I am seeing a problem in the monitoring charts, at 10.5v on the battery, the charge controller switches the load off. Which seems to happen at about 2200-2230. I need to get a good indication of how long it's been sunny the last few days, so I can see if this is a big problem with the positioning of the panel or simply that we haven't had much direct sunlight, or if I need to get the daytime consumption lower (actually I will probably do this anyway)

The meter monitor software I am writing is going to be essential here, I must be able to measure the current into the battery

Well, it’s all done. I installed the main controller this evening, and took control from the house. I need to install latest firmware on the boards that were previously installed so that blending and broadcast work, and at some point automate the lighting and dousing into my HA set up, oh, and do the cap where the weather station is but apart from these few loose ends it’s all over bar the shouting

Blending is cool, I'm glad I decided to do that. It really makes it feel like a finished item. I am still toying with the idea of turning the node clocks down when the lights are off (back to minimizing power consumption). I run them flat out (32Mhz) to keep the PWM frequency up, but when the duty cycle is 0% it won't matter how fast it is. I can turn them back up when needed too. I'll have to see how the battery lasts, particularly when there is no sun.

The panel is mounted, and connected to the charge controller. The battery is in its final resting place, and connected to the charge controller. The dock wiring at the moment in not connected to the charge controller. The main controller can read voltages on both channels and is mounted on the bottom of its enclosure on my bench. The remaining cap lights are next on the list, and will be installed tomorrow.

Reading the last post about how I will need to split intensive calculations over more that a single cycle, I can see people wondering I made the choices I did with the availability of the Raspberry Pi Zero and ZeroW. Here's some explanation

Was thinking about the blending this morning, decided that I need to use fixed point numbers, essentially 8 bit integral, and 8 bit fractional. As I will be dividing by 64 I will only use the top 6 bits in the fractional byte so there will be no rounding error at the end :). As I work the delta back in over the following 64 cycles, I will not worry about rounding the fractional part into the integer, and I will do all these calculations pre gamma correction, and correct each step of the way.

Following up on this, as I sat and thought through the problem (different calibration slopes between positive and negative trending values) I figured it out. The filter on the ADC monitor was at fault, it was doing a divide by 16 with integer math, so as soon as you got within 16 of the target, it stopped approaching. Which means that the same voltage going up and coming down could be represented by a value up to
32 counts apart. I could have applied an adaptive filter so as you approach the target, the filtering gets less, but I have only just now thought of that.

Solar dock power monitoring is working kind of, I empirically derived the formula for the counts/V at the main controller, after adding in some lowpass filtering on the ADC processor, and there is a slight issue.

So I finally got fed up with the flaky state-machine so I decided it was time to hook the ESP to a FTDI device. I should have done this some time ago, as I found out some very interesting things...

After some thought about the situation, making the lights transition (or blend) from one colour to the next, It's a little like the mp3 player dimming the front panel when you turned the headlights on. Not really required but definitely one of those hugely underrated feature that takes the project from "cheesy amateur hack" to "looks pretty slick". The more I think about it, the more it needs to happen, I can change my PWM frequency so you won't notice the math at the end of the duty cycle.


Subscribe to RSS - blogs