EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: V_King on August 26, 2013, 11:27:41 am
-
Hello,
bought D18S20 temperature sensor recently and was very disappointed by the response time. It takes at least a few minutes to reach the new temperature which is way too long for me.
I am looking for a temperature sensor, which could react within seconds to the new temperature, as I am trying to monitor some equipment, which gets hot rapidly.
I looked in a whole bunch of temp sensors in farnell catalogue, from most expensive ones to cheapest ones. Not much information is provided about reaction times. The ones which provide anything, show anything of 3 to 15 min response in the air to reach only 60-70% reading of the new temp. As I understand those sensors are just mainly used to monitor ambient temperature, where response times are low.
Can anyone recommend any temperature sensors with greater response times and reasonable cost?
Cheers
-
I would use a thermocouple (type K), but it depends on what resolution/insulation you need. Also, are you making a one-off, or going into production?
Also, what response do you need?
-
I agree, a thermocouple is fairly fast.
You could also try a small thermistor in a small package but you have to deal with self heating.
-
For a fast response you need a small mass, and the fastest are:
a) thermocouple : a naked junction one or an eyelet type (you must have a second sensor for the "cold joint" compensation)
b) very small Pt100 RTD sensor
c) small silicon diode (1N4148)
the sensor must be tightly coupled with object to be measured (both mechanically and thermally: grease).
Tightly coupling means low electrical insulation. I hope that this is not a problem..
Best regards
-
What equipment are you monitoring? There may be other ways to do it, for example, if it's just a copper winding, you could monitor the resistance (positive temperature coefficient, a semiconductor junction will have a negative temperature coefficient.
-
Thanks for the replies.
The sensor is one of, for the small individual research project in university.
I am monitoring a small oil reservoir temperature, which is attached to the damper (sensor is glued into the casing with low thermal resistance epoxy). Damper is being tortured by a testing equipment. Depending on the settings, the temperature can shoot from ambient to 80-90degreesC in less than ten seconds. So the temp range is pretty small 0-100degreesC should cover 99% of testing easily. And +/-1 degree resolution should be more than enough.
I chose the in chip sensor due to convenience, and ease of use. I watched Dave's one of the videos about thermocouples and it looks like it is very complicated to measure temperature (reference tables, cold junctions, conditioning circuit - I am more of a digital fan :) )
Would I be able to use the D18S20 as a cold junction reference?
I am also connecting the sensor to the arduino and have an extra 16bit ADC with one spare channel, so got some extra capability for measuring voltages if need arises.
One more thing to consider, the damper is connected to the massive linear motor (electromagnetic actuator). Would a thermocouple work as aerial and pick up all the noise?
EDIT. ciccio, your PT100 RTD sensor recommendation at first glance looks very promising. IST PT100 RTD sensors are relatively reasonable cost (3-7 quid) and response times are much better. Are there any catches with their use?
Thanks
-
a thermocouple will be your best bet for that kind of response time, you buffer and scale the input into something usable with a single op amp, capture with an ADC and bam digital, ,
for the readings your only likely to be interested in 20-150C so its a small look up table, as for cold junction, you take the thermocouple voltage, minus the thermocouple voltage for the current ambient and then match for the measured temperature,
-
I suggest finding a way to place the temperature sensor inside the oil, regardless of what kind of sensor you decide to use. As it is you are measuring the outside of the reservoir and I'll bet the differential temperature of the reservoir wall is giving you trouble. I had the same problem with a DS18B20 trying to measure the water temperature in a hot water boiler and I solved it by embedding the DS18B20 in an immersion well (http://www.grainger.com/Grainger/immersion-wells/ecatalog/N-m2x?fromViewAll=true&Pid=search) so that it is surrounded by the water it is trying to measure.
-
Just use IR sensor. Response time will be about 40ms and price of sensor is about $20-40. There are sensors with I2C interface if you want to hook it up to microcontroller directly.
This one from Molexis may work: http://www.melexis.com/Assets/IR-sensor-thermometer-MLX90614-Datasheet-5152.aspx (http://www.melexis.com/Assets/IR-sensor-thermometer-MLX90614-Datasheet-5152.aspx)
-
Don't glue the sensor to the chamber. You are interested in the oil temperature not the chamber. Use a very fine thermocouple of wire
or foil less than 0.005" thick. Place the thermocouple so it is in the center of the oil volume, the colder walls of the chamber will create a gradient that you don't want to sense. Very fine foil would be the best since it has a larger surface area for the oil to transfer its heat to.
pdf with response times here (http://www.omega.com/temperature/pdf/Unsheathed_fine_gage.pdf)
-
First, as mentioned, you absolutely want to measure the temperature of the oil, not the chamber. +/- 1 C in a couple of seconds is a hard requirement as is, and doing it with anything extra in the way is probably impossible.
