How to eliminate unexpected noise from LM35 output signal after passing through low pass filter?

Isaac Po

Joined Nov 27, 2022
26
Hi, I am currently trying to suppress the noise signal from 50Hz and above for the LM35 output pin.

To do this, I connect the output pin to a 2nd order Butterworth low pass filter.

The current issue I faced is that the output signal after going through the low pass filter had its noise signal suppressed(dropped from -90dB to -100dB) BUT it would leave behind 2 peaks(-90dB & -95dB) at approximately 400Hz and 800Hz. May I know how to eliminate the two peaks?

Original LM35 output signal:

LM35 output signal after passing through low pass filter:

2nd order Butterworth low pass filter design:

R1 = R2 = 6.6k ohm; C1 = 47nF; C2 = 22nF

2nd order Butterworth low pass filter bode plot diagram:

ericgibbs

Joined Jan 29, 2010
18,849
hi,
E

MrChips

Joined Oct 2, 2009
30,813
One way to reduce noise induced by power line frequency is to average over integer cycles of the AC waveform.
For example, if your AC supply frequency is 50Hz, one cycle is 20ms.
Take an exact number of samples to span n x 20ms, for example, 100ms, and take the average.
For example, take 20 samples every 5ms, or 100 samples every 1ms, and then compute the average.

BobTPH

Joined Jun 5, 2013
8,971
What is the ratio of the noise to the change in the voltage output of the sensor for a change of 1° C?

Ian0

Joined Aug 7, 2020
9,819
The LM35 is accurate to within 0.5°C and has a sensitivity of 10mV/°C, equivalent to an error of 5mV, and you are bothered by 56μV of noise?
What equipment do you have running at that frequency? Brushless DC fans, for instance? Is it being induced into the probes of the spectrum analyser because of the placement of the earth probe?

Last edited:

BobTPH

Joined Jun 5, 2013
8,971
The LM35 is accurate to within 0.5°C and has a sensitivity of 10mV/°C, equivalent to an error of 5mV, and you are bothered by 56μV of noise?
What equipment do you have running at that frequency? Brushless DC fans, for instance? Is it being induced into the probes of the spectrum analyser because of the placement of the earth probe?
Exactly what I was getting at.

MrSalts

Joined Apr 2, 2020
2,767
In LabView, change the voltage scale from db to mV and your noise will vanish.

Isaac Po

Joined Nov 27, 2022
26
Hi everyone, this is the practical circuit I am using to provide the results above:

For the 2nd order low pass filter, I am using MC3403 IC chip connecting according to the one above. The V+ & V- terminal is connecting to the +15V & -15V respectively.

This is the LM35 IC that has it's leg connected to 5V DC supply, ground port from the NI Elvis, & the output pin connect to the input of 2nd order low pass filter(MC3403).

My task is to try to suppress the noise from the LM35 even further through an active filter before implementing the filtered signal to another application.

*Yes I realized the noise signal from LM35 is already quite low. I hope there is a method utilizing an active filter(MC3403) to reduce it further or has slight changes from the original signal.*

Isaac

Attachments

• 7.2 MB Views: 0

MrSalts

Joined Apr 2, 2020
2,767
The long wires are acting like antennae and capturing radio signals, interference from fan motors and anything else in your area. Add a few ceramic caps (100nF, 100pF and a fatter 10uF electrolytic) between signal and ground.

even eliminate the wires and place the LM35 in the same strip on the protoboard.
Twist the ground and signal wire pair that goes to the LabView module.
Is LabView module making the noise? What is the signal in LabView if nothing is connected?

Isaac Po

Joined Nov 27, 2022
26
The long wires are acting like antennae and capturing radio signals, interference from fan motors and anything else in your area. Add a few ceramic caps (100nF, 100pF and a fatter 10uF electrolytic) between signal and ground.

even eliminate the wires and place the LM35 in the same strip on the protoboard.
Twist the ground and signal wire pair that goes to the LabView module.
Is LabView module making the noise? What is the signal in LabView if nothing is connected?
I placed a 10uF electrolytic capacitor between the output of the Low Pass filter and the other leg to ground. I managed to suppressed the 2 peaks from earlier.

Original LM35 output signal:

LM35 output signal after passing through low pass filter + 10uF Electrolytic capacitor:

May I know what happened previously that causes the 2 peaks to occur?

This is a temporary circuit that I built on top of NI ElvisMX's circuit board. I would install the LM35 & low pass filter on a separate breadboard afterward.

Furthermore, how can I reduce the interference from having long wires? Do I shorten each wire as much as possible?

_______________________________________________________________________________________________________________________________________

Signal in the Labview module if nothing is connected:

+5V DC Supply:

Ground port:

+15V DC supply:

Voltage range can be set up to +/- 10V only, therefore unable to observe the signal properties.

-15V DC supply:

Voltage range can be set up to +/- 10V only, therefore unable to observe the signal properties.

Joined Feb 20, 2016
4,476
You really cannot test for noise levels when building your circuit on a breadboard. Your construction there almost guarantees a noisy signal. Aim for short leads (as mentioned above), with very good power bypassing and careful board design. Not a breadboard! Separate the power and signal return paths so no other current shares the same traces as the signal as far as you can.
Do you need anything other than the LM35 and a, R/C filter after it? The LM35 will not respond fast so an R/C filter with a long time constant is really all you may need as long as its output voltage is in the range needed for measurement.

MrSalts

Joined Apr 2, 2020
2,767
Think how you can place the LM35 on the board in the same rows as the filter IC without using any (or very few) jumpers. You may need to spread the pins of the LM35 to get some legs to insert in the breadboard next to the pins of the filter chip - that can remove one or more jumpers.
next, as I said above, any wires you do have should be "twisted pairs" of a ground and a signal back to your National Instrument board's signal input. Avoid large open loops of wire that work as an antenna. Placing the wires flat against the breadboard also help.

Sensacell

Joined Jun 19, 2012
3,449
Now it's down to practical, real world construction and layout practice.

That fancy active filter is probably not helping, as others have said, a simple RC filter with a long time constant and better shielding/grounding/layout is all you need.

Ian0

Joined Aug 7, 2020
9,819
LM3403 is a poor choice of op-amp for the filter. It has a maximum DC offset of 8mV - that could add an extra 0.8° of error by itself. It is also so noisy that the noise figure is not listed in the datasheet.

crutschow

Joined Mar 14, 2008
34,452
Can you explain why you are concerned about the noise level for a signal that changes so slowly?
Just use a filter with a low corner frequency that will still respond to how fast the signal changes (how fast is that?), as dendad suggested.
That corner frequency will likely be much lower that 1Hz.

BobTPH

Joined Jun 5, 2013
8,971
Not to mention that the noise is orders of magnitude less than the uncertainty (error) in the signal.