While looking at some LED facemask projects online, I came across a neat new flexible RGB LED matrix with BLE connectivity. It has 32×16 RGB pixels but the mobile app it uses will only send full-on primary color combinations, i.e. it’s not full color. Cost is ~$40.


  • Markings: [HC89F0541 | 2011 | A7616511]. Qty: 1. Holychip 8051 MCU, Link.
  • Markings: [SM16206S | BZBCNNAC17]. Qty: 6. Sunmoon 16 channel constant-current LED sink driver, Link.
  • Markings: [D7258 | 2009CF]. Qty: 2. Double Microelectronics 8 channel LED source driver, Link.
  • 2mm x 2mm RGB LED. Qty: 512.


4 wires go from the BLE/battery pack labeled VCC, TXD1, RXD1 and GND. I used my Saleae Logic16 to see what the TXD/RXD waveforms looked like, focusing on length of data bit periods. You can view the below logic traces in Logic2:

To save time I used a web calculator to convert from period to frequency. Seeing a falling edge before every packet and bit periods of 28µs -> 35714Hz I figured we had a UART at ~38400 baud. Setting an Async serial analyzer in Logic2 showed what looks to be meaningful data. I exported the incoming data (in hex):

RXD, however, is another story. The shortest bit period on RXD is 101µs, with others being 202µs and packetized with a ~5ms preamble of zeroes. Assigning an Async analyzer at 101µs -> 9900Hz -> ~9600 baud yielded garbage (framing errors everywhere). 10K baud also yielded nothing. Given this is a product in the LED space, I assigned a WS2812B analyzer in Logic2 but that also produced nothing useful. Weird, I’ll have to look closer.

There are (32 x 16 x 3) = 1536 LED die on the display. Since this display only drives each die fully ON or OFF, we should be able to represent each die with a single bit. I expect the display data for a single frame will then be (1536/8) = 192 bytes.

I entered enough text to cover 3 full frames, so I expect to see (192*3) = 576 bytes go by, ignoring protocol overhead.

TO BE CONTINUED… Feel free to drop any insight into the comments!


2 Replies to “Asian Flexible LED matrix”

  1. Thank you for sharing this material!

    Have you been able to dig deeper into this matrix and ways to control it without the dedicated app? So far I’ve only been able to see that it appears the memory and some logic reside on the matrix itself and the PCB seems to only control the charging and WiFi.
    When trying to control the matrix with an Arduino I connected the wires (I assumed P was data and K is for clock?) and initially, the matrix didn’t respond but when trying some FASTLED library examples something triggered it to think it was being charged and it displayed a red charging battery symbol on the matrix.
    I’m working on something that could very much use a small curved LED panel, so I’m hoping to get to the bottom of it. Also, being able to control something like that outside of the dedicated app seems pretty useful too!

    1. Hey Sunny,

      I haven’t revisited this device since posting the article, but it would make sense for a device like this to interact with other WS2812-based products. Maybe the TXD line protocol can be reverse-engineered by trying/finding patterns in the data, and maybe the RXD line protocol is simply a much-slowed WS2812-like protocol – the WS2812 ‘1’ bits are roughly twice as long as ‘0’ bits on the line.

      If you make any progress, please do get in touch!


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.