Breadboard games 2020

With the Covid-19 pandemic ongoing it is not likely there will be any further Breadboard Games events this year. There was a small event in early March in Kevin St. public library with children from Singe Street primary school. On that day the attendees built a board very similar to the one we used in
St. Audoen’s school in 2017.

So what about the rest of the year? Well, a couple of events happened this year that made me think of a particular game: the passing of John Horton Conway and the pandemic so the game is of course “The game of life”. Lots has been written about this game and I will not repeat that here, instead I will focus on how this can be implemented using a low cost microcontroller (MCU), screen and breadboard. There will likely not be “event” based on this game but maybe a reader will be inspired to try it out for themselves.

game_of_life_breadboard

The schematic
game_of_life_schematic

This circuits makes use of components I had lying left over from other breadboard games projects. The MCU is the 30 cent (approx) STM32F030F4P6. This display is a 176×220 TFT board with a parallel interface. The databus runs between PA0-PA7 on the STM32 and DB0-DB7 on the display (it is not shown completely as it would have cluttered the schematic).
The specifications of the MCU are pretty meagre: 16kB of Flash, 4kB of RAM, 48MHz clock speed. All pins have been used to control the display.

The game
Conway’s game of life is all about automata (cells) that live or die depending upon what surrounds them. Cells can die of loneliness (less that 2 neighbours), starvation (more than 3 neighbours) and they can come to life if they are bordered by exactly 3 neighbours. These rules are applied repeatedly to the cells in the landscape. I’ve chosen to do this in two passes. The first pass decides what the state of each cell will be during the next time around based on current state. The second pass switches from the old state to the new state. This of course requires memory (RAM) and this MCU only has 4kB. The display has 176×220 = 38720 pixels. The game allocates one pixel per cell so without some sort of compression there is no way this MCU can store this amount of data. To further add to the storage burden I’ve decided to go with a coloured version of GoL called Quad-Life. Each pixel can be one of 4 colours or black. Where can this be stored?

Stealing memory from the Display
The RM68130 display has enough on-board RAM to store 18bits for each of the 176×220 pixels. All of this does not have to be used for controlling the display however. If the display is switched to 8 colour mode then only 3 of those 18 bits is required per pixel (1 for red, 1 for blue and 1 for green). The remaining 15 bits can be used to hold whatever you like without impacting on what you see on screen. At least that was the theory…. I’ve been using various small displays like this for a while and up until now have had little success reading data back from them. The SPI versions are particularly tricky; some don’t even have read-back functionality. Happily the RM68130 does allow you read back from it’s graphics memory but there’s a catch: the bits you write do not come back to you in the same order. Now I’m willing to accept I may have made a mistake here but I found it VERY confusing reading data back from the display. After many hours I figured out enough to store 1 byte of my choosing alongside each pixel. This means the MCU just got an additional 38720 bytes of RAM – plenty for storing the future states of each cell. By the way there are other displays that should work like this. I’ve tried the ST7775 and the ILI9225 (both parallel interfaces) and they both worked fine

Additional game details

GoL has to start somewhere. You need some way of supplying initial states for the cells (pixels). As you can see above there are no more pins available for user input (PA13 and PA14 are used for debugging). One further feature of the display is that also functions as a resistive touch screen. The user now simply draws the initial state using a stylus (or finger).

The GoL landscape is not meant to run in to a hard border so when a cell goes beyond an edge it “wraps” around to the other side.

Gameplay
The video below shows the game “in action”. It is not particularly fast and manages about 1 update per second. Consider it a slightly interactive desk ornament 🙂

The code
Code is as usual available over here on Github. There has be no attempt to optimize the performance of this code in any way. It is shown in what is hopefully the easiest to understand mode.

The STM32G030

The STM32G030 is similar in many ways to the STM32F030. It has approximately the same peripherals and is available for a similar price (in western countries). There are some significant differences however:
The ‘G030 however uses a Cortex M0+ (instead of the M0 in the F030),
The ‘G030 has a maximum clock speed of 64MHz (48MHz for the F030),
The model I chose STM32G030K8T6 has twice the Flash (64KB) and twice the RAM (8KB) that the ‘F030 has.
stm32g030_display
Full size image
I’ve put together a number of examples over on github that might help someone get started. These include a simple blinky, serial comms, serial comms with an ADC, systick interrupts and finally driving an ST7789 LCD display.

