RGB LED panel, or my first encounter with

A while ago, a friend asked me to help him buy the components for an LED scrolling sign, which he wanted to use to spruce up his little shop.
We'd only ever seen these signs at other places, so we started figuring out how to put one together.
We picked out what we wanted, I placed the order, received everything, and decided to try it out at home first.
There'll be an inspection, a teardown, and some tinkering with various bits of hardware—in other words, the usual :)

At first we were considering either a monochrome panel (usually red) or a full-color one. I said that if we were going to do it, we might as well go full color—it gives us more options for how to set it up.
We ordered everything from the same store, but the power supply first and everything else later. We ordered the PSU earlier because there was a sale—or so it seemed. As it turned out, it was fake; the price was the same.

So, in the end I received a small box with the power supply and a large box with everything else.



I had no complaints about the packaging of the more important components: everything was packed very neatly and snugly, so kudos to the seller. The parcel was shipped via Meest Express.



The LED panels were packed in one large sheet of polyethylene foam. They couldn't rub against each other, since they were essentially wrapped with a layer between them.



And here's the whole kit.
1. Power supply. Product link, price: $18.05, free shipping.
2. Onbon bx-5ql control board, product link, price: $16.88 ($22.73 including shipping).
3. Three P4 LED panels, product link, price: $13.98 ($57.17 for all 3 including shipping)

The whole kit came to about $98. I'll come back to the price later.



In addition to the items listed above, the package included:
1. A CD with the software. I was too lazy to dig out a DVD drive to see what was on it, since it was easier to download it from the official Russian-language website.
2. Mounting hardware. I didn't count it, but it looks like there's plenty. People usually order a frame to mount everything on as well, but it's not very cheap, so we decided to make do ourselves.
3. A set of cables: two power cables, each designed to connect two panels, and four signal cables—one long cable (to the controller) and three inter-panel cables. Since there are three panels, the third short cable is redundant, but it can be used if the controller is next to a panel.
4. The controller's connectors were given extra protection, which is good.



A small board controls the whole thing. When I placed the order, I knew almost nothing about it—maybe just that I needed to check the supported resolution and whether it could work in full-color mode.
As I understand it, there are three types: monochrome (one color), two-color, and full-color (256 colors). As you've probably guessed, we chose the last option.
The board is called the Onbon bx-5ql.



The board has some unpopulated areas, as well as contacts for connecting external devices.
I have no particular complaints about the build quality.
According to the documentation, the board can work with two hub75 sets or four hub08 sets.
The maximum supported resolution is 256x65 pixels.



The board has the following connectors for external devices:
1. A rather hefty terminal block for the power supply, rated for 5 V. It's "hefty" because the board's power consumption is very low, while the terminal block is huge :)
There are also footprints for humidity and temperature sensor connectors; we'll need one of them later.
2. To the left of the power contacts is a footprint for some kind of universal contact group, with the reset contacts routed there as well. A "test" button is next to it. Above it you can see a resettable fuse rated at 0.75 A (which gives you an idea of the terminal block's size). Though this is simply designed to make connections convenient: insert the wire, tighten the screw, and it's secure.
3. A USB connector for uploading the control program, with a spot for an IR receiver next to it (there's a version with remote control).
4. The board also has a footprint for a local network connector and its controller, but we bought the "lite" version of the controller.



But the biggest headache for me was the panel connectors. On the one hand, it's simple: four connectors of one color and two of another. But for some reason, there was no information about them anywhere. I understood that they were intended for different types of panels, but the connectors aren't keyed, and the panel itself doesn't indicate its type.
In the end, I simply checked the contacts with a multimeter (the pinouts differ noticeably) and connected it.



1. An ARM STM32F207 serves as the "brain".
2. The board also has W29N01GV flash memory with a capacity of 1 GB (128 MB) for storing the control program.
3. The outputs to the panels are connected through HC245 bus transceivers.
4. The board also has a battery to power the clock; a small watch crystal is visible next to it.



And here's the panel itself. It has a resolution of 64x32. When choosing a control board, remember that the panels' combined resolution must not exceed the controller's capabilities.
In our case, it was 192x32 (three panels). The controller supports 256x64 or 512x32, so we had plenty of headroom.
The panel measures 256x128 mm.



