Comments on this circuit please

[Morcego]

Hi all,

I want to build a simple loadcell amplifier with single supply. I had designed this circuit. Since I dont have that much experience, I will like to ear from you with more design experience.

The power supply will be 5 volt and the gain set to have a full scale of 4,096 volt (the output will feed a ADC with 4,096 volt reference).

One of my questions is regarding a low pass filter to suppress RF and the mains noise (50hz) I was thinking in about 40 hz. The filter is not in the schematic, because I dont know the best way to put it: in the loadcell sense outputs (before the AD623) or at the output (pin6). I had see in the net both designs.

Thanks in advance for any help

M.

 

[cbarn24050]

It shouldn't take you long to look up a circuit.

 

[itsmoked]

It is a time-sucking chore to proof a design like this.

I see pots which can sometimes be noise/drift problems but maybe not.

Filter wise can't say without understanding the physical layout and the usage, environment, etc.

Often the best starting point is as cbarn mentioned, find an appnote either by the bridge company or by experts on the devices to be used like Burr-Brown(slurped up by TI), or Analog Devices, or Linear Tech, or National. Their designs will generally help you avoid some pitfalls.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[jimkirk]

I agree with itsmoked about those pots. For the offset you need to be sure that the pot has the same temperature coefficient as R3. An alternative is to just have the pot span the whole voltage, that way the tempco doesn't even factor in.

Is the Vcc supply stable enough to be a reference for offset? You might need a reference IC.

It's worrisome having the absolute gain controlled by the resistance of a pot. Noise/drift.

What accuracy for gain and offset do you need?

I remember a Bob Pease article showing a technique where you have a fixed resistor (your R1) get you within a couple percent of the desired gain. Then with a handful of resistors in a network of links that can be cut you can hone in the accuracy. Takes something like 4 resistors to get within a small fraction of a percent. Be sure your circuit is thermally stable before you adjust it. Some measurements tell you which links to cut, re-measure, cut more links, etc. Cheaper than a pot, more stable (again, check the tempco of the resistors) and will not suffer noise and inadvertant adjustments like a pot will. This could work for both your gain and offset, but note that sometimes those things can be tricky, depending on how your circuit is designed, they can be interactive adjustments.

For a filter, keep in mind that it only removes noise that occurs before it. Any noise picked up after the filter will not be affected. Also consider the impedances involved. A simple RC filter with a reasonably sized 0.1uF plastic cap requires a 40K resistor. It's not something you can just drop onto the output of an opamp.

Also, it will take a very high order 40 hertz filter to remove any significant 50 Hertz. If that's really a problem, I'd consider a notch filter.

For the RF stuff, like Keith said, can't tell from here. Design for the worst, it's generally easier to remove stuff than add it. Is this a one-of, or are making thousands of these? It will affect your approach.

 

[Comcokid]

If I was doing this, I would use something from the AD77xx family. However, you would need a micro to setup and get the data from the chip. These parts have the programable gain, ratiometric correction, and filters for these applications. You could also buy one of the demo boards for a chip from this family which have software and serial interface for using a PC.

Otherwise, stability of your VCC, and TC of your pots and resistors if you expect the ambient temperature to change any. You could also put some RC in series with pins 2 and 3 of IC1 to remove any high-frequency noise that might happen.

 

[felixc]

Since you feed an A/D converter there must be some microprocessor in the rest of your design. I suggest that you work with fixed value resistors to a "safe" approximate gain,(maybe even using instrumentation differential amps that have a fixed internal gain to bring the voltage output to a range that will be less influenced by external factors) and use calibration values stored in non-volatile memory, to rescale the readings from your load cell into the real-world values that you will want to read, and get rid of the pots.

 

[richs]

Hiya-

I can't give a lot of references to op amp design for this
particular set of criteria, but, in general, I have found
the app note from microchip.com AN695:
"Interfacing Pressure Sensors to Microchip’s
Analog Peripherals"

To be of interest and it does discuss some of the tradeoffs
in external vs. software filtering.

You might find it of interest too. The microchip site is:

http://microchip.com

and you can do a search for the app. note.

Hope this helps.

Cheers,

Rich S.