The two shelves in the hallway and the interrupting door
A view from the underside. You can see the LED stripe and one sensor clearly. Note that the sensor is in fact not exactly positioned as described in the config file (it should be located exactly at the outer edge of the stripe. This bug is there for historic reasons and will at one point in time changed by moving the sensor closer to the edge).
Closeup of a sensor attached to a wooden square rod (covered in white tape) that holds also the LED stripe
The uninteresting looking box containing the Pi, the logic board and a off the shelf power supply to drive everything, including the attached LED stripes and the sensors.
The uninteresting looking box from the other side. I am using a centronics style connector to attach everything: Power for the LED stripes & sensors, SPI lines for the stripes and 16 signal lines for the sensors (The SPI goes to the ADCs which are in the case, not to the outside sensors)
I am using a Raspberry Pi 3b in my installation. This has plenty enough horsepower to drive the project. In idle mode (no lights burning/effect running, only the readout loop of the sensors is active) the CPU usage is around 1.7%. While the multiblob effect is running as configured in the default config file CPU usage goes up to around 4.3%.
The OS used is standard Rasbian, and - thanks to the fact that GO compiles to a static binary - there is not much else needed more than that. You have to enable the SPI interface on the Pi though, there is plenty of documentation on the net on other things you could do (like overclocking, which may be overkill. One thing I did was making sure that the CPU doesn't scale but instead keeps running at a steady frequency as it seems to help with getting the SPI timings right, but YMMV)
BIG WARNING if you don't feel competent to work with mains voltage - DON'T DO IT
FOLLOWING WHAT I DID OR TRYING YOURSELF YOU ARE DOING AT YOUR OWN RISK
While I may (or may not) have a little clue about current and voltage myself, I am in no means in the position to give anybody reliable help on what or what not to do when it comes to dangerous voltage and amp levels
The next picture shows the opened box. The big thing closest to the camera is the off the shelf power supply. It's 5V and enough amps to drive the LED segments attached safely (and yes, also the Pi itself and the sensors need their share of power).
Behind on the right is the Raspberry Pi 3b v1.2 used, and left next to it is the custom multiplexer logic board.
The next image show the connectors that are attached to the Pi (as far as it can be seen there. The one and lonely connection at the left has been added later when I realized I had to drive 2 LED stripes and needed to hack another device into the multiplexer)
This connection routes the SPI GPIO pins and the GPIO pins used for multiplexing the SPI port on the custom logic board (see below for more info)
This is a bloody rat's nest of wires here...
The block labeled "Ignore" is from an old experiment to detect the input of any attached sensor. The potentiometer and the small OpAmp where used as a trigger to set another GPIO pin high and - detecting this interrupt - the software would only then start to loop the sensors for data (and stop again after some time of inactivity). The idea was to put less stress on the Pi in the mostly idle time. After rewriting the whole thing in GO (coming from phyton...), CPU utilisation is so low anyway that I no longer implemented that.
The next picture shows the simplified schematic (things like Vcc, GND or e.g. the capacitors at the power inputs of the chips silently ignored...)
I know this is for sure not the most elegant way to do the
multiplexing. E.g. I see no good reason why I didn't do the CS
select on the two ADCs directly via a GPIO pin (instead of via two
OR gates) that can be set to LOW and ignore the CS pin from the
Pi completely (remember, CS needs to be pulled LOW to active the
SPI device). I can only assume this is for historical reasons - the
whole board is the result of me actually creating my first digital
circuit some years ago - and I simply don't remember anymore the full
process to arrive at the final stage.
Anyway - it seems to work. CLK is shared among all devices, MISO
between both ADCs. MOSI is connected to both ADCs, but gated through
the two AND gates to the LED segments connection, so only one of
them get's the input signal at a time (the ADCs will ignore it if CS
is high).
I am trying to learn KiCAD to produce some "real" board. I will make sure in this new version to support out of the box up to 4 LED stripes, but maybe only use one MCP3008 for up to 8 sensors (which would allow for an entry and exit sensor per segment in the case all four stripes are attached)
At best would be a board in the format of a Raspberry Pi HAT so it could just be put on top of the Pi, connected via the expansion headers and actually power the Pi via the 5V & GND GPIO pins.