Racking my brain, why am I seeing these tiny voltage fluctuations?

Thread Starter

MikeA

Joined Jan 20, 2013
231
Here is what is happening. I have a 16 bit A/D (ADS1115) converter that is connected to a differential pressure sensor that's not connected. So it should be recording just ambient air turbulence, a breeze coming into the window, or someone walking by.

But I see something that I can't explain. The recorded signal has much more noise between around 6:30pm and 7:15am. And the signal level abruptly drops/rises roughly 0.5mV around those times.

Temperature is stable. AC voltage is stable. Barometric pressure is stable. I can't pinpoint any external factor that would line up with those times. And then I realized that those are exactly the sunset and sunrise times. But with the A/D converter and the sensor are in the basement. With 2 floors of living space above it. How could the sun come into play here?

I could collects data for a few months and see if it shifts with the sun. But I'm not even sure I'm on the right path here. Ideas?

fluctuations.png

The converter and logger are powered through 2 switch mode power supplies in series, so it should be well insulated from the power grid. There are no solar panels on the house or in the vicinity.
 

MrChips

Joined Oct 2, 2009
22,113
I experienced two similar situations in my workplace.

1) We were collecting sensitive data and would experience a dip in temperature every night at around the same time. After investigation, it turns out that the cleaner was coming into the room every night at the same time.

2) In another experiment, the data collected was perfect during the evening but was noisy between 8:30am and 4:30pm on a consistent basis. The fault was in a PC switching supply. The user was in the room one floor below us and he would come in at the same time in the morning and turn on his PC and turn off the PC before leaving. Changing his PC power supply fixed the problem.
 

sparky 1

Joined Nov 3, 2018
384
Without any other description given just the ADS1115 by itself implies that it is the standard evaluation.
TI demo board guide has the schematic but the software control communicates with an MSP430F440 @
section 7.1.2 of the pdf standard eval. Messing with the settings, It starts with lower sample per second then it works upward.
When the project always remains in continuous mode it is 12 bit.
Raspberry Pi may have got around that not sure of the modifications from the MSP430F440 in order to stay in continious mode.
TI was clear as mud as evidenced by the many revisions on that ADC. The many struggles to the higher resolution ADC not only the ADS111X.
TI went in a different direction I think that a strip down running da vinci must exist but the ADC would be 83 - 89 series not ADS111X


A conversion rate that is stable.
https://www.ti.com/lit/ug/sbau157b/sbau157b.pdf
 
Last edited:

Thread Starter

MikeA

Joined Jan 20, 2013
231
I experienced two similar situations in my workplace.

1) We were collecting sensitive data and would experience a dip in temperature every night at around the same time. After investigation, it turns out that the cleaner was coming into the room every night at the same time.

2) In another experiment, the data collected was perfect during the evening but was noisy between 8:30am and 4:30pm on a consistent basis. The fault was in a PC switching supply. The user was in the room one floor below us and he would come in at the same time in the morning and turn on his PC and turn off the PC before leaving. Changing his PC power supply fixed the problem.
That's what I thought at first. It must be interference from something. First I thought it might be coming from the power grid, even though it's well isolated. Looking at voltage I think I can see where the grid operator turns on extra capacity during peak hours. It didn't line up. Then I thought of what could turn on a schedule. The lawn lights come and off at different times too. There is definitely nothing repetitive happening in the residence at those times, and it's a free standing structure with at least 100 feet to the closest neighbor.

It certainly could be neighbors. Perhaps a cheap switching LED flood light with a dusk/dawn sensor?

The A/D has 4 input pins, and I'm logging a pin that's not connected to anything, just pulled down with 10K. And it's as smooth as I would expect that A/D to be. So it's picking up something on the wire going to the sensor, or the sensor itself is picking it up.

unconnected-ad-pin.png
 

Thread Starter

MikeA

Joined Jan 20, 2013
231
The noise is one thing, but there is also a DC offset. Do you have the part# of the pressure sensor?
It's a Setra 263 series. DC off is expected, it's an analog device. It has DC offset adjustment, I never bothered since I can do that in software.

At this point it seems like RF interference from somewhere.
 

MrChips

Joined Oct 2, 2009
22,113
What @Chris65536 meant is that from 0715h to 1830h the signal DC level drops.
Also I see that the noise is lower in amplitude. This would seem to suggest that there is a reduction in gain during that period.
 

MrSoftware

Joined Oct 29, 2013
1,937
See if you can figure out where the noise is getting in. Is it coming from your power supply? The sensor itself? Is it picked up via EMF? So I would first scope the power inputs to the ADC and see if the noise level follows your ADC output. If it does not, then scope the output from the sensor itself to see if the noise is coming from the sensor itself. Figuring out how the noise is getting in will potentially help you figure out the cause. Also is it possible to run the whole thing from a battery? If yes, you could try running it for a few hours from a battery. If the problem goes away then you know it's coming from your power supply side somewhere. If the problem does not go away then I would start to suspect other causes such as EMF. Are you sure the sensor isn't seeing a pressure change, meaning the output is actually valid? Also if you've got any dangling wires, investigate whether they need to be tied to grounds. Dangling wires connected to high impedance inputs become great antennas for picking up EMF.
 

Chris65536

Joined Nov 11, 2019
240
What @Chris65536 meant is that from 0715h to 1830h the signal DC level drops.
Also I see that the noise is lower in amplitude. This would seem to suggest that there is a reduction in gain during that period.
The way I see it, the DC level increases at the same time as the noise. That could be coming from some other power-using device sharing the ground connection. Or it could be RFI, somehow getting rectified by the pressure sensor circuitry into DC. There is a very clear shift in the DC level.
 

Thread Starter

MikeA

Joined Jan 20, 2013
231
Are you using the same ground wire for any other sensors? A little extra current in the ground wire could raise the voltage of the sensor ground, and so also the output signal.
It's only the one pressure sensor, the ADS1115 and an ESP8266 that reads the ADS1115 over I2C and transmits the data over WiFi. Everything on the same ground, and powered from the same source.
 

Thread Starter

MikeA

Joined Jan 20, 2013
231
Also is it possible to run the whole thing from a battery? If yes, you could try running it for a few hours from a battery. If the problem goes away then you know it's coming from your power supply side somewhere.
I think logically the supply can be eliminated since the anomaly seems to happen on a schedule, and doesn't effect the other A/D input, which has nothing connected to it.

Perhaps I should hang a piece of wire on the other A/D input, which is pulled down with 10K, and see if it starts picking up RFI?
 
Top