external temp sensor for EZ430-RF2480

Hi Brennan,

Most of TI's current temperature sensor devices, including the TMP421 and TMP441, are digital interface: either I2C or SPI.  So they would not tie up one of the ADC inputs but would connect to the digital bus instead.  You can use the USCI port of the MSP430 on your target board for this.

The one exception is our TMP300.  Its primary function is to act as a resistor programmable temperature switch, but it does have an analog output you could use with the microcontroller ADC.

The TMP421 and TMP441 are both remote temperature sensors.  So they connected up to a diode (or transistor) that would act as the actual sensor.  They are especially useful for monitoring the die temperature of a processor, but they could be used with a stand-alone diode placed anywhere in the system.  You probably need only consider the TMP421.  They both have the same functionality, but the TMP441 adds a special compensation that's only useful when the device is used to measure the temperature of a newest generation processor.

The alternative are our local temperature sensors.  These have the sensor integrated onto the chip and are used to measure ambient temperature where they are located.  They do not require any additional circuitry beyond decoupling caps and, in the case of I2C interface, pull-up resistors.  Perhaps the simplest local temperature sensor would be the TMP121.  It is a read-only device with SPI interface.  There are many choices, though, depending on the accuracy, resolution, power, size, etc... needs of your system.  One other device I will suggest is the TMP102 since I've gathered you are working on a battery powered application.  It is our lowest power offering.

Rudye

Hello Rudye,

Thank you for the quick and helpful response.  After reviewing the datasheets of your recommendations, I settled for the TMP300 and TMP102 (as we are investigating analog and/or digital sensors in our application).  I ordered a few samples of each immediately and the shipment will arrive in the next few days I suppose.

Can you also address FW-related queries or is there another TI forum specifically intended for such?


Brenn

Brennan said:

Can you also address FW-related queries or is there another TI forum specifically intended for such?

assuming you're trying to interface the temp sensor to the MSP430F2274 on the RF2480, i would recommend that you take a look at the code examples for the device on www.ti.com/msp430codeexamples.  they include several examples of setting up the SPI to communicate to an external device.  this should take you most of the way there.  if you run into any problems, you can ask the folks on the MSP430 forum.

Glad to help out.

If you have any other questions about the temp sensors, the communication protocols, or about programming the registers; let us know.  On the other hand, if you have questions about programming the MSP430, you'll probably want to ping one of those forums.  You'll have a faster response that way.

Rudye

Hello Rudye,

The samples arrived last week.  After a quick inspection, I had a what-I-was-thinking moment regarding the devices' packaging.  Soldering the sensors would be quite challenging I should say.  I would be requesting our PCB guy to make me a simple IC fan-out later this afternoon.  But I was hoping if there are other options that you know of, in using the sensors for an easier breadboard implementation.

Brennan

Hi Brennan,

Yes, the tiny packaging is both a blessing and a curse.

I do have some mounted TMP102's that I can send you if it would help.  I can also ask the team here to mount some TMP300's onto a DIP adaptor for breadboarding, but I need to find out which package you are interested in for that device, either the SOT23 or the SC70.

To send you samples, I will need to get your address.  I will send you a private email requesting this.

Rudye

Hello Rudye,


I updated my profile with my personal email address.  You may reach from there.  By the way, I also made a 'DIP adaptor' this afternoon. But I'm not sure if I it is good enough for breadboarding (or decent enough for viewing that is).    

I am actually more interested in investigating how the sensor communicates and cooperates with the MSP430.  But I cannot do that without the DIP connection, so I really appreciate the help so far.

Brennan 

brief update:

I was able to power up the sensor today. The output at Vtemp is at 718 mV. This would mean a temperature reading of 71.8˚C. The Rset was 180 ohms for a ~0˚C Tset. I just followed Figure 8 of its suggested circuit application.

I think I should be expecting a value at around 200-300 mV instead. Or did I damage the IC during soldering? Any other ideas?

Brennan 

Brennan, an output  voltage at Vtemp with 718mV corresponds to a temperature of about 24-25C.  If you look at the speciifcation table in the data sheet under "Analog Output",  you can see that at room temperature (25C) you can expect between 720mV and 780mV.