The panels are quite thin, about as thick as a matchbox.



Each panel consists of a large number of LEDs (2048 in this case), partially covered by a frame.



There is a split down the middle into two sections, each measuring 32x32. The plastic frame is held in place by nine small self-tapping screws. They're barely visible, though—my wife spotted them right away, while I missed them :)
Here you can see the seam between the two halves of the panel; the fit is excellent.



The housing has beveled edges that let you fit the panels together flush, or even at a negative angle if you like—in other words, you can wrap them around something round, although the radius will be fairly large.



Let's see how it's put together.
The LED panel consists of a plastic frame, a circuit board with components, and a front grid frame that holds the whole assembly together.
The housing frame has M3 mounting holes, which let you attach the whole assembly to a special metal frame.



The board is matte black and looks quite neatly assembled.





A little about the components.
As far as I can tell, the board has DP5020 LED driver ICs, ICN2012 controllers, and HC245 bus transceivers, just like the control board.
Power comes in through a four-pin connector. There are also two identical-looking connectors for data: an input and an output.
The controller connects to the input connector, the next panel to the output connector, and so on.



While preparing the material, I completely forgot about the power supply. I originally planned to write a separate review of it, but decided to just talk about it and show it here.
The power supply came in a cardboard box, wrapped in bubble wrap inside, although its case is so sturdy that I doubt it would have been damaged even without the extra protection.

All the power supply photos were taken after the panels had been given some preliminary tests and a short thermal run. I'll explain later why I mention this.
The power supply is rated at 200 W, with an output voltage of 5 V.



This is what's known as a low-profile power supply. The photo compares it with a more conventional 240 W, 12 V power supply.
The power supply measures 190x84x31 mm.



The power input and load terminals are on opposite sides of the case, and neither has a protective cover. That’s bad—or, more precisely, unsafe.



There’s nothing else to see on the outside, so let’s open it up.
The cover is held on by no fewer than six screws and two stickers: one with the manufacture date, the other marked “QC.”



The input filter is almost complete: capacitors, an inductor, a fuse, and a thermistor.
The input bridge diodes are 1N5408. To me, that seems a bit on the small side, but in reality it’s perfectly adequate.
The input capacitor’s value isn’t specified. I measured it at 130 µF, which is too little for 200 W of power.



The power supply uses a 09N90 high-voltage transistor. I didn’t look up the datasheet, but I’d guess it’s rated for 9 A and 900 V.
In other reviews, I often come across comments like, “Why take apart something that works fine?”
Fair enough—I’d tried this power supply too, loading it while testing how it worked with the panels. But when I opened it up, I was genuinely stunned (I’d put it more bluntly, but I can’t). The power transistor IS NOT SCREWED DOWN. You can see in the photo that there’s no nut. I didn’t remove it; it was never there. On the other side of the screw is a sticker saying “QC test passed”—irony of fate, or just cruel sarcasm.
A complete mess, plain and simple. But as my inspection showed, that’s not all. More on that later.

The feedback circuit and output voltage adjustment are tucked between the case and the transformer.

Sometimes I think Chinese engineers really do read my reviews. They got spooked and installed a whole bunch of Y1-class capacitors—where they’re needed and even where they could have been left out, just to be on the safe side.



The transformer is very flat and wound with a busbar rather than the litz wire normally used in this kind of application.

The output inductor is wound with four strands of wire. It starts buzzing noticeably at around 40% load.

The output has four 3300 µF, 10 V capacitors, with a combined measured capacitance of about 14000 µF.



I didn’t remove the board from the case, but I was a little surprised to find four diode assemblies on the output, and of two different types, with different packages too. I might take the power supply apart and write a separate review of it later, but I haven’t decided yet.

And here’s more of the “mess.” One pair of diodes has an aluminum shim between the diodes and the case, and the shim is crooked. But that’s not all: the diodes were barely pressed against it. I loosened the screws without any effort—they were only loosely threaded in, just enough to hold the clamping bar in place.

The high-voltage transistor wasn’t just unscrewed; it had no thermal contact with the case at all. There was probably a gap of about a millimeter.

What are you supposed to do with this?