Second, you need to figure out if your requirements are even well formed. Unless you have strong forced mixing, you won't have +/- 1 C temperature uniformity within your reservoir. In that case it doesn't matter how accurate your temperature sensor is, the reading won't mean anything, at least in an absolute sense (the relative measurement might still be useful).
Anyway, for immersion sensors, thermocouples are pretty good. They can often be dunked directly in the liquid without any protective housing. Other sensors may need housings or insulation which slows the thermal time constant. IR sensors are also an excellent choice. Normally you have to be worried about being fooled by reflections, but the emissivity of oil is probably close to 1.
-
I've used a D18S20 - and I don't recall them being so slow. Even in air.
-
ciccio, your PT100 RTD sensor recommendation at first glance looks very promising. IST PT100 RTD sensors are relatively reasonable cost (3-7 quid) and response times are much better. Are there any catches with their use?
Thin film sensors are cheap. I suggest PT1000 sensors, such as these: https://www.labfacility.com/1000-Ohm-Flat-Film-Detectors-p211-pg100/ (https://www.labfacility.com/1000-Ohm-Flat-Film-Detectors-p211-pg100/), because the higher resistance will result in less influence from connecting wires' resistance and tempco (no need for a 3 wire connection).
The look-up table is quite simple, but if you have enough computing power you can to solve the temperature vs resistance equation.
Thin ceramic substrate is fragile, and can be damaged by vibrations.
Best regards
-
Check out this baby: http://www.ti.com/product/tmp006 (http://www.ti.com/product/tmp006)
There are some YouTube videos on it: https://www.youtube.com/results?search_query=TMP006 (https://www.youtube.com/results?search_query=TMP006)
There are breakout boards available on Adafruit and Sparkfun. I'm already waiting for mine to arrive.
-
Thanks everyone for the answers. Thermocouple seems to get most of the votes :)
Unfortunately, I can not get inside the oil container, due to the dampers costing a fortune and being borrowed for educational purposes only. On the other hand the oil volume is low and monitored oil container is designed to transfer heat from the oil as much as possible for cooling purposes. So I think the temperature gradient is minimum between actual oil temp and the container body.
Will probably try platinum film and TMP006 sensors, as they look really promising (oh the joys of BGA's) and don't cost much.
-
Definetly K type termocouple. Use heat conductive adhesive to glue it on the container, and depending on the project budget, buy a USB datalogger for it, or something like the ads1118evm for 50 bucks.
http://www.ti.com/tool/ads1118evm (http://www.ti.com/tool/ads1118evm)
-
Unfortunately, I can not get inside the oil container, due to the dampers costing a fortune and being borrowed for educational purposes only. On the other hand the oil volume is low and monitored oil container is designed to transfer heat from the oil as much as possible for cooling purposes. So I think the temperature gradient is minimum between actual oil temp and the container body.
I would bet that the opposite is true. If the container is designed to transfer heat from the oil, then the inside surface of the container will have to be lower (probably much lower) than the oil itself. That's because the thermal conductivity of the container will be much greater than that of the oil. This means measuring the outside (or even inside) surface of the container will not give you the temperature of the oil without taking into account many other correction factors.
If you really need to get the actual oil temperature (and especially if you need to get it quickly), then you will need to get access directly to the oil. If you can't place a sensor in contact with the oil, then measuring the surface of the oil through IR means might be an option if you can place the IR sensor in a position that can "see" the oil without touching it. However, if you have enough mechanical movement to heat the oil that quickly, then it is probably splashing around in a way that would make IR sensing impractical. You're probably going to have to find a way to get direct contact with the oil.
BTW, I have also used the DS18B20, and while they are not super fast, their response time would be measured in seconds rather than the minutes you're seeing. That's an indication that the problem is with the sensor placement rather than the sensor response time.
-
Don't sweat it.
If you can't get into the dampers/resouvoir then you can't. If it's a project for uni, then just state that you can't and why it would/maybe better if you could. It might not really matter.
Thermocouples (type K) are cheap and plentiful and straightforward/well documented. Perhaps glue a couple in different places on the damper (perhaps you can extract some meaning from this - my understanding of dampers is that the heat will be generated in the controlled restriction, so perhaps that's as important as the tank).
-
Any chance of finding the thermal resistance from the oil to the casing? This might allow you to make a reasonable approximation of the oil temp by measuring the case temp, much the same way as we do with junction temperatures and heatsinks.
-
Your university has some thermographic equipment, I hope?
See, very fast responding measurement.
Borrow it. Prepare your DUT for thermography - paint reflecting surfaces.
Find the most interesting spots.
The camera might have a import interface to a PC - use it.
If not, it might show the temperature at the center pixel, use a webcam to see it and use OCR to decode, or put a Fototransistor in front of the screen and measure the brightness of the display.
Greetings
Babysitter