Instrumentation Amplifier for Use in Biomechanics Laboratory

I am an intern in a Biomechanics Lab and we have an analog force sensor (http://www.jr3.com) with noisy output and we'd like to try to house clean as much noise without sacrificing any performance or inducing delay.

As far as the equipment associated with the analog force sensor, the inputs are shielded correctly. The grounding is sound, and good-grounding practices have been followed. Spike trapping is up for consideration, but I am told that an INA is preferable. The power supply has also been checked and is configured correctly. All of this has led the lab to the conclusion that an instrumentation amplifier would be the best "next-step" in cleaning up the signal.

I am new to circuitry, so I researched what I could and jumped in by building an INA from a design similar to the one on this link (**broken link removed**). I read through several articles on good INA design practice and built several iterations of this design according to them. I used the Op Amps from an LM324N chip, and tried many different combinations of resistor values, always ensuring (approximately) that R1=R3, R4=R6, R5=R7. To test the circuit, I varied Rg (R2 in Fig.) and calculated a theoretical output voltage from the equation on the same page (I kept R5=R4 because I read that that increased accuracy). The power supply to the chip was sufficiently more than the theoretical output, but still within the acceptable range. I measured the output voltage against a wall ground to determine the actual voltage. There was a relationship between the actual voltages and the calculated gains, but no where near the values of the theoretical voltages (especially for high gains >6). I can provide spreadsheets of the resistor values, gains, theoretical voltages, and actual voltages for the different circuits I have tested.

After failing to construct a functioning circuit from designs, I resorted to using an INA118P. Again, I varied Rg values and measured output voltage and compared it to the theoretical output voltage as calculated from the gain. I received the same kind of correlation between gain values and actual voltage output, but again they were no where near the theoretical values.

I searched for threads on instrumentation amps, but not found any that have been totally applicable to my predicament. I'm honestly not sure how to proceed, so any direction would be greatly appreciated.

Here are some specifications that have been requested on previous INA threads:
Signal Levels: +- 10V
Sensor Outputs: +-10V
Frequency of Signal (~20-60Hz)

Can you give us more info on the "noise", i.e., frequency, amplitude, wave shape, etc.? Maybe a short description of how it's affecting the basic signal and how it's screwing up your data for that point.

Better yet, an oscilloscope pic.

[EDIT] JR3's site claims "Internal electronics for enhanced noise immunity". Perhaps the sensor's at fault?!?

You said you "measured the output voltage against a wall ground to determine the actual voltage". Why did you do that? You should measure the output voltage against the circuit common (ground) which may or may not be the same as the wall ground.

1) What force sensor did you buy, so we can find the datasheet.
2) An LM324 is totally unsuitable.

There are lots of problems with data acquisition when ground loops are involved. If this is tied to a SYSTEM that has both analog inputs and outputs careful considerations are required to eliminate ground loops.

Some ways are differential or quasi differential reads. Isolation modules. Or when there are voltages to output use current rather than a voltage output to eliminate the ground loops.

If the output of the sensor that your interested in, you can level shift the output, so the low end of the A/D isn't all noise and you also increase the resolution of the sensor.

KISS,

UTHSC1 said:

I am an intern in a Biomechanics Lab and we have an analog force sensor (https://www.jr3.com)...

Click to expand...

Although the datasheet (at that site) had precious little data.

I apologize in advance for my lack of knowledge and lack of information I have about the noise. I have attached two images of the noise, "forces_FFT" and "sample data". The Y Axis in "sample data" is Newtons, and I believe it is a reading from the load cells. I am not sure what the Y Axis is in the "forces_FFT", and I know this renders the graph basically useless, but at present it is the best that I have. View attachment 65848View attachment 65849

The sensor model is the J3 45E15A4-I63-AF and the data sheet is at this link.
https://www.electro-tech-online.com/custompdfs/2012/07/spec_sheet_45E15A4_eng.pdf

To respond to crutschow's post, I used two bench-top power supplies. I used a BK Precision DC Power Supply 1710 to supply power the chip and a BK Precision Triple Output DC Power Supply 1651 to create the differential input. I then measured the chip's output against the wall ground because the two power supplies were plugged into the wall. I assumed that all the grounds were therefore the wall ground.

