Neural Networks (MNIST inference) on the “3-cent” Microcontroller

Bouyed by the surprisingly good performance of neural networks with quantization aware training on the CH32V003, I wondered how far this can be pushed. How much can we compress a neural network while still achieving good test accuracy on the MNIST dataset? When it comes to absolutely low-end microcontrollers, there is hardly a more compelling target than the Padauk 8-bit microcontrollers. These are microcontrollers optimized for the simplest and lowest cost applications there are. The smallest device of the portfolio, the PMS150C, sports 1024 13-bit word one-time-programmable memory and 64 bytes of ram, more than an order of magnitude smaller than the CH32V003. In addition, it has a proprieteray accumulator based 8-bit architecture, as opposed to a much more powerful RISC-V instruction set.

Is it possible to implement an MNIST inference engine, which can classify handwritten numbers, also on a PMS150C?

Decapsulating the CH32V203 Reveals a Separate Flash Die

The CH32V203 is a 32bit RISC-V microcontroller. In the produt portfolio of WCH it is the next step up from the CH32V003, sporting a much higher clock rate of 144 MHz and a more powerful RISC-V core with RV32IMAC instruction set architecture. The CH32V203 is also extremely affordable, starting at around 0.40 USD (>100 bracket), depending on configuration.

An interesting remark on twitter piqued my interest: Supposedly the listed flash memory size only refers to a fraction that can be accessed with zero waitstate, while the total flash size is even 224kb. The datasheet indeed has a footnote claiming the same. In addition, the RB variant offers the option to reconfigure between RAM and flash, which is rather odd, considering that writing to flash is usually much slower than to RAM.

Analysis of the Gen2 Addressable RGB LED Protocol

The WS2812 has been around for a decade and remains highly popular, alongside its numerous clones. The protocol and fundamental features of the device have only undergone minimal changes during that time.

However, during the last few years a new technology dubbed “Gen2 ARGB” emerged for use in RGB-Illumination for PC, which is backed by the biggest motherboard manufacturers in Taiwan. This extension to the WS2812 protocol allows connecting multiple strings in parallel to the same controller in addition to diagnostic read out of the LED string.

Not too much is known about the protocol and the supporting LED. However, recently some LEDs that support a subset of the Gen2 functionality became available as “SK6112”.

I finally got around summarizing the information I compiled during the last two years. You can find the full documentation on Github linked here.

Revisiting Candle Flicker-LEDs: Now with integrated Timer

Years ago I spent some time analyzing Candle-Flicker LEDs that contain an integrated circuit to mimic the flickering nature of real candles. Artificial candles have evolved quite a bit since then, now including magnetically actuated “flames”, an even better candle-emulation. However, at the low end, there are still simple candles with candle-flicker LEDs to emulate tea-lights.

I was recently tipped off to an upgraded variant that includes a timer that turns off the candle after it was active for 6h and turns it on again 18h later. E.g. when you turn it on at 7 pm on one day, it would stay active till 1 am and deactive itself until 7 pm on the next day. Seems quite useful, actually. The question is, how is it implemented? I bought a couple of these tea lights and took a closer look.

Controlling RGB LEDs With Only the Powerlines: Anatomy of a Christmas Light String

The RGB curtain predictably turns into a mess of wires when not used according to instructions.

As should be obvious from this blog, I am somewhat drawn to clever and minimalistic implementations of consumer electronics. Sometimes quite a bit of ingeniosity is going into making something “cheap”. The festive season is a boon to that, as we are bestowed with the latest innovation in animated RGB Christmas lights. I was obviously intrigued, when I learned from a comment on GitHub about a new type of RGB light chain that was controlled using only the power lines. I managed to score a similar product to analyze it.

The TransistorPixel

What would it take to build an addressable LED like the WS2812 (aka Neopixel) using only discrete transistors? Time for a small “1960 style logic meets modern application” technology fusion project.

Power Analysis: Probing WS2812 RGB LEDs

Power analysis is a technique to probe the inner workings of an integrated circuit by measuring changes in the supply current. Whenever a logic gate switches, it will cause a tiny current spike that can be measured externally. By inspecting the temporal variation, especially in reaction to an external signal, it is often possible to deduce information about the construction of the IC.

A few years ago I used a logic analyzer to investigate the protocol of the, then new, WS2812 RGB LED. So, why not revisit this topic to test my newly acquired deep sampling oscilloscope?

Intelligent 7-Segment Display

Time for another project with a super-low-cost microcontroller. But what to design? Ever since seeing a project where a $0.03 MCU controls $40 worth of intelligent RGB LEDs, I have been wondering whether this is the right place to use these devices. At this price point, doesn’t it seem to make more sense to dedicate one MCU to one LED each and use it to implement a fancy node-controller? It has always appealed to me to design my own protocol. However, just copying a WS2812 RGB LED or similar seemed to be a bit pointless…

The “terrible” 3 cent MCU – a short survey of sub $0.10 microcontrollers.

Like many others, I was quite amazed to learn about a microcontroller sold for only 0.03 USD via the EEVblog last year. How was this possible? Many assumed this was a fire sale of an old product. Digging a bit further, it became apparent that there is an entire market segment of ultra-low-cost microcontrollers. Almost all of them are products of rather unknown companies from China or Taiwan. This write up summarizes my findings in this rather peculiar niche.

We already learned that there is a large variety of very powerful $1.00 microcontrollers, but what about the $0.10 MCU? Are they indeed all “terrible”, as suggested elsewhere?

SK9822 – a clone of the APA102?

Two years ago I took a deeper look into the APA102. Although it was more expensive than the common WS2812, and harder to come by, it had some intriguing properties. The main benefits are a timing-insensitive  SPI interface, allowing easy interfacing to standard periphery, and a much higher PWM frequency of >19kHz, making the APA102 almost flicker free.

So much about that. Considering how things with LEDs from China go, it should not take too long for clones to appear? Indeed! Recently, several comments showed up on my blog, reporting about issues with APA102 LEDs they bought. It quickly turned out that these were SK9822, APA102 clones from the same company that already brought the SK6812 to us, a WS2812 clone.

One of these people was Mike. He developed the Weblight, a WebUSB controlled RGB LED. The prototype (shown below, red pcb) worked well, but when he commissioned a small production run (black pcb), the LED started to show odd update behavior. Mike was nice enough to share a couple of boards with me for further investigation.