29.07.2018 - Years later, 4 years in 16 seconds
Today the camera has died.
More precisely, i opened it up for the yearly battery replacement and photo download, and once it was put back together the error light kept lighting up.
Turns out the card in it was no longer writeable.
Gonna try to replace it some time this week, meanwhile i feel like it's the right time to post the footage collected so far.
Here is it a noon every day. The best part of it i how the trees poof in blossom suddenly.
|Timelapse: 4 years of apples, centered at noon
| Poof, blossom
And here is the same closer to evening. You can actually see how the sun's elevation change as the times of the year go by.
The jitter in the sun represents the jitter in the camera's onboard clock.
|Timelapse: 4 years of apples, sun motion
| Sun bobs up, sun bobs down
So, how was it performing?
The batteries tend to last about 9 months, after which it starts losing frames at night.
The clock on it is a bit erratic and have drifted almost a full day over 4 years. Fortunately the drift is consistent so i can correct for it in post processing.
The sensor shows it's age, being 640x480 in the days of 4K being the norm and HD being the expected minimum.
All in all, i'm facing with a hard choice: Add more batteries so it'll last 5 years, or add a better sensor as well, and have to keep replacing the batteries every year...?
08.11.2015 - Apple trees
Remember how i said "but at least the apples are going to grow"?
Well, guess which apple tree decides to die...
On the positive note, that camera was set up 246 days ago and is still running, so my longevity calculations seem to hold up so far.
But it's quickly getting colder now, so i don't think it will survive the winter.
In any case, i plan to run it to depletion (and post the video after that).
11.03.2015 - All coming together
First, the firmware.
It was uploaded quite a while ago, but i never posted about it.
Built with avr-gcc.
The bootloader is to be flashed first, then it would read the main firmware from a microsd card, if the file is there.
One thing i missed is a simple way to set the clock - you would have to uncomment the code for it, hardcode the time, update, comment it back and update again.
Second, the camera was fully assembled and installed.
The park build was finished too early to get anything interesting, but at least the apples are going to grow.
Not sure where to put the second camera now. I'll see what the first one would produce, in two months.
Anyway, now it's time to sit back, wait and see what happens.
| Fully assembled and sealed camera
23.12.2014 - Questions and assembled camera module
Lots of traffic lately, looks like it got on Hackaday. :)
I guess i'll answer the questions asked here and there, since there is little progress at the moment - i'm still thinking on the enclosure and testing the firmware.
-Will the firmware be available?
Certainly, once i make sure it's reliable. Right now it tends to make an empty file every few thousand shots on average.
I'm using FatFS module for the card - https://elm-chan.org/fsw/ff/00index_e.html, and the rest is trivial.
-Geez, put a solar cell on it already.
Ignoring locations where the sun only shines for half a year at any usable power, there is a question of cost.
For a solar setup to be practical, it have to have enough battery capacity to survive a night and several rainy days.
That means you will have to pay for the panel, pay for the batteries and pay for charge circuit, ending up with a lot of complexity, several times the cost and little real gain.
So running on batteries is neater and more practical.
Of course, you're free to add one in your version - it might be practical in different climates and scenarios like faster frame rates.
-There`s already a measure of power usage called "Ampere hours" perhaps you can convert all the joules references to it?
Assuming 4 D cells in series, the consumption would be 0.2 mAh per frame.
AH is a specific unit that makes many assumptions, like how many batteries are there and their average voltage.
Joules on the other hand measure actual energy, so it's more practical and less confusing in this case.
-What is the music in the Moscow video?
"Ether" from Youtube audio library.
The camera module assembled with the batteries is below.
All that's left is to make a box for it, and get them into the field.
I'm planning on two for the moment - one will observe an apple orchard growing, and the other observing a park under construction.
Later next year i hope to have a couple more locations - perhaps a construction project on beach resort area.
| The camera module, assembled with the batteries
19.12.2014 - Open-source one year time lapse camera
Sensor: 640x480 UART camera module.
Sample video: https://www.youtube.com/watch?v=aYzQma_5TN0.