Why is the LM324N totally unsuitable?

We have tried to eliminate all ground loops. Is there a way to verify this?

Yea, but the moment I hear the word Bridge, it immediately means trouble. These sensors are complex as in this one: https://www.electro-tech-online.com/custompdfs/2012/07/5979_SENSOR_W_EXT_ELECT.pdf and there are so many different varieties from the manufacturer that things gets messy, The "calibration sheet" or the sheet that came with the sensor will tell you a lot more. e.g. **broken link removed**

In a link under the datasheets yields a 404 error and that's useless too: **broken link removed**

Their datasheet link has been fixed after I contacted them: **broken link removed**

Blind leading the blind, I'm afraid.

So are there any other questions I can attempt to answer? Any pertinent information that I can provide? Could you recommend a better chip (since the LM324N is unsuitable) with the information that I have provided? Is an instrumentation amplifier applicable in this scenario?

First, what sensor? e.g. Poin to the links of the manufacturer data.

Second, a scan of the datasheet received with the sensor.

Look here for some insights on bridge amplifiers: https://www.electro-tech-online.com/custompdfs/2012/07/49470200sscsect2PDF.pdf

For some reason, I THINK you may need a bridge amplifier for EACH row of the sensor, but I'm not gonna pretend I know what sensor you have. Assume makes an A$$ out of U and ME and it's just not worth guessing.

Multiplexing would be difficult because of the likely low R of the sensor (see, I'm guessing again). I would also consider the use of sample and holds, but again I'm guessing. I'm also guessing that someone wanted to save some money and you have a mess to "clean up".

Again, I saw a matrix of x,y data and a voltage vector but I didn't read into it. So, I'm ASSUMING that there is some way to address the element. If there is, then using SAMPLE/HOLDS on the voltage vector might be the way to go. Now you have a snapshot of the values of the sensors in the matrix over a very short interval.

That's what my xtal ball sees. It could be all wrong or broken.

You might as well have said, I bought a car from xyz dealership and I need tires. How much are they? It's a stupid question and impossible to answer. Same with "I need shoe laces. How much are they?

And from what I can tell, "I need an OP amp"; Oh I have a LM324 and I can get them down the street. OP amps have optimized characteristics for different purposes. You have to SELECT one.

We also know nothing about your data acquisition system. At least some of the basics. A 4 bit D/A converter connected to earth ground or something different.

You need a clear set of specs that you have to meet. This defines the system from sensor, acquisition system to presentation.

Environmental will be assumed to be laboratory room temperature conditions unless otherwise specified. Power will be assumed to "as required" unless otherwise specified.

Are you prepared to use matrix math? You want to do this on a PIC micro, RIGHT?

The internet encyclopedia you can use is www.google.com. An example of such a task is: https://bit.ly/Ow6oGa

UTHSC1 said:

...............................

To respond to crutschow's post, I used two bench-top power supplies. I used a BK Precision DC Power Supply 1710 to supply power the chip and a BK Precision Triple Output DC Power Supply 1651 to create the differential input. I then measured the chip's output against the wall ground because the two power supplies were plugged into the wall. I assumed that all the grounds were therefore the wall ground.

We have tried to eliminate all ground loops. Is there a way to verify this?

Click to expand...

The wall ground is a noisy ground and may not be connected to the power supply commons. It is never used for signal ground. You always want to measure to circuit common, not wall ground. The common of all your power supplies must be connected to the common of the amplifiers. It's generally not desirable to connect that common to the wall ground.

You can determine if there is a ground loop by opening the ground connection and then measuring across the connection with an ohmmeter (power off to all devices). If you read an open circuit then you have no loop.