Noise reduction in a low voltage 24bit adc

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
Hello,

I am looking for suggestions here for my circuit. I have a 3 Wheatstone bridge, 3 instrumentation amplifiers (INA333), One 24-bit ADC (ADS1220) sampling those INA333 outputs at 1000 SPS.

Here is the design and signal flow of the circuit for the result I have uploaded in attachments:
my circuit operates at 3.3V from a laboratory dedicated power supply.

1. I have used OPA333 to set the reference of INA333 at the mid supply of 1.65 V(since I am using a potentiometer to set voltage divider to set mid supply for OPA333 input there is some shift I belive in mid supply due to human hand error)

2. 3 INA333 inputs at grounded since my strain gauge is not yet ready to have experimented hence the voltage difference between inputs should be zero(theoretically).

3. I get the ADC input from ina333 outputs and the reference of ADC is power supply reference that is 3.3 V hence ratiometric measurements.

4. Now I have separated the power supply of ANALOG and DIGITAL part of ADC in order to separate noise to a certain degree but still, I am getting noise of 0.7 amplitude at mid supply

I have not got my hands on LP5907 component yet but I have i have included it in the cirucuit( please consider that the plot I got was from two seperate power supply in laboratory which will be replace with LP5907 once I get them)
And Yes I am right now using breadboard to prototype my circuit hence capacitance can be of significane?

Here is attached image of result of above mentioned circuit conditions and schematics.

It would be great if I could remove noise in order to use full 24bit(20bit ) of ADC resolution, thank you.3 INA333 AND 1 ADS1220 AND OPA33.JPGnoise.jpg
 

Papabravo

Joined Feb 24, 2006
19,266
Have you identified the source of the noise?
Does the noise have a wide bandwidth?
Are the signal levels comparable to the noise levels?
 

ericgibbs

Joined Jan 29, 2010
16,364
h Hiral,
There is inadequate power rail decoupling, also it appears the power line and Ground commons appear Daisy-chained.
Any chance of a photograph of the project.
E
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
Have you identified the source of the noise?
Does the noise have a wide bandwidth?
Are the signal levels comparable to the noise levels?
no, I am pretty new to electronics circuits relatively.
I don't know about bandwidth either, i will check how to know the bandwidth meanwhile can you suggest how to do that as well?

by means of signal levels comparable to noise level then if i have understood it correctly you meant by input signals at INA333 right?(I measured these signals when the inputs where grounded)
 

Papabravo

Joined Feb 24, 2006
19,266
The reason I ask is that eliminating the source of the noise is easier and more effective than any other measure you can take without exception. If the noise has a limited bandwidth, then filtering may be effective. The problem with filtering is that you may degrade the quality of you signal as well. If the signal levels are comparable to the noise levels and they have the same bandwidth characteristics, then you have no choice but to eliminate the source.

This is speculation on my part, but the resistors in the Wheatstone bridges may be the noise source you are looking for. If that is the case, you are going to have trouble eliminating them.
 

crutschow

Joined Mar 14, 2008
30,774
Frquency of noise in the data; depends on sampling frequency of adc?
Not in practice.
Any noise above 1/2 the sample frequency is aliased back into the passband.
You need an anti-aliasing RC low-pass filter at the ADC input to suppress any analog noise above the highest signal frequency of interest.
What is the highest signal frequency you need to see (how fast does the bridge signal change)?
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
h Hiral,
There is inadequate power rail decoupling, also it appears the power line and Ground commons appear Daisy-chained.
Any chance of a photograph of the project.
E
IMG-0801.jpg


I am not at lab hence I am not including power supply here for analog part, inputs of INA333 are floating in this picture but try to assume that I had attached it to nearest ground of respective INA333.
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
Not in practice.
Any noise above 1/2 the sample frequency is aliased back into the passband.
You need an anti-aliasing RC low-pass filter at the ADC input to suppress any analog noise above the highest signal frequency of interest.
What is the highest signal frequency you need to see (how fast does the bridge signal change)?

