The Main Tolerance is on Vref Spec, which is : 1.20 1.25 1.30 V ie -> ~4%
VREF, even though it may vary from part to part, is directly measurable and can be used in the temp calculation.
LM317L (fairchild)
Minimum Load Current to Maintain Regulation VI - VO = 40 V 3.5mA typ Spec: 10 mA
LM317L (OnSemi)
ILmin − 3.5mA Typ Spec : 10 mA
The minimum load current for the OnSemi and Fairchild parts are spec'd for VDIF < 40V. This is similar to the TI parts, which are futher characterized, as noted above, for VDIF < 15V. Fifteen volts is plenty of headroom for this app.
The TI data I have seems to have changed again... Minimum output current to maintain regulation VI – VO = 35 V 1.5mA typ Spec : 2.5 mA
So if you specify TI LM317L only, and also make sure you do get parts that match that 2011 data, you can base on 2.5mA Min load.
This means it is guaranteed to give correct current mode only under 500 ohm sensor values, typical will work to 833 ohms.
Phil, I'm not sure if this is your intent, but I'd point out that Vin when deselected should be open circuit or diode isolated, not grounded. If a Vin is grounded to deselect it, then I believe current will flow backwards through the LM317 chip from adjust/output to Vin. I don't have an 'L version to test, but I did just check it with a standard '317.
The AD590 and AD592 have become expensive, but there is still the venerable half-buck LM334, current source/temperature sensor. The output is linear with temperature, single point calibration, and it has the advantage of transmission over two wires as a current.
I was looking to see if there was a way to replace the diodes in the circuit I posted with a transistor to produce a constant current circuit until I saw your circuit using the 3 terminal regulators. Very nice idea. There are a few 3 or 4 terminal precision regulators in small packages that look like they would work for this. Don't really have any use for it at the moment but I am going to add a few 1K and 10K thermistors and precision regulators to the next order so I can try it out.
Here's a diagram to go with the LM334 scheme. In this circuit, the sensors are common bussed to Vdd = 5V, and current from the selected sensor passes through the CD4051 8-channel analog mux to a 10kΩ resistor.
The output across the 10k resistor is ~10mV/°C, with an offset of ~2.73V. That is suitable for a 12 bit or better ADC, or even an 8-bit ADC0838, which has a flexible psuedo-ground terminal and can use a 1V reference.
Resistance of the mux switch does not affect the accuracy, because the signal is a current. The signal current is ~1µA/K, and can be trimmed with the 226Ω resistor, or in firmware. The 226Ω resistor should be a precision type soldered on right at the LM334. See the data sheet. The LM334 operates down to 1.3V from its (+) to (-) terminals.
It would also be possible to do the mux switching on the high side, but as a matter of principle I like to isolate the active temperature measuring node from noise that might be picked up on other sensor cables. It is amazing what unshielded cables can pick up out in the environment. It is best to use shielded or coaxial cable for the sensor leads and to be aware that pickup might exceed common mode limits of circuit components. In my circuit, I'd consider running the mux at a higher Vdd than the sensors, to keep the voltage from even deselected sensors more centered in the common mode range of the mux.
I have used AD590 sensors for a long time. They use only 2 wires with a wide supply voltage range ( 3-15 volts ). They should be scanned and read in about 20 ms to prevent self heating from distorting the reading. An array of scanned sensors can all return their reading on a single wire because the reading is in the current mode at 1 microamp per degree Kelvin. You could hook them up to a decoded counter, say a chain of 74HC4017 chips for as many sensors as you want and then just select them by sending pulses to the counter. Boop oop!
