power feeders, dual 14 AWG for the strand (2 positive, 2 negative), and I'm not sure what I'm going to use for data. Martin shared some of his experience with various methods of data wiring, and I plan on testing with different types of coax to see if there's any room for improvement. With 22 AWG cat 5 he was getting close to 50ft with no amplifier iirc, and if I can reliably stretch that to 100ft then I wouldn't need anything between the controller and first pixels. Being that I'm using a bunch of buck converters, noise is a concern, so I'm not taking any chances and will definitely shield data wherever I can.

I too plan on running it at 50% brightness, but I try to make a habit of designing everything to be able to operate at 100% continuously, just in case.

[MartinMueller2003]

I tend to run 10-12 16 AWG wires out of a 60A PSU. Each one can easily drive over 100 pixels. However, I locate my PSUs at the base of my props so the runs are as short as possible.

Data wires have almost no current. As a matter of fact, the drivers are pretty weak and a large cable has enough capacitance to cause issues rather than reduce issues. For the Data and its companion ground, the goal would be to have as little current going through the two wires as possible. All of my data plus V- wires are made from cheap 24AWG wire. Beefy grounds sit at both ends of the data wire set since that is one of the locations of my injection points.

[Sparky95]

I tend to run 10-12 16 AWG wires out of a 60A PSU. Each one can easily drive over 100 pixels. However, I locate my PSUs at the base of my props so the runs are as short as possible.

Data wires have almost no current. As a matter of fact, the drivers are pretty weak and a large cable has enough capacitance to cause issues rather than reduce issues. For the Data and its companion ground, the goal would be to have as little current going through the two wires as possible. All of my data plus V- wires are made from cheap 24AWG wire. Beefy grounds sit at both ends of the data wire set since that is one of the locations of my injection points.

I plan on having a centralized PSU, but since it's a relatively high voltage the wires won't have to be too large. 10 gauge should be sufficient (read overkill) for a 50ft run at 48V 10A with relatively negligible voltage drop.

I honestly overlooked the capacitance issue with the data line, thank you for pointing that out.

I'm aware that data lines carry little current, but as your results show going from a 24awg cat 5 cable to a 22awg cat 6 cable allowed for a longer run nonetheless. Do you think it's the larger wire gauge, the shielding of cat 6, or both?

I have a good amount of various coax types laying around, so I'm inclined to see if it offers any advantages. Should keep capacitance low at least.

[MartinMueller2003]

IMO it is the higher twist per foot count and the extra shield.

[Sparky95]

IMO it is the higher twist per foot count and the extra shield.

I'm thinking I can probably use cat 6a s/ftp, which should allow me to go directly from the controller to 4 separate strings. Each twisted pair is shielded so there shouldn't be crosstalk. Then I could run a single heavy power feeder up the tree and branch off at each string. (String being 12 strands, or 600 pixels)

Have you used the WS2818 strings at all? It'd be rather nice to not have to worry about one pixel failure way up in a tree taking the whole string down.