Small signal DC amplifier - please critique my design

[daqq]

Hi guys,

I'm making a small DC signal amplifier for strain gauges and such - the input should be somewhere about +/- 0.5mV. The amplification should be 11 (not a nice number, but it's OK). The differential input is designed to be floating - these are the only two connections between the sensor and the system - which is battery powered. The output is differential as well. The schematic is in the attachment.

The divider will be one of these: http://www.farnell.com/datasheets/652744.pdf - a fixed ratio resistor array.

The whole system is effectively just a non inverting op amp with feedback. A virtual ground is created by the pullup resistor R15 and the D13-D15 diodes. These also act as protection against ESD/overvoltage on the differential input (with the help of D11 and D12 they cover both of the input signals).

C12, C8 R10 and C13 should act as a filter.

A little leakage current from the diodes is expected, but since input is designed for relatively strong sources it should not of concern.

In simulation, the design work well when loaded with a 1k. The main source of DC error is the leakage current of the diodes at higher temperatures. I'm also concerned about noise from them.

In the final design there will be more decoupling caps, a voltage regulator and a common mode choke for the input. In the layout I shall take into account http://ww1.microchip.com/downloads/en/appnotes/01258b.pdf as well as other good practices. (I hope )

My question: Is this a good/acceptable design? Where is there room for improvement.

Links for convenience:
Op amp: http://cds.linear.com/docs/en/datasheet/1050fb.pdf
Divider: http://www.farnell.com/datasheets/652744.pdf

Thanks,


David



EDIT: The power should be +5V. Sorry.

[pinkysbrein]

Central semiconductor has some low cost ultra low leakage diodes available on mouser btw.

This will get you pa range leakage.

[daqq]

Thanks, I'll look into those.

[Gall]

How precise should it be? The weak point of all such designs is that the input is not really differential. The parasitic capacitance between input wires and power wires is usually equal for both input wires but the input impedance is not. That's why the input cable may act as an antenna even it is well made. It's Ok for a shielded cable but it is not for a twisted pair.

[fcb]

I'd put an RC filter in-front of the opamp and possibly ditch the input protection or move it behind the RC filter.

Nothing inherently wrong with the design - depends what your driving, a micro ADC perhaps?

A gain of 11 seems low for strain gauge input, and there are some natty little parts that make life easier.  Try the INA125, it's worked for years in a strain-gauge amp in something I designed.

[calexanian]

I am most likely wrong cause these circuits are not my specialty but the first thing I see in the input clamping 1n4148's have a drop of half a volt or so and you only have a gain of 10. You will shunt your input signal before you get anywhere near your full output swing, Unless you are also using it as a limiter or have a design where that would be advantageous. Also depending on what your output impedance the 10 ohm output resistor kinda rubs me the wrong way. If you want the protection of some resistance there fine, but move the feedback loop to after it so you will have no drop across it for precision work. That chip most likely has output protection, and for high stability DC stuff it would not hurt to throw a zobel network on it to bat down any high freq rings from long output wire runs and stray capacitance. If you are only concerned with C it will be fine. If you are sampling fast or have a finicky ADC after it for get that plan as it will cause settling problems. In fact, better forget everything I said.   O0 Other than that I got nothing. If I missed any other huge things then face palm myself. 

[daqq]

Hi, thanks for the replies.

How precise should it be? The weak point of all such designs is that the input is not really differential.

Well, it should be as good as it can be  As to specifics, well, it would be good if out of the 0.5mV I could get as much as possible - limit the noise.

Nothing inherently wrong with the design - depends what your driving, a micro ADC perhaps?

A 16 bit external ADC. Its input impedance is really high (10Meg+).

Try the INA125, it's worked for years in a strain-gauge amp in something I designed.

I'll look into it, thanks. Though it does seem to have a far bigger offset voltage.

calexanian: The input is meant to be +/- 0.5 mili Volts. Those diodes are against overvoltage and ESD.


I've built the circuit on a breadboard, and there seems to be a lot of noise. I'll experiment a bit more with this.

[Gall]

Try the INA125, it's worked for years in a strain-gauge amp in something I designed.

I'll look into it, thanks. Though it does seem to have a far bigger offset voltage.

Maybe INA128, INA126 too.

The best differential input design is the classic three-opamp differential amplifier. Exactly this:

Its output may be used as differential, too. I once even saw this made from one TL064 and this was medical equipment (ECG).

On a breadboard, try to "shield" all sensitive wires by putting extra ground wires above them. This is the only way to build a high-sensitive circuit without etching the proper PCB.

[dannyf]

Is this a good/acceptable design?

Hard to tell as it is very application specific. Your design, inspite of the "differential input" designation, seems to be single-ended - utilizing a true differential (bridged) configuration would be helpful.

Where is there room for improvement.

Assuming that the strain gauge signal comes off a bridge configuration, I would do a few things:

1) do away with the diodes and utilize a regulator to create a virtual ground. Something like a 7805 would work here.
2) balance out the non-inverting input: use a 1k//10k pair to ground to minimize out dc errors caused by input bias current;
3) lift the non-inverting input off the ground;
4) unless you are absolutely sure of the load, put a small serial resistor on the output.
5) I would pick a jfet opamp (a single) so you can adjust dc output.

[Marco]

The best differential input design is the classic three-opamp differential amplifier.

Can't you go one better by making an auto-zero version of that? (Which google tells me do seem to exist.)

[Gall]

The best differential input design is the classic three-opamp differential amplifier.

Can't you go one better by making an auto-zero version of that? (Which google tells me do seem to exist.)

It is usually done by input chopping. The input is short-circuited for a very short time, the output is measured (i.e. stored in a capacitor) and then subtracted from the regular output. The idea is best illustrated with this schematic by Maxim:


Since this is not trivial, it's better to use a readily made IC for this purpose. I believe Analog Devices, Texas Instruments, Linear Technology and Maxim manufacture them.