I know I am late, just found this posting via Google. I am working on the exact same project, monitoring composting. My goal is to develop a system that I can apply to Community Gardens. I am coming from the Compost managers biological science aspect. I have come up with the compost environmental factors that help guide a systematic selection process of temperature monitoring options. I am at the beginning stage but have done a bit of research on the criteria for selecting a sensor option. Most Compost systems use the three bin wooden box 3 foot by three foot by three foot (one cubic yard) most ideal for reaching required temperatures. I believe you are looking for efficiency in the composting process and the best tool or method is use temperature sensors inside the compost. Temp range ambient temperature to 185 C. Ideal high end 180 C to 185 C. Not to exceed 185 C since it causes destruction of beneficial organisms. I found taking readings every hour for two weeks would work well. Number of sensors 5 sensors per layer with three layers and 2 ambient sensors = 17 sensors per bin times three bins 45 plus 2 ambient is 47 sensors. Probe distance about two meters to hub. Remote independent DC power source working up to solar power method. I like measuring temperature in degrees F (LM34 IC) but may have to settle for Celsius and convert to Fahrenheit since LM34DZ accuracy for cheapest is +/-3 degree not good enough. I was looking at inserting my sensors into a Dollar Tree kids Hoola-Hoop to make it more durable, provide uniformity in placement and easy to insert and remove when turning. I am toying with Series or octopus arrangement. I would be very interested in working with you on resolving this issue.
Comments
Minimum load is 1.5 mA (typ.), 2.5 mA (max.) when VDIF < 15V.
Perhaps you were looking at the plain LM317 (TO220 version)?
-Phil
The Main Tolerance is on Vref Spec, which is : 1.20 1.25 1.30 V ie -> ~4%
LM317L (fairchild)
Minimum Load Current to Maintain Regulation VI - VO = 40 V 3.5mA typ Spec: 10 mA
LM317L (OnSemi)
ILmin − 3.5mA Typ Spec : 10 mA
The minimum load current for the OnSemi and Fairchild parts are spec'd for VDIF < 40V. This is similar to the TI parts, which are futher characterized, as noted above, for VDIF < 15V. Fifteen volts is plenty of headroom for this app.
-Phil
Minimum output current to maintain regulation VI – VO = 35 V 1.5mA typ Spec : 2.5 mA
So if you specify TI LM317L only, and also make sure you do get parts that match that 2011 data, you can base on 2.5mA Min load.
This means it is guaranteed to give correct current mode only under 500 ohm sensor values, typical will work to 833 ohms.
The AD590 and AD592 have become expensive, but there is still the venerable half-buck LM334, current source/temperature sensor. The output is linear with temperature, single point calibration, and it has the advantage of transmission over two wires as a current.
I was looking to see if there was a way to replace the diodes in the circuit I posted with a transistor to produce a constant current circuit until I saw your circuit using the 3 terminal regulators. Very nice idea. There are a few 3 or 4 terminal precision regulators in small packages that look like they would work for this. Don't really have any use for it at the moment but I am going to add a few 1K and 10K thermistors and precision regulators to the next order so I can try it out.
BTW, I like the LM334 idea, too!
-Phil
The output across the 10k resistor is ~10mV/°C, with an offset of ~2.73V. That is suitable for a 12 bit or better ADC, or even an 8-bit ADC0838, which has a flexible psuedo-ground terminal and can use a 1V reference.
Resistance of the mux switch does not affect the accuracy, because the signal is a current. The signal current is ~1µA/K, and can be trimmed with the 226Ω resistor, or in firmware. The 226Ω resistor should be a precision type soldered on right at the LM334. See the data sheet. The LM334 operates down to 1.3V from its (+) to (-) terminals.
It would also be possible to do the mux switching on the high side, but as a matter of principle I like to isolate the active temperature measuring node from noise that might be picked up on other sensor cables. It is amazing what unshielded cables can pick up out in the environment. It is best to use shielded or coaxial cable for the sensor leads and to be aware that pickup might exceed common mode limits of circuit components. In my circuit, I'd consider running the mux at a higher Vdd than the sensors, to keep the voltage from even deselected sensors more centered in the common mode range of the mux.
Duane J