[PIC]Is this design good to measure 2mV to 8mV ? [Load Cell]

[FvM]

Load Cell has three wires.

One wire already lost? Link says

Out-Wire: 4-Wire

as expectable. Respectively the differential output voltage is centered around half bridge supply.

 

[Okada]

The data provided to me is that Load Cell has 3 wires and one wire gives 2mV to 4mV. I don't have the load cell. The client is not giving full information.

@FvM

According to you if 4 resistors are used for the Load Cell circuit what voltage range I get as the output ?

 

[KlausST]

Hi,

What is the best way to design an amplifier for this using either single or dual power supply ?

Now with your three wire sensor (I still doubt that you get a 2mV ..6mV output signal referenced to GND. At least I´ve never seen such a sensor.)
there is no need for an instrumentation Amp.

A single non inverting OPAMP circuit can do the same. Simple RR with single supply.

Klaus

 

[Okada]

I understand Non-Inverting Amplifier but what is RR ? Rail to Rail ?

This is what I designed. It uses Internal Rail-to-Rail OpAmp in Non-Inverting mode. But what is wrong in it. My client says the adc value keeps fluctuating a lot.

They say that they are providing 2mV to 6mV to Op Amp input and adc value fluctuates a lot.

 

[KlausST]

Hi,

RR = rail to rail.

Maybe:
Bad layout. Noisy sensor signal. No low pass filters. Noisy supply voltage. Noisy reference voltage...

Klaus

 

[Okada]

This is the layout.

 

[KlausST]

Hi,

I assume Ground_bounce in the PIC.

All the display current goes through PIC´s GND pin.
Inside the chip there is a thin bonding wire with a noticable resistance.

I don´t know how much. Let´s say 100mOhms.

Outside the chip we have clean 0 V.
but across the bonding wire depending how much segments are ON a GND current is between 20mA and 70mA (estimated).
those 70mA x 100m Ohms give a voltage drop of 7mA.
This 7mV is where the internal OPAMP refers to. Outside maybe 5mV sensor voltage, referenced to internal 7mV gives an OPAMP input voltage of -2mV.
I assume this is out of range... in either case I assume it can not amplify it.

Now it depends on how many display segments are ON. And maybe this is what your customer sees. It is jumping from one value to the next.
Every time another display current, another PIC_GND current, another GND_bounce.

***
How to solve this?

You could use an external amplifier. This is not optimal, but an improvement.

Or you could use an extra display driver, then the display current doesn´t flow through PIC_GND pins. Causing no GND bounce.

***

Am I sure about this? Not 100%.
--> Measure GND_bounce. How? You can´t measure inside the chip?
True. But maybe you can switch one unused PORT pin as OUTPUT_GND. This means the PIC switches it´s internal GND voltage to this port pin. Now you can measure it.
Millivolt-meter between this port pin and the external GND plane.

Maybe you don´t see the calculated/estimated 7mV, but even if you see 2mV it gives a large error in OPAMP output.

Klaus

*******
Added:
Regarding your layout.
It seems there are two layers. And in each there is a copper pour. On one side GND and on the other side VCC. Is this correct?
If yes: Why are there still GND wires and VCC wires in both layers?
And: there are extremely thin wires. From power input JP1 to rectifier AC1 ... are they really that thin?

 

[BradtheRad]

This illustrates how 2-8 mV can be amplified by a transistor in common base mode. It is a concept, not a working circuit. It accepts an incoming signal which is within a few mV of ground. The output is always in the positive. It never wanders into the negative. No negative supply is required.

Usefulness depends on the ability of your signal wire to sink a few mA of current. It may require additional components, temperature stabilization, filtering, etc.

Resistor values are low because transistor junctions are operating near their threshold. High resistances would not work so well.

The bias voltage must be carefully adjusted. You must not let bias current burn up anything. (The bias resistor is shown as 20 ohms to indicate it works best through a low resistance.) Changing the bias raises and lowers the output. Perhaps a silicon diode can be used for creating a voltage regulation network.

For greater amplification, use a high gain transistor, or perhaps a darlington.

 

[Okada]

How about using this 24 bit ADC ?

**broken link removed**

Any cheaper solution ?

 

[Aussie Susan]

Adding bits to an ADC does not solve problems related to noise on the input signal. The output of the ADC can only be as good as the accuracy of the input voltage.
You have already stated that the initial input voltage is from 2mV to 8mV which corresponds to a range from 0 to 100. Therefore you are looking at measuring the 6mV range to an accuracy of +/-60uV.
Even if you amplify the signal to be from 0V to 5V, you still only have a range of 100 and therefore an accuracy of +/-50mV.
To measure a value to 100 steps requires 7 bits (0 - 127). Adding one or 2 more bits than that might help a bit so any 10-bit ADC will do the job. Adding more bits than that will simply take more time for the conversions and add cost (ADCs with more bits generally cost more).
Therefore I'd suggest you stay with a 10-bit (or whatever you have in the MCU) ADC and concentrate more on getting the signal as clean as possible. That means short traces that are well shielded for low voltage lines, very clean power supplies etc..
Susan

 

[Okada]

Ok Susan. I will try with external Op Amp.

 

[Easyrider83]

This is the layout.

Very unprofessional. I don't recommend you to give any advise before you will get more experience yourself.

 

[KlausST]

Hi,

I'd use an input range of 0...10mV,
and amplify it just by 500. (Or any other value to meet your ADC input range)
Adjust the sensor range to 2...8mV with software. This is easier, more stable and better to adjust (self calibration).
And maybe this gives the information to your microcontroller whether the sensor is connected (correctely) at all or not.

Use input low pass filter, RR I/O Opamp with low input offset voltage drift (noniverting circuit).
Maybe an additional filter capacitor across the feedback pin.

For sure still the GND bounce exists. But it isn't amplified anymore. Remaining some LSBs of uncertainty.

To further improve the quality you may test a difference amplifier circuit which refers to the internal GND (switched out via port pin).


Klaus

 

[Okada]

That is just the proto board for testing. it is not used for final product.

 

[FvM]

We haven't yet seen a datasheet or specification of a single ended load cell that could work with the circuit shown in post #24 an #26. In so far the discussion is somehow insubstantial.

 

[KlausST]

Hi,

I agree.

But we have to believe in post#22...

A couple of things are not clear: ratiometric output or not? Output impedance, desired signal frequency ...

Klaus

 

[FvM]

But we have to believe in post#22...

I don't believe it. There's no known load cell design that could achieve a single ended mV output in a three terminal configuration.

If the device is no regular load cell, we should at least see the datasheet.

 

[Easyrider83]

Well, I don't see any other options except using dual supplied opamp with low pass filter with 1000 gain. If you have AC voltage on input it is so easy to get negative voltage source. But analog ground should be separated from digital one and advanced VCC filtering have to be implemented also to reduce noise ratio. Even so, the noise filtering alghorithm have to be used anyway.

 

[Okada]

Wait. The client has made changes to the interface circuit. They now said that they have a 4 resistor network connected to two signal lines of Load Cell and two wires go to Excitation Voltage. I will update this thread after I get correct information about the interface circuit.

 

[Aussie Susan]

This thread was started near the beginning of October. Now at the end of November (nearly 2 months later) the client makes a fundamental change to their requirements!!!
I hope you are charging them full rates on a T&M basis.
Time to get a new client perhaps???
Susan