Versions
We developed multiple iterations until we reached the current version of Fencyboy. This page provides some useful information on previous versions.
Last updated
We developed multiple iterations until we reached the current version of Fencyboy. This page provides some useful information on previous versions.
Last updated
The v1.3.0 improves on some bad things of the v1.2.0 version. Most importantly, it solved the voltage offset error between several Fencyboys. This defect caused multiple Fencyboys to measure highly different voltages at the same position of the fence. This improved version comes with improved peak voltage detection circuitry, making the measurement more accurate. Unfortunately, the new components also consume a lot more energy. This is why we introduced the Smart Sleep functionality to limit power consumption.
The v1.2.0 Version was the first version we made several prototypes of. Some of them with a single-use battery and others with a rechargeable battery and a solar panel. We're still running some of these in live environments. For example, the one providing the public dashboard is also a v1.2.0 prototype.
The following core components were used:
Heltec CubeCell Module Plus as the main MCU and LoRa Radio. We used the Plus version because it offers more GPIO pins. We're pretty happy with this chip in total. However, we had one bad chip that just would not join the LoRaWAN network (probably caused by a hardware fault) and two other chips that had absurdly high power consumption in low-power deep sleep.
Microchip MCP3461 16-Bit ADC as the main ADC to measure the fence voltage. We downgraded from the MCP3561 with 24-Bit because of supply issues and the fact, that we don't really need the additional precision. However, we made sure that it is possible to easily go back to the MCP3561.
Texas Instruments LM2904 Op-Amp for the peak voltage detection circuit. We picked this one primarily for its low quiescent current per amplifier of just 90 μA which made it possible to leave them running in deep sleep. Additionally, the slew rate of 1.5 V/μs, the input offset voltage of ±1 mV, and short-circuit current of ±40 mA were aspects we looked for.
Texas Instruments LMV551 Op-Amp for the impulse trigger circuit. Similarly to the previous op-amp, we looked for a low quiescent current per amplifier. This chip has a very low consumption of just 37 μA allowing us to leave it running in deep sleep. This is very important because we need it to detect incoming impulses at any time.
Although we spend hours developing this version, there are some known problems:
Static errors between different builds in the peak voltage measurement. We observe significant measurement differences between builds of this version if they are connected to the very same fence at the same position. One Fencyboy may constantly show 5200V while another one constantly shows 5400V. Changes are observed by all of them in a similar way but the static offset error between these builds persists. At first, we thought this was caused by the huge voltage divider and tolerances of the used resistors. But even after measuring the resistance values, with high-quality equipment, and calculating the theoretical error caused by them the situation did not really improve. At the moment we believe this is caused by the LM2904 OpAmp and the voltage sampling capacitor.
The first PCB-based version was a failure, unfortunately. Something went wrong ordering the PCBs from our manufacturer and the GND plane was not manufactured. We had to throw away these PCBs.
