INA333: INA333 output is not amplified and is odd on an oscilloscope for an EOG circuit

Hi Roshan,

There are a few comments/suggestions the circuit:

- Please add low-pass filters at the INA333 input with a low frequency corner to filter all the extrinsic high-frequency noise.  The EOG signal of interest is relatively low-frequency.  The differential capacitors of the low-pass filter will also help minimize the effects of the high-source impedance with the chopper instrumentation amplifier input bias (IB) current spikes. 

- The EOG signal is an AC +/- differential signal, therefore, the INA333 REF pin needs to be biased at Vs/2 = +1.65V to amplify the signal properly. Connecting VREF=GND will not work.

- The integrator of the feedback REF circuit needs to be referred to a reference set to mid supply or +1.65V.  See a similar integrator example below. You may need to adjust the circuit depending on your probe impedances and requirements.

Below is a link for reference design for an ECG application using the OPA2333 in a difference amplifier + ADC circuit.  I realize that your circuit is for an EOG application, and the reference design is for a different ECG application.  Nevertheless, please review this document in detail, as many of the concepts are quite similar. 

https://www.ti.com/lit/ug/slau516/slau516.pdf

https://www.ti.com/tool/TIPD116

Please let us know if you need anything,

Thank you and Regards,

Luis Chioye

Hi Luis,

Thanks for the detailed response, a few questions based on your explanations:

- It was always my understanding that the EOG was a DC signal, but it is made AC to remove the dc offsets, do you mind clarifying? Or do you mean that since I'm already AC coupling the circuit, that we treat EOG as an AC signal. So, would the other way to approach if I were to treat it as a DC signal not require any high pass after the first gain stage and just a high precision ADC?

- I understand the need of a low pass filter at input for the intrinsic noise, in that case I assume there is no need for a low pass after the first gain stage right? But it is also my understanding that placing any components in front of INA333 might result in an increase in offset voltage. Where IB current spikes coming from closing/opening of the front-end switches is converted across the mismatched input impedance into voltage causing an increase in the offset voltage. Hence doesn't it seem like adding an EMI (LPF) filter at the output of INA333 the better approach. But that brings me back to where I'm at with this weird output, only making sense to try something different.

- INA333 REF pin needs to be biased at Vs/2 = +1.65V to amplify the signal properly -> why can't it be biased after the amplification?

Sincerely,
Roshan

HI Roshan,

Roshan Sivakumar said:

It was always my understanding that the EOG was a DC signal, but it is made AC to remove the dc offsets, do you mind clarifying? Or do you mean that since I'm already AC coupling the circuit, that we treat EOG as an AC signal. So, would the other way to approach if I were to treat it as a DC signal not require any high pass after the first gain stage and just a high precision ADC

The EOG signal is a low-frequency signal. Most literature describe the EOG electrode signal in the low frequency range is DC 0-Hz to 50-Hz, or other articles define it in the range of ~0.1Hz to ~20Hz. The signal is relatively small in amplitude in the 100s of microvolt to millivolt range. Since the signal of interest is a small-amplitude, low-frequency signal, the user needs to eliminate any external EMI and high-frequency noise interference from the environment.  Using low-pass filters at the instrumentation amplifiers such as the INA333 is standard practice, as shown in the proven design listed on the application note.  

Roshan Sivakumar said:

- I understand the need of a low pass filter at input for the intrinsic noise, in that case I assume there is no need for a low pass after the first gain stage right? But it is also my understanding that placing any components in front of INA333 might result in an increase in offset voltage. Where IB current spikes coming from closing/opening of the front-end switches is converted across the mismatched input impedance into voltage causing an increase in the offset voltage. Hence doesn't it seem like adding an EMI (LPF) filter at the output of INA333 the better approach. But that brings me back to where I'm at with this weird output, only making sense to try something different

The low-pass filter at the instrumentation amplifier input is used to eliminate external or extrinsic noise, but not the intrinsic noise of the circuit. 

The original post circuit uses only resistors with no capacitors, which will not work. The high-impedance resistors will amplify the IB current spikes of the chopper amplifier and increase noise.  In contrast, the example provided on the reference design uses the high-impedance resistors with a differential and common-mode capacitors, creating a low-pass filter. 

The filter uses a differential 1nF capacitor across the instrumentation amplifier inputs and 100pF common-mode capacitors.  These capacitors at the INA333 inputs offer a low equivalent impedance at the instrumentation amplifier inputs, or a relative low-impedance path at the chopper frequency, and therefore, the capacitors help mitigate the effects of the chopper IB current spikes.  Also, the low-pass filter help reducing EMI noise at the inputs of the instrumentation amplifier, which is absolutely required, specially when attempting to amplify small low-frequency signals.

"The integrator of the feedback REF circuit needs to be referred to a reference set to mid supply or +1.65V" - I actually did that in my attached circuit, unless the one you shared is fundamentally different. Do you mind giving me some advice as to what to change based on electrode impedance measurements

Yes, I see that you referred the integrator to +1.65V.  The impedance of the electrodes is dependent on the electrode/probe type. If you provide the electrode impedance model, then we could check the integrator for stability.   

Roshan Sivakumar said:

- INA333 REF pin needs to be biased at Vs/2 = +1.65V to amplify the signal properly -> why can't it be biased after the amplification?

 Keep in mind that the amplifier output can only swing within the supply rails and requires a minimum ~50 millivolt headroom above or below the rail supplies, where the INA333 is powered with uni-polar supply of +3.3V and GND. The minimum output of the INA333 in this circuit is around +50mV.  Therefore, when using REF = GND on the INA333, this setup limits the output swing to only positive input signals, and limits the output range of the INA333.  This becomes an issue for small differential signals.  Also, the INA is required to amplify the voltage difference on the electrodes, which in most applications, the difference on the electrodes can be positive or negative voltage.

Thank you and Regards,

Luis

In this link, https://www.ti.com/video/6269751746001?context=1148465-1148591 they rather recommend a buffer at input, is this because of the differences in ina333 and ina826?

I'm using dry electrodes, so something like this is the equivalent impedance model, can you please inform me of the stability:

Best,

Roshan

Hi Roshan,

Dry electrodes may require higher input impedance, and in some instances, certain dry electrodes may not be able to drive a chopper amplifier. Non-chopper, CMOS or JFET input amplifiers such as the OPA392 can be used as buffers for these applications. These CMOS amplifier exhibit typical input bias currents well below ~10pA. Although the INA333 is a chopper, it still offers relatively low average input bias current amounting to ~70pA typical (200pA max). Nevertheless, if your specific electrodes require a very high-input impedance, you can certainly buffer them using a CMOS input amplifier such as the OPA392. 

The actual impedance (R, C) values on the electrode model above are required for the stability analysis. The impedances on the model are specific to the electrode on the application and need to be defined on your side.

Thank you and Regards,

Luis 

Would it make a difference to place the buffers before or after the input filters?

I am modelling a range of electrode impedances based on material and pressure that I will be dealing with, i'll drop in a reply when I get that.

HI Roshan,

Please place the buffers after the low pass filters, at the inputs of the INA333.

Thank you and Kind Regards,

Luis