Matt

Hello Matt,

I misinterpreted the "analog out:10mv/˚C" portion in page 1. Thank you for pointing this out. It is basically the sensitivity at room temperature.
Now I can do the FW codes.

Bren

I'm back. [:)]

Let me clarify if I got the datasheet values right this time. For voltage values that fall within the 720 mV - 780 mV, the temperature is 24-25C. If the temperature sensitivity is 10 mv/C, and I got a 680 mV output this time around, then the temperature reading would be 21-22C. This is confirmed when I checked the actual room temperature using an IR thermometer.

Is the behavior of the output this linear for all temperature readings? Although figure 5 in the TMP300 datasheet shows typical analog output error values, I am looking for a more straightforward temperature vs voltage graph.  There seems to be none provided. This is particularly helplful because the application I am working on deals with accurately monitoring temperature in a closed system.    

Please advise,
Brennan

Brennan,

This slope is a typical slope given as an ideal reference; at any given temperature the error can be +/-3C.  The temperature vs. voltage graph that you would like does not exist because from 1 production lot to another the characteristic slope and linearity can change but the total temperature error at any given temperature will still be within the aforementioned tolerance.  If you need better than +/-3C accuracy you will either need to calibrate each temperature sensor or you will need to go to a better temperature sensor like the TMP275 or the TMP112 which are both digital out temperature sensors.

Matt

What I have been trying to do is to connect a temperature sensor to one of the available ports in an EZ430-RF2480 kit.  Back then, the choice between an analog or digital sensor was easy to make because the communication (ie SPI, I2C and the like) for a digital interface is not enabled on the demonstration kit.  By 'enabled,' I meant an easy access to the necessary MCU pins. The other alternative of course is to carefully solder hook-up wires into the rather small MCU package.

It looks like we would have to go with the digital temperature sensors eventually.  My follow up question would have to be posted in an MSP430 or EZ430-RF2480 thread (re: clock source among other things). After a quick glance of the datasheets (TMP275's and TMP102's), I would have to go for the latter because it is a low-power solution.  However, if I could not provide the necessary digital interface, I would probably go with TMP275.  Did I miss something?

Brennan

Brennan,

If you need low power and better accuracy, consider using the TMP112.  FYI--a 2 wire interface will be necessary for communicating with the TMP102, TMP112, and TMP275.

Matt

Hey Matt,

To be honest, I thought you meant 'TMP102' and you just had a typo with 'TMP112' in your previous message.  I learned this afternoon that the latter is actually a different digital temperature sensor and a better upgrade at that (in terms of accuracy). I ordered some samples right away and hope to play with them next week. I'd let you know my thoughts by then.

Thank you once again,
Brennan

Quick question: What is the best sampling time for the TMP300?

Because there is none provided in the datasheet.  I measured though a period of 1.5 us from the analog output pin of the said temperature sensor.  I assumed that that is the effective sampling time as well.  However, our ADC code (we are using the EZ430-F2012 by the way) is generating the wrong value everytime.

Regards,
Brennan

Brennan,

Most of our digital temperature sensors that have an on chip ADC have about a 20-100mS conversion time.  Temperature itself does not change that fast, and this conversion time is largely due to the sampling and averaging of the converter to reduce the noise and improve the measurement resolution.

When you say that your ADC get the "wrong" value, how wrong do you mean?

Matt

Hello Matt,

The converted output hex values exceeded the sensor's output analog voltage by ~200mV. My colleague is also suspecting that the reference voltage was set too high for our application, which involves values in the mv range. We set it at 1.5V and we are using the F2012 by the way.  We'll see the waveforms and timing sequences next week to verify.

Brennan

This is not related to my previous query. We are still ironing out the FW for the temperature sensor as of this post.

I was just wondering if it is possible to increase the sensitivity of the TMP300. Because like all temperature sensors in plastic packages, it could be less sensitive to the temperature of its ambient environment than to the temperature of its leads.

Two solutions I had in mind is: to increase the sizes of the solder paths and to make a limited copper fill in the PCB. However, any better ideas are most welcome.

Quick update: We were able to perform A to D conversion (for the temperature sensor) using the MSP430 chip.
Thank you so much TI guys for the valuable insights and helpful suggestions.

Cheers,
Brennan