Differential amplifier noise

Hi all, I have a problem with my INA122.
I've done different test:

  • 4 Res. 1/4w => output is always stable at 1.,20v +/- 0,05v
  • 4 Regolable Pot. => output is instable and decreases or increases steadily, if i'm near to resistances there are additional noises (+/- 0,100/0,200v)
  • 4 Strain Gage on the same piece of alluminium => output is instable, +/- 0,100v when i'm near or not to resistances
  • 3 Strain Gage and a separated loadcell with 1 Strain Gage => output is very instable, +/- from 0,500v to 1,000v if i'm near to the loadcell, if I touch it, it stabilize and the noise is +/- 0,100v

Now, i've tryed to shield the circuit and i've noted this:

  • 4 Res. => nothing change, it was already stable
  • 4 Regolable Pot. => same problem, it's related to Pot. (this is why I have not used them in this circuit)
  • 4 Strain Gage on the same piece of alluminium => If alluminium is outside the shield, there still same noise(+/- 0,100v), if inside the shield, no more noise( +/- 0,005v)
  • 3 Strain Gage(inside the shield) and a separated loadcell with 1 Strain Gage => If loadcell is outside the shield, there still same noise( +/- from 0,500v to 1,000v), if inside the shield, little noise( +/- 0,020v)

Shield was connected to ground.
I've also tried to connect the body of loadcell to ground, i didn't notice any change.

Also, the output of wheatstone bridge is:

  • 4 Res. => stable
  • 4 Regolable Pot. => little instable couse Pot. inaccuracy problem
  • 4 Strain Gage => stable at 1,8mV
  • 3 Strain Gage and a separated loadcell with 1 Strain Gage => stable at 1,7mV

The loadcell is working, if i press the wheatstone bridge output change correctly.

So, my conclusion is: INA122 amplify all the noise that are on loadcell body or on straingage much more than the pressure on the load cell so it becomes unusable. Create a shield that cover circuit and loadcell is impractical, and there is anyway a bit of noise.
Now, what i can do? I need to create a precise digital balance with a self made amplifier that work with a wheatstone bridge and a selfmade lowcost loadcell.
As I said loadcell is correctly made and working, the INA122 seems working with 4 normal resistances so i deduce that INA122 is working(i've tryed 2 of them with the same results).
There are no supply interferences.

Maybe very precise is not really possible I guess, but working yes as i saw from other peoples.

The only method that came to my mind to solve this problem is to place a filter on wheatstone bridge output or another amplifier with and integrated filter. But I don't know what type and how.

Here's some picture and scematics:
(Images are big, I don't know how to resize them so rightclick on them and press "Visualize image")







(Near A, B, C, D in the last images are resistance, don't care about rectangles near them)

Please help me, I'm freaking out with this.

Your breadboard has lots of antennas, which is not helping with your noise problem. Twist the two wires going into in+ and in- together and see if that affects your noise level. If those wires are closely coupled, the effect is that the coupled noise is equal on both wires and will cancel at the +/- inputs of the differential amplifier. If that helps then maybe you can figure out a way to mount your amplifier on your load cell and shorten up/twist the wiring to good effect.

Not sure what to suggest about the one separate gauge. I can't tell how it is wired or whether or not the cable is shielded.

I don't see any decoupling capacitor between the power supply pins of the amplifier, nor any filter capacitor on the power lines to the board (neither in the schematic nor in the pictures of the breadboard). Between the open wiring, noisy pins of a breadboard, and lack of bypass capacitors, I think a lot of noise is to be expected.
Ciao,
Lenny

^ what they said.

Just use the breadboard to verify if your circuit is correct/working.
But design a proper PCB for your project, prototype it again on the PCB and do your noise evaluation using the PCB prototype.

Filtering out any RF on inputs is always a good idea for sensitive DC circuits like this - 1nF ceramic across all the load-cell input leads
should kill any RF pickup without affecting the frequencies of interest. Ferrite toroid on the whole lead close to the breadboard is
also another possibility.

Definitely keep all the wiring on the breadboard short and neat and supply decoupling capacitors are obligatory
for a high-gain amplifier, otherwise it is very likely to oscillate via supply-rail feedback. A ground-plane would be nice,
though obvious difficult on a breadboard. One can add a grid of ground wires across the whole breadboard to
improve matters.

Reducing the gain in the circuit will reduce risk of instability/oscillation BTW.

Reducing the gain in the circuit will reduce risk of instability/oscillation BTW.

This is an important point in quality instrumentation front end circuitry. In amplifying very low level signals it's often much better to use two stages of op-amps using lower gains in each stage to get the desired overall gain rather then forcing one stage to supply all the gain. That also allows one to select stages for the most important specifications needed for the application, such that the first stage having the best low noise qualities where as the last stage might be selected for it's output current capabilities.

Lefty

MarkT:
Filtering out any RF on inputs is always a good idea for sensitive DC circuits like this - 1nF ceramic across all the load-cell input leads
should kill any RF pickup without affecting the frequencies of interest. Ferrite toroid on the whole lead close to the breadboard is
also another possibility.

Definitely keep all the wiring on the breadboard short and neat and supply decoupling capacitors are obligatory
for a high-gain amplifier, otherwise it is very likely to oscillate via supply-rail feedback. A ground-plane would be nice,
though obvious difficult on a breadboard. One can add a grid of ground wires across the whole breadboard to
improve matters.

Reducing the gain in the circuit will reduce risk of instability/oscillation BTW.

I want to try to place a filter to power lines and wheatstone bridge output, it's good 1nF? how i need to connect it, between + and - like a short circuit? but a filter don't need a resistance?

A mechanical earth ground to the metal of the strain gage itself. It is acting like an antenna. Not knowing what your local sources of RF noise are, it would be hard to say what might be inducing the noise. You could also try running the wires through some ferrite beads as well as twisting the wires from each strain guage. All are ways of reducing the level of interference. Shielding around the wires from the strain gages. Ground shield wires AT ONE END ONLY. Preferably the end closest to the amp.