All right, enough of the sad stuff. Let's move on to the panels and how to connect them.
Since the panels are quite large, I'll demonstrate with two of them.
The panels are connected with the included ribbon cable. The connectors are keyed, so it's very hard to plug them in the wrong way.
Power is also connected using the included cable, with one cable for every two panels.
Connect the ribbon cable from the controller to the Data_IN connector. In this case, I'm using the long cable.
Additional panels are connected in exactly the same way.



Connect the power wires to the power supply. The cables come with fairly convenient fork terminals. But no wires for powering the controller were included, so I had to make do with what I had on hand.



Near the beginning of the review, I wrote that I couldn't find instructions online for connecting the board. There are plenty of videos and photos showing how to configure the software and calibrate the panels, but not a single photo of the cable connected to this controller model.
After checking the connectors with a multimeter, I figured out that they should be connected as shown in the photo. Presumably, monochrome panels connect to the black connectors.



The ribbon cable is a little long, probably designed for a thick chassis. You can fold up the excess as shown in the photo.



First power-up. A program is preloaded that displays the scrolling text AAAAA BBBBB. You can also enter test mode, which cycles through Red, Green, Blue, White, Off, and various stripe patterns.



And here are my first attempts with different programs, as well as displaying an image. Sorry about the shaky camera—I was filming with a camera without image stabilization, holding it in one hand while trying to plug in a flash drive with the other.
When you plug in the flash drive, the program is automatically loaded onto the controller. The current time at the moment the program was created is stored there too, so if you plug the flash drive in again an hour later, the clock will be an hour behind. In theory, this feature can be disabled in the program; I haven't looked into it yet.
The image didn't come out very well because of the glare; the colors are much more vivid in person.



There was also a minor annoyance.
One of the panels had an intermittent red dead pixel.



If you touch it with your finger, it lights up. It can also turn on of its own accord—in short, it has a mind of its own. I marked it with a marker, though it still turned out to be hard to find when it was off.



I shot this video for the seller to show what it looks like in real life.



That’s what prompted me to take a closer look at the LEDs.
Undo the nine screws (the center one is shorter) and remove the frame. The frame fits quite snugly, so the screws are probably just there as a precaution.



There are a lot of them. A whole lot :)



I found the dead LED, thinking I could simply reflow it.



But unfortunately, it’s soldered properly. Testing it showed that it goes out even when power is connected directly to its leads.
By the way,
the photo shows the common lead and the red lead.

Let’s keep playing with the panel—it really is a great toy :)
First, I tried out the different operating modes.



Then I remembered that we were looking for a controller with thermometer support. I found a Dallas 18B20 sensor in my stash and simply plugged it into the connector holes. I didn’t want to solder on a connector yet, especially since it might not work.



But everything went smoothly, and the controller detected the sensor right away.






As I mentioned, the panel can display images—in theory, even GIFs, though I couldn’t find a suitable one to test. Still, displaying static images works without any problems.
The image quality here is closer to what you see in person than in the video.



I also tried displaying a picture of my cat :) Since the matrix resolution isn’t very high, you need to choose the image resolution carefully so it’s clear what’s in the picture.
The photo shows two versions of the cat picture; I think the difference is obvious.

I also noticed a glitch here: if you display the temperature reading—or, more precisely, just enable this feature in the software—interference periodically appears in the image. It happens even when no sensor is connected, though a little more often with one. You can see it in the last photo as bright horizontal lines. It doesn’t happen every time, either, and I haven’t figured out what triggers it yet.



During the selection process, and even after the tests, my friend asked how much power a panel like this uses, since electricity is very expensive where it will be installed.
So I ran a number of power consumption tests. At first I told my friend that consumption depends heavily on what is displayed on the screen, but saying it is one thing and measuring it is another.
The photo shows the current draw on the 5 V rail. I measured it with a clamp meter because the voltage is low, and connecting a regular multimeter would have skewed the readings. All measurements were taken at 100% brightness with three panels connected.
1. Red
2. Green
3. Blue
4. White



1. Off
2. A little graphics from the built-in test
3. Text
4. Text + image covering about 40% of the screen.



Power drawn from the mains:
1. Red
2. Green
3. Blue
4. White
5. Off
6. A little graphics from the built-in test



I also checked the temperatures after running it for an hour.
On the right, an image and text were displayed in turn; the clock was shown continuously on the left, with the temperature in the middle.
The panels themselves don't get very hot, but where the image was displayed, you can clearly see the locations of the drivers that power the LEDs.