Performance improvement for STM32F030/ST7789 graphics library

stm32f030_st7789_nrf24l01

I’ve been working on a new project involving an STM32F030, an ST7789 display and an NRF24L01 radio link. As part of this project I took a good look at the graphics library that I used in the Dublin Maker badge in 2019. It turns out that there was plenty of scope to improve it’s performance. Tweaks included flattening function calls and using the set/reset registers in the STM32F030. Here’s an excerpt from the old library:

void display::fillRectangle(uint16_t x, uint16_t y, uint16_t width, uint16_t height, uint16_t Colour)
{
    openAperture(x, y, x + width - 1, y + height - 1);
    for (y = 0; y < height; y++)
    {
        for (x = 0; x < width; x++)
        {
            writeData16(Colour);
        }
    }
}
void display::RSLow()
{
    GPIOB->ODR &= ~(1 << 1); // drive D/C pin low
}
void display::RSHigh()
{ 
    GPIOB->ODR |= (1 << 1); // drive D/C pin high
}

The new version of these functions looks like this:

void display::fillRectangle(uint16_t x, uint16_t y, uint16_t width, uint16_t height, uint16_t Colour)
{
    
    register uint32_t pixelcount = height * width;
    uint16_t LowerY = height+y;
    if ((LowerY) <= VIRTUAL_SCREEN_HEIGHT) 
    {
        openAperture(x, y, x + width - 1, y + height - 1);
        RSHigh();
        while(pixelcount--)
            transferSPI16(Colour);
    }
    else
    {
        // Drawing a box beyond the extents of the virtual screen.  
        // Need to wrap this around to the start of the screen.
        uint16_t LowerHeight = (VIRTUAL_SCREEN_HEIGHT-y);
        uint16_t UpperHeight = height - LowerHeight;
        openAperture(x, y, x + width - 1, VIRTUAL_SCREEN_HEIGHT-1);
        RSHigh();
        pixelcount = LowerHeight * width;
        while(pixelcount--)
            transferSPI16(Colour);
      
        openAperture(x, 0,x + width - 1, UpperHeight);
        RSHigh();
        pixelcount = UpperHeight * width;
        while(pixelcount--)
                transferSPI16(Colour);
        
    }
}
void display::RSLow()
{ 
// Using Set/Reset register here as this needs to be as fast as possible   
    GPIOB->BSRR = ((1 << 1) << 16); // drive D/C pin low
}
void display::RSHigh()
{ 
// Using Set/Reset register here as this needs to be as fast as possible     
    GPIOB->BSRR = ((1 << 1)); // drive D/C pin high
}

The new version is a good deal bigger for a couple of reasons:
First of all, the fill rectangle function has been extended so that it is usable with display scrolling (a new feature)
Secondly, the call to writeData16 has been eliminated (removing the function call overhead). This means that lower level SPI function calls have to be used. Also, the nested loop for x and y co-ordinates has been changed to a single loop that fires out the pixels as a continuous stream – the display hardware itself looks after the x and y coordinates.

So how much faster is it? To test this I wrote a simple program to fire a full filled rectangle at the display 50 times and measured how long it took.
The results:
The old driver :
50 rectangles (240*240) took 8.6 seconds. This corresponds to a pixel write speed of 334883 pixels per second.
The new driver:
50 rectangles (240*240) took 4.6 seconds or 626086 pixels per second. Nearly twice as fast as the older library. At this speed it takes 92 milliseconds to fill the display. Not stellar by PC standards but good enough for my needs.
Code is available over on github.

NRF905 Revisited

I have worked with the NRF905 radio transceiver in the past as part of a weather station. At the time I wasn’t entirely happy about the code for the radio as it felt a little hacky. Like many others before me I vowed “This time I’m going to do it right” 🙂 so I have set about rewriting the radio part in a more object oriented way. I have put together a transceiver pair as shown below and I intend working on the code to simply the NRF905 application interface as much as I can. Right now there is little abstraction going on – that needs to change.
The transceiver pair consists of a couple of NRF905’s wired to some STM32F030’s mounted on breakout boards. These are wire-wrapped together and a full schematic will eventually be posted along with the code over on github


Update 19-Nov-2019
I have updated the code on github with a tighter interface to the nrf905 class. I also did an initial range test and was getting error free line of sight range of about 200m @ 10dBm. The transmitter was on top of a car and the receiver was at about 1m above the ground. Transmitter and receiver had 1/4 wave monopole antennae. A longer range is possible if you are not concerned about the odd missing packet.

