AnalogDataLogger

intro

The data logger is designed for the Pepito WindMill project to gather data about the windmill's performance. Althougth I have abandoned this project in favor of WindLcdMeter but I was quite pleased with the code I wrote for this project for a number or reasons. This was the 1st time is wrote to and read from EEPROM with an AVR (it's pretty trivial honestly but still) as well as the serial shell was pretty neat and it allowed the user to write ADC samples to EEPROM and later download the sample data. In addition the system offered 2 distinct execution paths. If a key is pressed within 2-3s of powering up the system then the user is taken into an interactive shell, otherwise the system goes into a deep sleep state and periodically wakes up to take a sample from the ADC and write it to EEPROM.

spec

ATmega32 RISC uC

• • too much processing power (using it anyway)
  • free via sample
  • built-in USART (rs232)
  • built-in 8 channel 10 bit ADC
  • built-in 1024 bytes EEPROM

Serial Shell

• The command shell is entered by rebooting the datalogger and pressing 's' within 5 seconds.
  • If this does not happen the datalogger will return to logging mode and start recording a new record.
  • Once done downloading/erasing or other simply disconnect the power, there is no need for a shutdown procedure

Shell Options

• d - download flash
  • download all records over the serial port
• e - erase flash
  • clear all records in the logger to make room for more
• c - configuration options
  • this feature has yet to be implemented. not sure what options even need to be set short of the sample rate which will be set through an external RC circuit.

• Secret Option
  • E - test eeprom writing and reading (used in development of application)
  • a - test analog sampling. An analog sample is taken and echoed to the terminal

Logging Mode

this mode almost entirely consists of sleeping to conserve power (we work hard to generate it so no point in wasting it). When a sample is to be taken the processor is woken up, the system time is incremented and then the sample is read and written to eeprom before the processor returns to sleep.

• wakeup
• increment system time
• take sample
• store sample in flash
• return to sleep

File System

each data point consists of a 10bit analog reading and an 8 bit 'timestamp'. ideally this means that only 18 bits are needed per sample. for ease 3 bytes will be used in rev1, 1 byte for the timestamp, and one word for the sample. this eliminates the need to develop a packing algorithm but wastes EEPROM as well.

• 1 byte - timestamp (abandoned, only data sample is recorded)
• 1 word - data sample

System Time

at power up the system time is set to 0. each time the processor wakes up to take a sample the clock is incremented. therefore the datalogger reports a relative time that could be seconds, minutes, or hours.

raw logs

the documentation will be partially formed from this:

mike: It seems to put out about 18volts doing a high wind, which I think is good for charging 12volt batteries, and makes me worry less about high-voltage issues that I thought I might face. As long as she doesn't throw a blade or over-charge my little battery (I don't have a charge controller or a dump-load yet) it should be good. Unfortunately it's right next to the road so if it did throw a blade it could be ugly (luckily not much traffic out here).

chris: to simplify things i will just use a more powerful uC with all the hardware that i need onboard (i was going to go from a smaller uC and external EEPROM but would rather deliver something sooner than never).

the data will be stored in onboard EEPROM and downloaded thru the serial port on your computer. the chip i'm using (free sample, hand built board - staying within the budget of the project) has 1024 bytes of EEPROM and i need 18 bits for each data sample (assuming 1 10bit voltage and a 'timestamp') so that means that:

(1024 bytes * 8 bits/byte) / (18 bits per sample * X samples/day) = Days of operation before download is needed

X	Days	Sample Frequency
24	18	hourly
12	37	every 2 hrs
8	56	every 3 hrs
4	113	every 6 hrs

the samples/hour should be able to be modified 'in the field' either thru the serial port or via an external switch. the device will be low power and probably be parasitic (powered by the windmill) so as to not need an external battery. it is expected to be placed on the battery terminals, not on the windmill tower (easier to get to. as valuable?) - the timer will start each time it is powered up and take samples every X hours.

would it be more effective to measure the voltage output by the windmill itself (above the blocking diode)?

chris: i'm making good progress on it but i'm starting to get into the things i've never done before (writing to EEPROM for one - that seems easy enough though). batterys do tend to leak badly when they are overcharged (at least caps do) so anything much over 13V will probably be lost. that being said the mill producing 14V means the battery will charge faster (and probably damaging it to boot).

measuring at the battery terminals would tell you how quickly the battery is charging/discharging (i.e. if the mill produces 14V for 5 min you may only see 5V on the battery). This is best measured slowly to give an overall trend (i.e. 1 sample every 2 hrs would allow for 1+ mo of data before it needed to be downloaded). measuring at the mill motor outputs would give you an instantaneous measurement of the voltage produced but totally depends on the wind at that moment (best to sample this in bursts - like every 5 sec for 40 minutes (gathering more data would will require more flash than what's in the uC i'm using).

it would be good to take current measurements too but will double the amount of data (thus half the samples can be collected). usually current is measured thru a very small resistor (like a 5W 1ohm - then V = I). it will be easiest to start off with just voltage sampling using internal EEPROM. if it proves useful it is easy enough (but not as easy) to interface to a seperate EEPROM and take a lot more data samples including current.

rev 1: 1 voltage data sample with pseudo timestamp rev 2: 2-3 voltage samples with real timestamp (1 at the motor, 1 at the battery, 1 for the calculated current. the 'real timestamp' would come from a real time clock if you want to gather that sort of data, otherwise it'll be the same relative timestamp)

it's a fun project. glad to be on it.

-- ChristopherPepe - 11 Apr 2006