I have far too many ideas to follow them all through. So I decided to take them to the proof-of-concept state, make sure they not just idle speculations, and list them here. Most of them are too big to make them alone, others are just waiting in the queue.
Maybe someone will be interested and help me. I offer to make them, or similar ones based on the same idea, for a reduced price.
CAN is a beautiful interface. It is simple, fast, and has a high noise immunity. It is no wonder that it is used extensively in cars and in big robots.
Unfortunately the cheapest CAN driver, the SIT1040T costs €0.20, which is far too much for price sensitive commercial products, small robots and toys. I addition to the CAN driver some microcontroller is needed to process the I/O signals, and to handle the communication. As a result, it is difficult to make a CAN-node for less than €0.5.
The Padauk PMS152 microcontroller costs only €0.04. It has a few high sink pins, which connected in parallel, can drive similar currents to a real CAN driver. It also has a comparator, which can be used similarily to the input comparator of a real CAN driver. The end result is not as good as a real CAN driver, but it can be connected to a CAN bus, and it costs only 1/10th of the price.
PCF8574 costs €0.34. It has a 100kHz I2C interface and gives us 8, not too spectacular, I/Os.
Padauk PMS152 microcontroller costs only €0.04. It can emulate, with some effort, a 100kHz I2C interface and could give us 8 high quality I/Os.
STM8S003F3P6 costs €0.16. It has a true, high speed I2C interface and could give us 8 or 10 high quality I/Os. And some additional bonus features, like non-volatile storage, firmware upgrade over I2C, etc. Using it with USART or SPI is just as easy.
Why on earth does anybody use PCF8574? And why can't we find any, reasonably priced, I/O expanders?
Use encrypted FW upgrade even on STM8S003F3P6, which costs only €0.16. The algorithm could be some sort of steam ciphers, preferably a variation of something that has been proven to be safe, like the ChaCha algorithm. More details are in my white paper.
It is not a stand-alone project, but part of a bigger project, and the foundation for a new, safe design environment.
Cooperative multithreading is possible even on the smallest and cheapest microcontrollers, like STM8S003F3P6, which costs only €0.16. My proof-of-concept code can do context switching in 8 cycles, or 0.5µs.
I have not tested the same concept on any Padauk microcontroller yet. I think it would work, and because most code for these processor do a lot of independent bit banging, it would simplify the integration of small, undelatad tasks.
Traditionally an I/O in a small robot, or in a household appliance, made of a few resistors, a few capacitors and a transistor or a single gate IC. Then the signal is routed to one of the pins of the main CPU.
The problems with this approach are:
Using the cheapest Padauk microcontroller the I/Os can be grouped, multiplexed, and filtered. One PMS152 can handle up to 12 I/Os, and thus the cost of processing one signal is actually lower than using even the simplest discrete I/O circuit.
Using multiple pins, or even multiple processors, to create redundancy, increases safety.
With two microcontrollers, checking each other all the time, it is possible to satisfy the requirements of IEC 60335 and ISO 13849.
Traditional dimmers use triacs and phase splitting, which does not work well with most elecronic loads today, and even if it works, it produces high switching noise. Quite often we find a big, bulky inductor to filter those switching transitions. It is neither nice, nor cheap.
Using FETs for switching, a microcontroller to control, and an advanced, high frequency PWM algorithm makes the undimmable dimmable. A filter inductor is still needed, but it can be smaller. An additional bonus is that the microcontroller can make the user interface smaller, cheaper and smarter. A capacitive slider can be used instead of a real potmeter, which simplifies safety isolation.
A simple and cheap soft start circuit module to be used in kitchen utensils.
These extremely price sensitive household appliances are often sold without a soft start circuit, because those circuits were too expensive. A really cheap circuit could make them better, and still competitively priced.
The use of solar panels and batteries will continue to increase in the next decade. It means the demand for efficient, and cheap power converters will increase too.
If you open a junction box today in your home you will probably see a bunch of cables there. Designing and routing such a wiring is a work for professionals. Modifying them is difficult, sometimes impossible, even for them.
What if there were just five wires and they would be connected in parallel to every outlet, every switch and every light socket? What if any switch, anywhere could control any light source anywhere?
Such systems do exist, but they are not cheap enough. The idea is to make the electronic switch cheaper than a mechanical switch, and the wiring simple enough for DIY installation.
And once the system is ready, we can make internet bridges, and RF433 bridges, to extend the control range even further.
RF 433MHz is the past. IoT is the future. Connect them simply and cheaply.
Household internet distribution networks are not built for reliablility. Sometimes they hang, and have to be restarted manually. But as the number of IoT devices increases, it will be increasingly important to have your home connected to the internet all the time, even if there is nobody there to reboot the router.
Making an IoT watchdog is little more than a hobby project. But, nicely packaged, it may have a market.
Make it look like an old product, but make it safe and dirt cheap. Use two wires only for power, and the same two wires for secure communication between the keypad and the lock control electronics.
A surprisingly large number of personal safes are sold with numeric code locks that are not safe at all. YouTube is full of videos showing how to open such safes in seconds. Let's make our safes safe!
GPS is good for global positioning in the open. But a cleaning robot, or lawn mowing robot, needs higher accuracy near the docking station. A drone needs higher accuracy for landing.
Use ultrasonic beacons near the critical area to help navigation. Ultrasonic guidance is cheaper than visual guidance, simpler than RF guidance, and more reliable than mechanical guidance.
A drop sensor is simple, usually optical, sensor at the perimeter of a robot helping it to detect sudden drops in its path.
Surprisingly, there is no cheap and universally accepted solution for this, seemingly simple problem yet.
I have experimented with different type of sensors and I think I know a solution. If it works, and can be properly packaged, it could be sold to every robot manufacturer, who do not want to spend much time designing their own. If such a sensor could fulfill the requirements of IEC 60335 and ISO 13849 it could have even a bigger market.
Quite often a robot has an outer shell, or a movable fender, and an inner shell. When the outer shell makes contact with an external object, the outer shell pushes mechanical switches built-in between the two shells.
This is an accepted and relaible, but comlicated solution. The mechanical switches have a finite lifetime, they are sensitive to environmental effects, like dust, water or vibration. Optical sensors are better, but more complicated and just as sensitive to dust.
A capacitive sensor could detect the change in electric field before the actual contact is made, or detect the slight deformation of the shell and would make the mechanical construction simpler. A capacitive sensor is also cheaper and less sensitive to environmental effects.
While I worked on the waterproof capacitive switch, TapTile, I discovered a sensor that was less suitable for wet environment, but ultra sensitive in dry environment. I think it may have some commercial application.