The STM32F030 driving an ST7789 display

The ST7789 display used in this example is a 1.3 inch 240×240 colour LCD display. It interfaces to the STM32F030 at the pretty fast speed of 24MHz. The display is connected as follows:
stm32f030_st7789_schematic
My intention with this circuit is to create a low cost conference badge (actually for Dublin Maker 2019). The target bill-of-materials cost is €5 – most of which is attributed to the display.
stm32f030_st7789
Full size image
Connecting the hardware is only part of the story however and a software library is also required. This was based on the Breadboard Games code from last August in Wexford. The main changes that had to be addressed were the new initialization code and the openAperture code. These were based on the Adafruit library code and the manufacturer’s datasheet. This library includes some graphic drawing primitives as well as other functions to support game play (e.g. timing and random number generation). Additional functions will be added as the badge develops.
Code is available on github
A video of the library in operation is available on YouTube.

Weather station part 3: putting it all together

Introduction

The previous two posts on this topic dealt with the BMP180 pressure/temperature sensor and the NRF905 radio transceiver separately. This post brings it all together and shows how energy consumption was reduced to prolong battery life. The original plan was to run this from two solar powered garden lights. These lights come with coin cell size NiMh batteries and a simple charging circuit. After a bit of fiddling I decided to put this to one side having concluded that the internal resistance of the coin cells was quite high resulting in a large voltage dip during transmission. Furthermore the capacity of the batteries is so small (40mAh printed on the case) and the quality so poor that they were unable to power the station through the night. For the moment I have switched to a pair of rechargeable 1300mAh AA batteries that are charged externally. I may revisit the solar power option later and replace the tiny NiMh batteries with some decent AA or AAA ones.

Reducing power consumption

The three main power consuming elements in the weather station are:
(1) BMP180 pressure/temperature sensor module
(2) NRF905 radio transceiver
(3) STM32F030 MCU

According to the datasheet, the BMP180 has an idle current of 5uA and an active current of up to 650uA. This device is mounted on a breakout board which has a voltage regulator (XC6206) with a quiescent current consumption of 3uA. A MHC5983 magnetometer is also on the board with a quiescent current of 2uA and an active current of 100uA. The total quiescent for this board should be the sum of these : around 10uA. Only the BMP180 is used so the peak current should get to around 650uA.
The NRF905 is powered down when not in use. The datashseet states that its current consumption should get down to 2.5uA when powered down, 32uA when in standby and 20mA when transmitting at +6dBm.
The STM32F030 is put in to deep sleep mode between transmissions. The LSI clock and RTC are left running and periodically wake the system up. It is a little difficult to determine what the current consumption should be in this mode. According to the datasheet, the current consumption should be between 2 and 5 uA. The power consumption of the RTC is of the order of 1uA according to ST literature. During active mode at 8MHz with code executing from flash, the current consumption should be around 4mA.
Adding all of the standby currents up we get to about 10 + 2.5 + 5 + 1 = 18.5uA.
Active mode current flow should be dominated by the STM32F030 and the NRF905. I would expect active mode currents of 5mA with no radio transmission and 25mA during transmission.
The main program loop is as follows:

while(1) {            
    Int2String(readTemperature(),&Msg[0]);    
    Msg[10]=',';
    Int2String(readPressure(),&Msg[11]);       
    TxPacket(Msg,0x20);
    low_power_mode(); // Sleep and wait for RTC interrupt
    resume_from_low_power();        
}

This transmits a packet with the following format:
0000000233,0000099804
This is interpreted as a temperature of 23.3 C and a pressure of 998.04 milibars
The RTC is configured to wake system once per minute (the weather does not change that quickly)
Low power mode is entered and exited with the following two functions:

void low_power_mode()
{					
    PwrLow(); // Put NRF905 into low power mode
    // Turn off GPIO B,A and F		
    RCC_AHBENR &= ~(BIT17+BIT18+BIT22);			
    RCC_APB1ENR &= ~BIT21;   // Turn off clock for I2C1
    RCC_APB2ENR &= ~BIT12;   // turn off SPI1 	
    RCC_CFGR |= 0xf0; // drop bus speed by a factor of 512
    cpu_sleep();      // stop cpu
}
void resume_from_low_power()
{	
    RCC_CFGR &= ~0xf0; // speed up to 8MHz		
    // Turn on GPIO B,A and F
    RCC_AHBENR |= BIT18+BIT17+BIT22;
    RCC_APB1ENR |= BIT21;   // Turn on clock for I2C1
    RCC_APB2ENR |= BIT12;   // turn on SPI1 	
    PwrHigh(); // power up the radio
}