RTC (DS1338) included, with CR2032 battery backup.
Stores images on a microSD card, which also holds the config file.
Power is supplied from 4.8+ V, over a 3.3V LDO, so 4 alkalines is ideal. I thought of using a switching regulator to increase efficiency, but it just isn't worth it on this scale - at best you can get 20% increase in run time.
One frame takes:
3 seconds at 100mA for the camera to stabilize and get a picture.
7 seconds at 60mA to write the picture down.
Between frames it stays in deep sleep, consuming 91 uA.
In other words, it uses 3.5-4 joules per picture.
On 4 D cells that would be around 50000 frames - at one frame every 10 minutes it will run for a year, at one per minute - for a month.
The intent is to set it and leave it be, coming back once a year - i like making long time lapse videos.
You can get the schematics and layout here:
Current firmware and MicroSD bootloader:
Appearance of the naked device itself below.
I'm still thinking on what sort of an enclosure to make for it. 3D printing is not going to cut it here. :)
15.12.2014 - Splitting the projects
The camera have been running for a day, and all seems to be well.
It made me wonder, whether the RF module is even needed.
All the settings are stored on an microSD card, and it is supposed to sit in some remote location.
At first all i wanted is an RF-accessible camera, which grown into a time lapse one.
Time to split the two.
Practical numbers on performance now are 3 seconds at 100mA for the camera to stabilize and get a picture, then 13 seconds at 60mA to write the picture down.
4 joules per picture, if we ignore the RF.
That is, 56Wh per year, or 4 D cells.
With an LDO, it doubles to 8, so i might add a buck converter to power the big loads.
The time lapse camera, as it looks now.
Several problems with it - RF module sticks out, the card is in an inaccessible place, camera is not lined up with the mounting holes, etc.
Time for a new board.
14.12.2014 - Low power camera
So, low power time lapse camera.
The sensor is a JPEG camera common on Ebay, modded for 3.3V.
It's 640x480, with 1/3" sensor - good performance at night compared to cell phone ones or OV7670.
Averaging 48Kb per frame at max quality.
Storage on an microSD card.
An RF module to get status or download an odd frame.
With some design changes from the RF one i can make a camera that does 1 frame every 10 minutes for just under a year, on 8 alkaline D cells.
The limit is actually the 2 GB microSD card (if i won't bother with FAT32 support).
Big issue was the camera module, which used up 50mA even in power save mode.
With a P-FET above it, i can do a frame&write to microSD for 15 joules, or 90 joules per hour.
Add 1 joule/hour to ping the radio every 10 minutes and round off, that's 100 joules per hour.
A D cell stores about 60000 joules, and i need 4 of them to get the voltage.
So, 2 strings of 4 will give me about 200 days of runtime, in pessimistic estimate.
With that in mind, i built a new board and tried it out.
It writes a picture to an SD card without problems.
Power, however, does not want to play nice - the theory and practice diverged a bit in what is being considered "sleep mode" by various parts.
Namely, the RF module does not want to play ball and go below 2mA, and microsd card sucks lots of current from MOSI line when it's VCC is out.
A redesign incoming, but the basic idea seem to work.
13.12.2014 - A radio controlled camera
An $20 JPEG camera, an $4 RF module, Atmega and some batteries.
Yesterday i've put them together into a little radio-controlled time lapse camera, and wanted to see how long the batteries will last.
So, it was left staring out of the window from 2:50 to 20:30 taking about 1 frame each 30 seconds.
The result happened to illustrate the mood of a winter day in Moscow.
|Lapsing through a snowy day in Moscow
| That`s actually a good stand replacement
On 4 AAs it runs for about 30 hours, on 4 AAAs - 12 hours.
This got me thinking - how about a camera for time-lapse photography, that would run from batteries for a few months (i.e. from a bunch of D cells)?
The idea is kind of to leave it be in a nice spot and come back next year, without worrying about getting power there.
This one seem to be quite inefficient on that front, having to transmit the frames, and thus keeping the camera module on all the time.