Finally, a photo of the power supply. I took it before I opened it up and discovered that the power transistor wasn't screwed down.



As I mentioned at the very beginning, the package included a CD with the software, but it was easier for me to download it from the official website.
The software is called LedshowTW 2016, and I'll give you a quick look at it now. I initially downloaded the version with GIF support, but it was kind of "half-Chinese"; then it got completely messed up, so I had to install another version. That one works fine and has a much better Russian translation.
So, download the software, install it, and launch it.



Before you start, though, you need to at least set the controller type and operating mode.
By default, the software is set to monochrome mode and a different controller.



Select full-color mode; the name of our controller, bx-5ql, will be there too.
This brings up the following settings window. It will complain that there's no connection to the controller; you can ignore that, it's normal.



After this, the startup window changes, and colored strips—a "border"—appear around the perimeter of the virtual screen. I later turned these strips off because they take up two of the 32 pixels.
The zoom in/out buttons let you increase or decrease the preview size.
At first, it worked as a preview, so you could take a look without writing the software to the controller, but then this feature stopped working :(



Next, click Image/Text to create a new item in the program.
Selecting "Subtitles" gives you a scrolling marquee; "Text" gives you a static line with switching effects.



Clicking the folder icon prompts you to open an image.



When you select an image, a preview appears with a rectangle showing our resolution and, accordingly, what will appear on the screen.
Move this rectangle around the image to choose what will be displayed on the panel.



Select Clock or Thermometer at the top to display the time or temperature, respectively. The temperature can be shown in different colors depending on the threshold value.



The software has quite a lot of different settings. I won't go into all of them here, since that would be difficult in a review; it's easier to find a tutorial video online.


Once the project is ready, you need to transfer it all to the controller. Since our controller only works with USB, we prepare a USB flash drive with the program.
It’s very simple: take a USB flash drive formatted as FAT16 or FAT32 and plug it into any USB port on your computer. In the software, select USB, then click Save.
The controller program is written to the flash drive as a boot file.
As soon as the software displays a notification at the bottom of the window saying the write was successful, disconnect the flash drive from the computer and connect it to the controller. Everything else happens automatically, as you could see in the earlier videos.



The software settings include a brightness control. Besides the regular one (16 levels), I also found a less obvious automatic setting.
In any case, after adjusting it, click “Close,” not “Install,” since “Install” means connecting the software directly to the controller, which will result in an error message.
By the way, I noticed that lowering the brightness also slows down the running text, by about 1.5–2 times, so keep that in mind.



I mentioned that you need a FAT32 flash drive, and as it happened, I had no fewer than five of them at home :)



If you think there’s nothing unusual about that, you’re wrong. I have five fake flash drives with capacities ranging from 11 to 46 MB. Nobody knew what to do with them, but now they’ve come in handy.



The thing is, programs are usually quite small, and images take up most of the space. For example, the program shown in the video is only 160 kB :)



I could stop here, but I’d like to tell you a little about how I made a large clock and then never finished it :(
I had a project in mind: I wanted to make a large clock. I even soldered two large circuit boards for it—after all, the clock was going to be big :)





Well, I think that's everything.
Pros
The kit works
More than enough functionality
You can build your own running text display without using a soldering iron
Low cost

Cons
The power supply is extremely poorly assembled
Interference when the thermometer is activated (I hope to fix this)
One dead red pixel :(

My take. First, the price. When I counted the LEDs on one panel and worked out what they cost, I was really surprised that the manufacturer managed to fit all this into $20:
1. 2048 RGB LEDs
2. Plastic chassis
3. Circuit board
4. Set of ICs
5. Set of wires
6. Mounting hardware
7. Shipping
And still make some money on it. I can't imagine how you could pull this off here; I think it's practically impossible. The panels are made in huge quantities, but still...

Overall, I liked the system: simple, attractive, and it works. Even someone with no experience can put it together, and the software is in Russian. What more could you ask for? :)

As always, feel free to ask questions in the comments.

Almost forgot: I downloaded the programs here; the review used this version.