[SOLVED] INA111 JFET instr amp application noisy?

I'm making a 16-bit app. I have an AC signal of 100 KHz, 5 mV from a sensor at 1 meter. Noise measured with an oscilloscope is 300 uV. After using INA111 at 1 gain I have 2000 uV noise. My V+ rail (12V) has 4000uV (I guess 500 KHz) noise, uses the LM317. Should I be expecting this?

I never thought the power supply would be critical for a 16-bit app and and instr amp; I've never seen appnotes or books bothering with other than LM317. Should I be expecting that noise from the LM? Should I use other regulator?

I also see some strange high-freq noise from the supply electrolytic capacitors. Is this a sign that they may be damaged? They're 8 years old. Is it a good practice changing them?

INA amps don't perform that well a those high freq. I have used the INA2126 and INA126 for years
on intrumention projects for aircraft at max 100hz. Try decoupling caps on each supply leg of amps. .1uf to gnd.
Keep as close to amp as possible. Really need more info to help. E-mail schematic if possible
Lloyd

The schematic is like this...

L1 and L2 are actually a single inductor coil, whose center is attached to ground. I'm trying to measure an AC magnetic field at 100 KHz (can make some effort to lower that to 40). I don't know what to do with cable A and cable B. I could either measure close to the inductor coil and send the amplified signal two meters away, or let the cables run for two meters. I'm just afraid my 5 mV signal will become swamped by noise and cable inductance (the coil resistance is 5 ohm I think). Is an instrumentation amplifier good enough for this? I'm considering using the AD8523, rated for 10 MHz. Do you know any other choice I might consider?

The term "noise" is far from clearly describing your observation, I fear. What do you see? Amplifier noise? Interferences from switching power supplies or AM radio? Depending on the nature of the unwanted signal, the measures to reduce it would be quite different. True amplifier noise of INA111 isn't very low however, particularly at unity gain. But it's impact depends on the noise bandwidth of your signal processing system. If it's critical, a lower noise amplifier may be preferable.

I see 2 mV noise which seems high frequency, and 2 mV looks like 60 Hz. It looks a lot like the supply noise, but I thought the op amp rejection ratio would take care of that.

If you see 60 Hz "noise", it clearly can't be introduced by the supply voltage because you have 95 dB PSRR at 60 Hz. So it seems to indicate, that you have built another issue into your circuit. But high frequency PSSR is rather low, a good power supply filtering is needed, as already suggested.

Looking at your schematic, I can't determine how "AGND" is connected to the amplifier circuit.

If I turn the supply off so that my circuit runs on the 10000 uF capacitors alone, the 60 Hz noise goes away. Maybe my sensor is picking up the supply magnetic field. But it's 2 meters away.

Can you advice on whether it's better to amplify the signal at the sensor or at the ADC? My shielded cable has 5 nF capacitance and the sensor is 5 ohm. However, I can't shield the INA111 at the sensor.

I did forget the AGND pin in the schematic, but the circuit is OK.

I just found out that if I connect all three points of the oscilloscope to ground, I get the 60 Hz noise, like shown in the pictures. The scope is at 1 mV/div..

There are many current sense and mag field sense I.C.s that have a low dc output that can be more easily amped.
As for cable noise try the bootstrap trick (bias the shield ) some INAs have a bootstrap output just for that reason.
INAs by Burr Brown ---I think Texas Instruments bought them out but still have the INAs avail.

I just found out that if I connect all three points of the oscilloscope to ground, I get the 60 Hz noise.

Click to expand...

I don't know why your oscilloscope has two ground clips, but it's apparently a case of a ground loop, most likely through protective ground. It can be avoided by redesigning the grounding scheme. You should also use reasonable gain with the INA111 to increase the output signal.

P.S.:

As for cable noise try the bootstrap trick

Click to expand...

I think you're on a wrong track. The problem doesn't lock like "cable noise".

I don't understand too much about these kind of ground loops. I thought that it was an in-circuit concept.

Is this a solution? I'm not sure I understand it. Does it mean that I can connect the ground clip of my oscilloscope probe to protective ground with a capacitor and a resistor? The circuit is a modification from figure 7.69-7.70 from Horowitz & Hill 2nd ed.

Do you know of some app note with a solution?

Ground loops with single ended oscilloscope probes can't be avoided, if the circuit has a protective earth connection (and the oscilloscope too, for electrical safety reasons). But it's a problem in oscilloscope measurement, not in normal circuit operation.

Possible solutions are:
- using a differential probe
- breaking the ground loop by disconnecting one of two protective earth connections (if acceptable according to safety rules)
- breaking the ground loop for AC by a common mode choke, e.g. a toroid core at the probe cable

It should be noted, that high frequency common mode interference are injected in a single ended measurements even without a protective earth connection respectively ground loops. Common mode chokes can partly reduce it, but for sensitive measurements, differential probes are mandatory.