Entering low power mode, the code shuts down the NRF905, disables clocks to the I/O port, I2C and SPI peripherals, slows the peripheral bus clock and stops the CPU. Resuming from low power reverses these steps. The CPU can be placed in shallow or deep sleep modes depending upon the setting of bit 2 in the system control register. The code below enables deep sleep mode which stops the CPU clock and powers down flash memory.

    SCR |= BIT2;  // enable Deep Sleep mode

On receipt of an RTC interrupt, the CPU exits sleep mode and resumes execution.

Results

CurrentConsumption

The measured quiescent current was a lot higher than expected; around 170uA @ 3.6V and 90uA at 2.4V. Why is this so much higher? Perhaps passive components on the breakout boards such as pull-down or pull-up resistors are raising the levels. All of the peripherals on the STM32F030 may not be powered off. A current drain of 90uA at 2.4V should allow the station to idle on 1300mAh for more than a year and I felt that this was probably good enough for now. The actual run time is likely to be a lot lower than this due to self discharge in the battery and the energy used during the brief transmission interval. I’ll see how it goes with an extended test.
The active mode current is lower than expected. The MCU current just before transmission is about 3mA. It would seem that the radio is only drawing (19-3) = 16mA during transmission. The datasheet suggests that this level of current should be typically 20mA. During experimentation I found that the current levels during transmission were consistently less than suggested by the datasheet – maybe the datasheet typical values are a little overstated.

Looking at the graph, the 5mA (max) region consists of three approx equal length periods: I2C read of temperature, I2C read of pressure and SPI transfer to NRF905. The NRF905 data packet consists of a preamble (10 bits), address (32 bits), data packet (32 bytes) and a CRC field (16 bits). In total this is 538 bits. The data rate if 50kbps which means the transmission time should be 10.76ms. This agrees with the graph.

Debugging issues

I developed this code using my ST-Link V2 debugger salvaged from a Nucleo board and reflashed with J-Link firmware. OpenOCD and GDB allowed me to debug and download code. A problem arises when the CPU is in a deep sleep : OpenOCD can not talk to it in this mode. It complains and times-out repeatedly. Occasionally, it catches the CPU when its awake allowing you halt it and to download code. If the sleep interval is long (e.g. a minute or more) this can be quite frustrating. There is a way around this however. If you pull the Boot0 pin high and power cycle the MCU it will boot to ISP mode. Openocd & GDB can happily attach to this allowing you to download fresh code. When download is complete, set Boot0 to zero and power cycle once more to run your code. You don’t need to stop OpenOCD or GDB when you do this power cycle (I just pulled a wire from the breadboard and put it back again)

Radio range

The NRF905 radio modules were supplied with some short antennas which seemed too short for 433MHz. I found that replacing them with quarter wave length wires slightly improved signal range. I settled on a power output level of 6dBm as it gave me an (urban) range of about 100m – enough for my needs. The radio module seemed to misbehave at the higher power level of +10dBm. This could be as a result of the breakout board or perhaps poor supply regulation. Running it a +6dBm produced stable results.

Code download

You can download the code for the weather station over here on github: https://github.com/fduignan/Weatherstation1

Weather station part 2: The radio link

Introduction

As mentioned in the previous post I’m hoping to put together a weather monitoring system. It will consist of an outstation in the garden which will be solar powered. The outstation will send data over a radio link to a base station located indoors. This post outlines the radio link code and setup.