Here is the thing about the frequency, my project is all aboout me apply the rossette strain formula to find angle of strain with respect to strain gauges hence the I have no specific requirement for frequency here but you can say that since I am sampling each channel(4 channel in total) at 1000 sps I am getting sampling frequency of approx 250 Hz.
 

crutschow

Joined Mar 14, 2008
30,774
Here is the thing about the frequency, my project is all aboout me apply the rossette strain formula to find angle of strain with respect to strain gauges
And here's the thing about sampling frequency--
It is only important as compared to the highest analog signal frequency of interest.
If you need a 250Hz signal response, that's fine, but it would seem you don't.
It sounds like it settles to basically a DC level after you apply the rossette stain, is that correct?

So I repeat my question:
How fast does the signal level change and how often does it change?

We need to determine the highest signal frequency, because it's critical to limiting the noise bandwidth and thus the noise in your sampled signal.
If you can't answer our questions, then the best solution to your problem cannot be suggested.
 

ericgibbs

Joined Jan 29, 2010
16,364
hi Hiril,
The Photo in post #9, indicates you are powering from the USB via the Arduino.
Also, the wiring is like a birds nest and will give you the noise problems, the layout is unsuitable.

E
 

crutschow

Joined Mar 14, 2008
30,774
As noted by others, it is likely not possible to get low noise with such a spread out circuit, with all those long flying leads.
It needs to be built on a small, soldered vector type board with a copper ground plane, and short, direct connections.
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
And here's the thing about sampling frequency--
It is only important as compared to the highest analog signal frequency of interest.
If you need a 250Hz signal response, that's fine, but it would seem you don't.
It sounds like it settles to basically a DC level after you apply the rossette stain, is that correct?

So I repeat my question:
How fast does the signal level change and how often does it change?

We need to determine the highest signal frequency, because it's critical to limiting the noise bandwidth and thus the noise in your sampled signal.
If you can't answer our questions, then the best solution to your problem cannot be suggested.

OK, so a while ago I conducted a little experiment with function generator in order to see how nicely the sampling frequency of adc responds to the function generator. I was able to get a smooth non edgy sin wave at 10 hz of function generator frequency beyond that it started to get edgy little by little. hence I decided to keep all my furthur experiment with strain guage not above 10 hz. Of course I could go for higher sampliing rate adc but it was not necessary
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
hi Hiril,
The Photo in post #9, indicates you are powering from the USB via the Arduino.
Also, the wiring is like a birds nest and will give you the noise problems, the layout is unsuitable.

E

I am not powering it with arduino, as I am at home rigth now so I use it for some other purposes but the first plot that I shared was from laboratory power supply not with arduino. Yes the wiring looks like a nest, I would be switching to soldering breadboard not the PCB when I finalize my circuit.
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
What did you expect? With this layout the overhead fluorescent lights could be a problem. Sheesh!
I apolozie for this, is there any nicer way in order to better this layout? Since I am still experimenting with component I kept my circuit to breadboard level for ease of changing anything.
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
hi Hiril,
Check the Red wire to the bread board track, no other connections on that track.?
E
View attachment 273896
Again I would like to mention I am at home right now and I promise you this that I am not using this red wire 3.3v arduino supply for anything( right now its connected to digital power supply of ADS1220(its hard to see but its there in the picture(sorry for messy circuit))), I am just using it right now to do some changes with my circuit not related to this topic and yes its not going anywhere because I am experimenting with analog power supply of ADS1220
 

Thread Starter

Hiril Patel

Joined Dec 25, 2021
31
As noted by others, it is likely not possible to get low noise with such a spread out circuit, with all those long flying leads.
It needs to be built on a small, soldered vector type board with a copper ground plane, and short, direct connections.

Ok I will switch to that once I am done with power supply of analog part of the circuit. but is there any suggestion for filtering signals? as we are talking about 10hz of signals
 
Top