Grrrrrrrreat! It never fails to amaze me how often the solution to a hair tear problem turns out in retrospect to be something so simple. Mmmmm. It will be worth a trip to Victoria if only to taste the brew and see your setup! I accept the rain check!
@RussH – It's good to see that you have finally found the cause of your problems. It is also nice to see I am not the only one who just can't let go of something like that. Just have to poke, prod, and chew at it until I figure it out. A mild case of OCD I guess.
After reading the postings and the app note it looks like I have just been very lucky in avoiding this kind of problem.
@Tracy – Thanks for the info you posted and the link to the app note. I have both filed for future reference, and will be adding some notes to any schematics where I use these sensors.
@ Falcon - I put a 2k resistor about 8 inches back from the LM34. I tried both 2k and 3k, and didn't notice any difference. On some start ups there is a bit of ripple, but it's only less than 10mV (1 degree F). If I leave everything powered and unplug, plug-in the sensor, it goes away every time. I'm going to try 3k on the other probes and see if that changes. Before this fix, the unplug, plug-in trick worked maybe 5% of the time. I can deal with 1 degree of fluctuation.
@kwinn - Even if I moved to another sensor, this would have bothered me until I sorted it out. Very glad I stuck to it, and I hope this thread helps others with implementing these sensors.
What kind of wire did you end up using in the end? And how long from probe to Stamp?
I made a new probe with a 2.7KOhm resistor (that's as close as a had) about .5" from LM34, but am still getting 30 degree fluxuations.
Right now the circuit is as follows:
LM34 w/ 2.7K resistor in shrink wrap→1.5' Cat3 phone cable (un-shielded) → terminal block→7 pair shielded(braid and foil)STP 15' → Terminal Block→Jumper wires to ADC0834→BS2
I previously tried the following but still had the same fluxuation:
LM34 (no resistor)→15' Cat3 phone cable (un-shielded) → terminal block→Jumper wires to ADC0834→BS2
LM34 (no resistor)→1.5' Cat3 phone cable (un-shielded) → terminal block→7 pair shielded(braid and foil)STP 15' → Terminal Block→Jumper wires to ADC0834→BS2
I have another LM34 probe without the resistor in a water line 4' from the ADC0834 using the Cat3 and it's rock solid.
Could the terminal blocks be a problem?
Does the wire need to be shielded?
That diagram in Post 28 looks like a coax cable carrying the signal after the 2k resistor. Is coax preffered, or required?
If 60 Hz A/C is too close, what type of filtering would help the most?
BTW, I'm also glad you got your rig working. I've been watching this thread for a solution I can apply to my problem.
Thirty degrees of fluctuation is waaaaayyy too much! Yes, the terminal block could be a problem, as could any connection along the way between the sensor and the ADC. The key in Russ's case came when he posted an oscilloscope picture. That showed the high frequency typical of self-oscillation. Fluctuations can also come from AC line pickup, or from a nearby radio station, or from other AC signals on the line (e.g. motor PWM), or from noise on the power supply or ground lines. An oscilloscope helps to resolve what kind of noise is really there. If you have a hand multimeter, connect that to the sensor output at your intermediate terminal blocks and try it on both the DC and AC voltage settings, and on both the power supply and signal lines. Is it the same at all points? The ground connections are especially important.
The coax and the shielding are not necessary in many situations, unless there are strong external AC lines or other signals. Are there other signals running on the CAT cables? Ground unused pairs. Don't think about 60Hz software filtering until you nip it at the source. A big capacitor at the ADC input can also help, but again, identify the noise source first.
Tracy,
There are no other signals on the Cat3 cable going to the first terminal block. There are two other wires used in the shielded cable. They are the +5VDC to, and the signal from the leak detector probes seen in this circuit. They should be pure DC.
I will check all points with the multimeter on AC and DC.
I'll also try with the caps.
I was refering to hardware filtering as in caps, not software.
I might make up another LM34 probe using that second diagram from post #28.
I checked the LM34 signal at the second terminal block against ground and it was rock solid. No flux at all. Of course that was with a multimeter so I couldn't get the resolution of an o-scope. I then added a .o1 cap between the ADC input from that fluxating LM34 signal and ground and the flux vanished. That temp has been steady on my LCD ever since. The resistor no doubt helped but the cap takes the credit.
It's so easy to read a number of suggested solutions and areas to check and skip the minor one to jump on the one you haven't heard before.
Comments
After reading the postings and the app note it looks like I have just been very lucky in avoiding this kind of problem.
@Tracy – Thanks for the info you posted and the link to the app note. I have both filed for future reference, and will be adding some notes to any schematics where I use these sensors.
falcon
The resistor goes at the LM34. Take a look at the diagram and follow the link in post 28 for more detail.
@kwinn - Even if I moved to another sensor, this would have bothered me until I sorted it out. Very glad I stuck to it, and I hope this thread helps others with implementing these sensors.
What kind of wire did you end up using in the end? And how long from probe to Stamp?
I made a new probe with a 2.7KOhm resistor (that's as close as a had) about .5" from LM34, but am still getting 30 degree fluxuations.
Right now the circuit is as follows:
LM34 w/ 2.7K resistor in shrink wrap→1.5' Cat3 phone cable (un-shielded) → terminal block→7 pair shielded(braid and foil)STP 15' → Terminal Block→Jumper wires to ADC0834→BS2
I previously tried the following but still had the same fluxuation:
LM34 (no resistor)→15' Cat3 phone cable (un-shielded) → terminal block→Jumper wires to ADC0834→BS2
LM34 (no resistor)→1.5' Cat3 phone cable (un-shielded) → terminal block→7 pair shielded(braid and foil)STP 15' → Terminal Block→Jumper wires to ADC0834→BS2
I have another LM34 probe without the resistor in a water line 4' from the ADC0834 using the Cat3 and it's rock solid.
Could the terminal blocks be a problem?
Does the wire need to be shielded?
That diagram in Post 28 looks like a coax cable carrying the signal after the 2k resistor. Is coax preffered, or required?
If 60 Hz A/C is too close, what type of filtering would help the most?
BTW, I'm also glad you got your rig working. I've been watching this thread for a solution I can apply to my problem.
falcon
The coax and the shielding are not necessary in many situations, unless there are strong external AC lines or other signals. Are there other signals running on the CAT cables? Ground unused pairs. Don't think about 60Hz software filtering until you nip it at the source. A big capacitor at the ADC input can also help, but again, identify the noise source first.
There are no other signals on the Cat3 cable going to the first terminal block. There are two other wires used in the shielded cable. They are the +5VDC to, and the signal from the leak detector probes seen in this circuit. They should be pure DC.
I will check all points with the multimeter on AC and DC.
I'll also try with the caps.
I was refering to hardware filtering as in caps, not software.
I might make up another LM34 probe using that second diagram from post #28.
I'll report my findings: good or bad.
Thanks for the help.
falcon
It's so easy to read a number of suggested solutions and areas to check and skip the minor one to jump on the one you haven't heard before.
Thanks to all that helped.
falcon