The outstation radio wiring

stm32f030cct1
The figure above shows the connections between the outstation MCU, the BMP180 pressure/temperature sensor and the NRF905 radio module. There are lots of connections to the radio module and I’m not sure they are all strictly necessary at the moment but for now they stay.
The NRF905 can transmit at a number of different frequencies. After some trial and error I decided to go with a frequency in the 433 ISM band as it seemed to work best (the NRF905 is on a breakout board which already has inductors and capacitors in the antenna circuit which I suspect were tuned for this band). The actual module is supplied from dx.com (link). Overall I found this module much easier to use than the NRF24L01.
The code for the radio link is available on github over here. This code transmits an counter value once per second at maximum power (10mW). This current consumption while transmittiing is quite high (about 29mA) but this current burst is pretty short lived. Early indications are that the range is sufficient for my purposes though I may fiddle with the antenna later to see how far it can be pushed.
A (messy) prototype is shown in the photo below
Stm32f030Bmp180NRF905
You may notice the ST-Link debugger salvaged from a Nucleo board attached to the target. This debugger was reflashed with the Segger/J-Link firmware and behaved very well.

The base station

The base station consists of an STM32L011 Nucleo board attached to an NRF905 module. A photo is shown below (wiring details in the code only for the moment)
stm32l011nrf905
The Nucleo outputs any data it receives over its built-in USB/Serial converter. This is displayed on the host PC.
One problem arose with the STM32L011’s SPI interface which is covered in its errata sheet. There is a timing problem with the SPI pins and ST provide a simple fix for this (which I found out about after two days of head scratching :). Code for the base station is available here

Next post

The next post will deal with the issue of power consumption on the outstation. This needs to be kept as low as possible if the solar power cells I salvaged from cheap garden LED lights are to work out.

Weather station part 1: Pressure and temperature

Introduction

I’ve decided to try to put together a solar powered wireless weather station this summer. It will consist of two parts: an outstation comprising of a pressure/temperature sensor attached to an STM32F030 MCU; all powered by a cheap solar panel and a rechargeable battery. The MCU will use an NRF905 radio module to send data back to a base station within the house which will display it on screen. Power consumption in the outstation will have to be kept to a minimum.
The base station will consist of an STM32L011 Nucleo board and an NRF905 module. It will interface to the host PC using the built-in UART.

The pressure/temperature sensor

About a year ago I bought a BMP180+HMC5983 module from dx.com. The BMP180 component is a very sensitive atmospheric pressure and temperature sensor with an I2C interface. The HMC5983 is a digital magnetometer may have a future role to play in the weather station.

The outstation MCU

I have previously posted a description of how to mount the STM32F030 TSSOP-20 MCU on a breadboard friendly breakout board. These MCU’s can be configured to consume very little energy. I had several of these lying around so the seemed the logical choice.

Wiring

stm32f030_bmp180
The image above shows the wiring involved. The bmp180 module has built-in pull-up resistors so its quite simple. The code on the other hand is quite complicated. The BMP180 has a very strange conversion routine that takes the sensor outputs and converts them to pressure (in Pascals) and temperature (in degrees C x 10). The measurement results are output via the UART TX pin on PA9. I connected this to the host PC using a USB/Serial converter.
The code can be found here. It includes a lot of debugging code that was useful during development. This is very much version 0.1 as it does not take power consumption into account in any way. That will be the subject of a future post.

A very low cost STM32F030 dev board

Aliexpress have begun shipping a low cost breakout/development board for the STM32F030. The board can be programmed using ISP and a USB/UART interface as shown below. The boards cost varies but I got mine for 1.58 Euro.
stm32f030board
I had previously worked on some suitable examples which can be found over here.
These examples are built using a Makefile however I have found that the following script is a lot easier to work with (just make sure your PATH environment variable includes the directory where the arm compiler programs and utilities are stored).

arm-none-eabi-gcc -static -mthumb -g -mcpu=cortex-m0 *.c -T linker_script.ld -o main.elf -nostartfiles 
arm-none-eabi-objcopy -g -O binary main.elf main.bin
arm-none-eabi-objcopy -g -O ihex main.elf main.hex

You then program the chip by linking the Boot0 pin and 3v3 with the jumper, hit reset and enter the following:

stm32flash -w main.hex /dev/ttyUSB0

To run your program, move the jumper so that it links Boot0 to GND and hit reset.
The USB/UART interface appeared as /dev/ttyUSB0 on my system, yours may vary (on Windows it will be something like COM3)
stm32flash can be downloaded from a number of sources. On Ubuntu it can be installed with

sudo apt-get install stm32flash

The ARM cross compiler suite can be downloaded from launchpad.net

Aliexpress link to the board.