This page has some old data logging graphs I had on my website years ago.
I have some Dallas Semiconductor DS1820 Onewire temperature sensors (the
old ones, before they realised they had a design flaw and brought out the
DS18S20 instead). They're interfaced via the serial port to a PC (running DOS), and logged with software I got off the web.
The sensors are in a transistor-sized (TO92) package , have inbuilt analog to digital conversion, serial data output, and can work off parasite power. Although they're not wonderfully cheap, they do give digital temperature readings with virtually no effort - from this perspective, they're pretty good value for money.
Dallas provide a schematic for a serial port interface (providing power & comms), using 4 diodes & a resistor. This simple interface has some limitations on cable length etc, but it's not a problem for basic setups. Multiple sensors can be attached to the common buss and addressed individually.
The variety I have are calibrated to 0.5 degrees Centigrade, but the resolution is closer to a hundredth of a degree if you do an extra read of the internal registers. The high resolution makes them quite good for showing small temperature fluctuations, even if the absolute measurement is a little off. The temperature conversion (to digital) takes about 200ms.
The following plot shows the temperature at my work, over a 2.5 day period. I setup a sensor about a foot above the floor, one about halfway up a wall and the last one about a foot below the ceiling. Readings from each of the sensors were logged at 1 minute intervals. The log file (plain text of temperatures with timestamps) was imported into Excel to produce the graph. (this was all before I started using linux)
You can see when the sun hits the building in the morning & the temperature starts to rise. When I leave the building at night (I'm normally last to go) the temperature starts to fall, looking very much like a typical exponential decay. From 7pm to 8pm in the middle of the plot, I was out of the office and the temperature fell - when I arrived back, the ceiling temperature rose by most of a degree until I left. Presumably this rise is from a combination of:
- PC & monitor coming out of standby - maybe 200watts change.
- me - maybe 60w-100w (estimates from various places on the web, eg. here)
- fluourescent lights - maybe 45w of heating.
Next time I do some logging, I'd like to have an outside air temperature sensor as well.
My 1kW oil-column heater has a simple on-off type of thermostat, as is obvious in this graph. The red line is from a sensor clamped to one of the fins of the heater; the pink line is the air temperature in the room, about a metre from the heater; and the blue line is the temperature in an adjacent room, just to show that ambient temperature in the rest of the house is fairly constant throughout the test.
The heater temperature in this case is ranging over about 13 degrees C - assuming my thinking is straight, this is a combination of two things:
- the hysteresis in the thermostat.
- how much the heater manages to heat up before the temperature rise is noticed at the thermostat.
If the heater had a better thermostat system (eg. proportional control), then the heater would be on pretty much full power until about the 20 to 25 minute mark, then on reduced power, supplying just enough heat to make up for the heat which is leaking out of the room (through windows, doors, etc.)
What I'd like to test next time:
- run the test for longer, to see where the room temperature ends up.
- record the state of the thermostat, to see exactly when it turns on and off.
Heater Thermal Resistance
This test shows operation of the heater in a room which is too big for it to have much heating effect. The heater is on full power for the whole time and the room never warms enough for the thermostat to come into action.
From the graph, we can see that the heater is about 44 degrees (Celsius) hotter than the room temperature. Since we know that the heater is supplying 1kW of heat (nominally), we can calculate the thermal resistance of the heater, as being 44 degrees divided by 1kW = 0.044 degrees per watt.
This is a measure of the effectiveness of the heater in transferring heat to the air - a lower number is better. Making the heater more effective won't make the room any warmer in the long run, or save any power, but it will mean that the heater runs at a lower temperature, and therefore be safer to touch.
Things which could improve the effectiveness:
- have more fins on the heater (a greater surface area).
- paint the heater black instead of white (improved heat radiation characteristics - won't make much difference at these temperatures).
- maybe change the heater surface texture to increase surface area.
- change the shape of the heater fins to improve airflow.
I have no definite explanation for why the heater temperature fluctuates so much in the later part of the test. I suspect that the sensor may not have been firmly clamped to the fin.
I borrowed a basic thermal data logger from a friend and stuck it in the letterbox for 5 days, logging temperature every minute. Surely everybody wants to know how hot it gets in the letterbox? OK, I admit that it's not very useful.
The thick red line is the average temperature. Typical peak ambient temperature was probably about 18-20 degrees C. We can see that sunlight hits the letterbox about 8am at this time of year, and the temperature climbs rapidly. What I find most interesting is the temperature on the Monday - it starts climbing about 4am. The sun definitely wasn't up at that time of the morning, but presumably there was a warm breeze.
(The average line is derived from slightly more data points than are shown here, hence the slight discontinuity just before 5am)
Car Boot Temperature
This is probably more useful than logging the letterbox temperature, but not by much. Two probes were used in the car boot (trunk). One on the floor of the boot, just above where the exhaust pipe runs - that's the nice hot red line. The other probe (blue) was about mid-way up one side of the boot. The third line (green) is the car speed (logged with GPS) in units of 10kmh, to fit in neatly with the temperature scale - it's only really there to show when the car is running. The logger used has a maximum temperature of about 42 degrees C, so the red plot is clipped at its peak.
Unsurprisingly, the temperature above the exhaust climbs rapidly once the car is started, and starts falling rapidly about 20 minutes after the engine is stopped. The temperature peak on the blue line is partly from heat radiating from the floor of the boot, but is mainly from being parked in the sun.
The car is a 1990 Mazda MX5 (Miata / Eunos Roadster).
In 1999 I bought an OEM GPS module from a surplus dealer. The module was meant for car navigation, so it's somewhat appropriate that I've installed it in my car.
The GPS produces a serial datastream containing current latitude & longitude, as well as current speed and various other things, updated once per second. My logger hardware displays these variables and stores them to 32k EEPROM at configurable intervals. There's capacity for 1638 readings - about 27 minutes if logging at full rate. Most of the time, for recording my day-to-day driving, I have it set to log every 30 seconds, giving capacity of about 13.6 hours - one or two weeks of travel, depending on how much driving I'm doing.
This plot shows activity for a fortnight, mostly travelling to work and back, but also various social & shopping trips. What is it good for? Very little, but it looks arty.
GPS path - MX5, Snells Beach
This plot shows where we (MX5 club) went for a drive to Snells Beach for lunch. Not the most direct route from A to B, but that's not why we were doing it. Who wants to drive on a boring straight motorway when you can have nice twisty roads instead?
I was logging coordinates every 5 seconds, so the plot has quite good resolution - certainly good enough for display at this scale.