Usb 6008 + LM35 temperature sensor problem

[east star]

Hi, Mark,

I was desperate this morning before I saw your post. I am working exactly on LM35DZ and USB6009. I think I met the same problem with you. A real puzzle.

Actually I have two types of LM35. One is LM35DZ (To92) and the other is LM35Dm (SO8).

First, I did experiments on LM35DM. The pins are Pin 1(Vout), Pin4(Gnd) and Pin8 (+Vs). I connected Pin4 to two diodes (1N4148) in series instead of power GND directly. (Exactly as the Figure 7 in the NI PDF manual, see the attached PDF), I connected Pin8 to +5.8V (Output from a battery pack). A 18K resistor was connected to Pin1 as a small load. I measured the differential voltage between the Pin1 and Pin4. The output is perfect, equaling the room temperature multiplying 10 mV/C. I used USB6009 and Multimeter to cross validate the value.

However, when I did the same experiments on LM35DZ. The output is always between 0.31~0.33V no matter I cool the LM35DZ using a ice pack or a hot solder iron. I designed my PCB boards based on the experiments on LM35DM but it did work on LM35DZ. You can imagine how sad I was.

I cannot get good results on LM35DZ so I have to resort to LM35DM. I repeated the experiments, It works ok. However, when I connected the Pin4 (Gnd) to power Gnd instead of connecting to two diodes (Figure 1), the output becomes 3.75V.  It is Unbelievable.

I also tried to connect the LM35DZ pin 1 (+Vs) to +5.8V. Pin 3 (Gnd) to power Gnd (Figure 1). The output of Pin2 is also between 0.31V~0.33V. I tried 4 LM35DZs and got the same results.

I even doubted the NI LM35DZ is a temperature sensor.

Help me, please. Mark, If you get any progress on this problem.

East Star.

[AnalogKid2DigitalMan]

Interesting.

Mark- what is the input capacitance of the 6008/6009? The LM35's may get quirky above 50pF. Has anyone tried the decoupling/damping on pg 6 of the pdf supplied in the above post?

[east star]

I tried Mark's method 1. It works.

The problem for me is that  I made a PCB to amplify this signal with difference to 0.25V. The temperature range I am measuring is around 20~30 degrees.

I have to find the problems of my PCB.

[east star]

There are many figures on Pg 6. Which figure you mean?

[AnalogKid2DigitalMan]

Sorry, actually page 5, figures 3 and 4 and related discussion below them.

[Troy]

I am having a similar problem only using an LM20 as opposed to the LM35, and using the USB-6008.  I tried all of the suggestions listed eariler in this thread with no success.  The LM20 data sheet, like the LM35 suggested using the capacitance compensation for long line lenghts or high capacitance.  I tried this as well with no change in performance.

Specifically, the LM20 has a negative temperature curve and I am unable to get it to read above +66.7 degrees C or 1.06Vdc output.  Everything below that point works perfectly.  I have verified the software has the correct ranges, I checked the voltage at the input to the DAQ with both a meter and an oscope.  The meter reflected exactly what the DAQ card said, and the Oscope showed less than 25mVpp of noise.  My software is doing some averaging of samples, so the noise effect should be very minimal.

I am going to try some series resistance of a high value, and try a high value resistor at the input of the DAQ, but I don't hold much hope at this point.  Does anyone have any other suggestions?  I have also tried several new LM20s to be sure I wasn't trying to fix a bad temp sender.

[AnalogKid2DigitalMan]

Hi Troy:

Have a look at Fig 8 in the app note:

http://cache.national.com/ds/LM/LM20.pdf

All I can suggest is to try to add a series resistor and parallel cap (try 470 ohm and 0.1uf). It looks like you have a match between the DAQ reading and scope. What happens when the DAQ is disconnected, does your sensor output rise above ~1V?. What voltage are you supplying to the LM20 and is supply ground connected to DAQ input ground? What mode is you DAQ set to (DIFF, RSE, NRSE)?

Sorry for all the queries

[Troy]

All I can suggest is to try to add a series resistor and parallel cap (try 470 ohm and 0.1uf). It looks like you have a match between the DAQ reading and scope. What happens when the DAQ is disconnected, does your sensor output rise above ~1V?. What voltage are you supplying to the LM20 and is supply ground connected to DAQ input ground? What mode is you DAQ set to (DIFF, RSE, NRSE)?

To answer your question, I did try the filter with no success.  If I disconnect the DAQ from the output of the sensor, I get the correct reading from the sensor (about 919mV for +80C)  The LM20 is getting power from the 5V source on the DAQ and is using the ground from the DAQ as well.  Since I am doing a voltage measurement and have all 8 channels used, I wasn't given a choice on setup, but it is set to RSE. 

I have been able to solve the problem, but have introduced some other minor issues.  I ended up using a single op amp in a voltage follower configuration (I had some in an SOT23-5 package that was easy to dead bug into the circuit).  This causes a bit of gain error as the voltage follower is not a perfect 1:1.  Since my temperature coeffecient is negative I am about 3 to 5 degrees high across the temp band.  I think I can correct this with a tweak to the gain of the op amp.  I also ran into one other problem of picking up noise on the long cable runs (about 50').  I ended up having to put a small value cap at the input of the DAQ to eliminate the noise being picked up by the cable.  Both of these issues are fairly minor and can be worked around.

If anyone has any other suggestions, I am definately open to them.  The voltage follower idea came from reading this and other posts that have had issues with using temp ICs to drive the DAQ inputs.  I felt that the voltage follower would provide the boost in current necessary to ensure there was enough to traverse the line with no problems and feed just about any load presented to it.   

[santiore]

hI

Just put a 220 ohms resistance between Vout and GND in the LM 35.

Use Differential connection.

If you get too much noise, take several samples and take the mean or DC value.  You'll get 0.1 degree of precision.

See the attached.

Regards

Santiago Orellana

HighLights - National Instruments Ecuador

[LBLguy]

I think I figured out what is going on. The USB 6009 has a 144 k input resistance, and a source of +1.4V at the bottom end of this resistance (or eqivalent). I think the same is true for the USB-6008. The LM35 can put up to 1 mA into a load connected to ground or some negative voltage (to read below 0 C). However, if the load resistor is tied to a positive source (such as the 1.4 V terminating the input resistance), the output will be pulled up higher, since it appears that the LM35 can only source current, but cannot sink it. This is why an opamp buffer works (which can either source or sink current). A 1.5k resistor to ground should help here. It should be OK down to 1 C and up to 150 C with this.