TI E2E Community
Precision Amplifiers Forum
INA333 output problem (EMG active electrode)
Hi,I'm designing an EMG measurement device to control a robotic hand, for a research project. The instrumentation amplifier I'm using as a preamplifier is the INA333, since is a widely used biopotential amplifier due to its characteristics. I'm having some trouble with it, the output signal is not as it sould be. I've done some tests with OrCAD Pspice, using sinewave signals as the inputs. Also I've done tests with the real circuit assembled in a breadboard, using sinewave signals too. The results are the same, the amplifier only outputs the positive (or negative if the difference is negative) half of the signal. This happens when I use a single power supply and a reference voltage with a value of half the power supply. When I use a dual power supply and a reference of 0V, everything works as expected and I am even able to measure the EMG signal.
I don't know what is the problem. The schematic is based on the ECG one that is in the datasheet of the INA333, and I think it's OK. I've attached the circuit schematic and the output signal generated by Pspice, in case you find any mistake that I've missed.
Thank you in advance, and excuse me for any mistakes in the text as English in not my native language.
Notice the data sheet CMV specification:
"Common-mode voltage range V CM VO = 0V (V–) + 0.1 (V+) – 0.1 "
Regards, Neil P. Albaugh ex-Burr-Brown
Neil is correct that the common-mode voltage range of the INA333 is being violated. Try increasing the offset voltage on both V3 and V4 to 2.5V and you should see proper operation of your circuit. This means that inputs of the opamp must be biased to a DC voltage within the common-mode range for proper operation of the INA333. The ECG circuit in the INA333 datasheet accomplishes this by using the right leg drive circuit (RL output) to bias the body to a known potential.
Analog Applications Engineer
PA Linear Apps
Hi,You are right, guys, that was the problem. I feel so stupid right now, the only thing I needed to do to solve the problem was to read the datasheet carefully. I've done some simulations with the electrode-skin offset of 300mV (if I use an offset of 2.5V, the protection transistors chop the signal to an amplitude of about 600mV) and everything works fine.
In my breadboard design I have built the RLD circuit (in the attached schematic you can see the Vcm output that goes to the summing amplifier of the RLD) to compensate the 1.5V of the 50Hz common-mode, so I'm sure that if I would have tested the prototype with the real EMG signal and not only with a function generator, it would have worked properly.
Since we are talking about the RLD circuit, I have another question. The attached schematic corresponds to an active electrode, that is, an electrode that includes the user protection circuit, the preamplification stage and the circuit to remove the DC component of the signal. The common mode voltage of the instrumentation amplifier has to be buffered prior to the summing amplifier. My question is, is it better to have the buffer in the active electrode to reduce the effects of the motion artifacts, or is it better to have it installed on the main analog board?
Thank you very much for your answers, you really have helped me a lot.
Because instrumentation amplifiers can be sensitive to parasitic capacitance on the gain resistor pins I recommend keeping the common-mode buffer close to the INA333, this will help prevent the capacitance from a cable from affecting the stability of the INA333. Also, this will allow you to use 1/2 of an OPA2333 that is shared with the DC Servo circuit as is shown in your schematic.
Hopefully this helps!
Hi John,You're being very helpful, thank you very much. I will follow your advice.I have one more question, and the last one I hope. As the INA333 has RFI filtered inputs, I think I don't need the C1, C2 and C3 capacitors (they act as a RFI filter). The INA333 inputs are also diode clamped to the power supply rails, according to the datasheet, so my guess is that also the user protection transistors can be eliminated. Maybe a couple of 100k input resistors will be enough to protect the user in case of an electrical failure. What do you think?Regards,Alvaro.
Indeed, the INA333 has RFI protected inputs, this device shows a high level of EMI rejection over our tested frequency range of 10MHz to 6 GHz. However, it may be useful to leave the footprints for your input filtering caps on your pcb in case your testing shows additional filtering is required. If you are not in a space crunch, I don't think it hurts anything to leave the footprints on the pcb and not populate the capacitors if you find you don't need them. You do bring up an interesting point however, EMI rejection is measured as the change in the DC offset voltage at the output due to an applied EMI signal at the inputs (think of the opamp or instrumentation amplifier acting as a rectifier for the signal). Because your circuit uses a DC servo around the INA333, it may be fairly resistant to EMI-induced changes in the output offset already, I'll have to investigate this further in our labs here before I'm positive on this though.
You are correct that the INA333 has input ESD diodes that will clamp input signals to approximately one diode drop past the rail voltages, however because they were not designed for patient protection I can't recommend them for that purpose.
Hi,It's been a long time since my last post here and first of all, I want to thank you guys for all your help. Thanks to your responses, I have built a fully functional EMG prototype and, in my opinion, it works really well. Here is an image taken from an oscilloscope where you can check the quality of the EMG signal.
Although the signal quality is good, and the signal serves my purpose of controlling a robotic hand, something weird happens at the output of the circuit. The signal is referenced to a voltage level of 2.5V to have the full signal range using a single supply voltage, instead of using a dual supply voltage. Accordingly, the output signal should be centered around 2.5V. However, what happens at the output is that when the gain is increased (increasing the value of the potentiometer of the second amplification stage), the signal amplitude is increased (as expected) BUT the reference value drops. If the gain is high enough, the reference voltage reaches 0V.
I've made some simulations in OrCAD and everything works as expected: the reference voltage stays at 2.5V and the signal amplitude changes with the variation of the potentiometer value. Knowing that spice components are ideal, I think the problem is related with the non-ideal behavior of OpAmps. Maybe it has something to do with the amplification stage impedance. Or maybe the reference level at the output of the first amplification stage is higher than the reference level of the second amplification stage (2.5V) and that negative difference is increasing as the gain increases.
The total gain of the circuit is 500V/V and my idea was to split the gain as follows: 2V/V at the instrumentation amplifier, 5V/V at the first gain stage and 50V/V at the second amplification stage. Due to the gain-reference problem, the gain in the real prototype is split as follows: 50V/V at the instrumentation amplifier, 5V/V at the first gain stage and 2V/V at the second amplification stage. The operational amplifiers used in the gain stages are OPA2188. I've attached the circuit schematic so you can check it.
I have the feeling that I am missing something obvious but I'm not sure what the problem is.
Usually in these types of inverting AC gain stage / filtering systems, each stage is AC coupled from the stage before it. This prevents DC errors from getting gained up and causing issues.
This is likely the case in your circuit, so to quickly test if this is the problem or not please try placing a 1uF capacitor between U5B pin 7 (output) and R19 (10k). Then place another 1uF capacitor between the U6A output and R21. This will AC couple these gain stages from the rest of the signal chain.
I'm not sure about the bandwidth requirements of your system, but these 1uF capacitors form high-pass filters with the feedback impedance. If the 1uF capacitors fix the issue then a more appropriate value can be determined to meet your actual signal needs.
Let us know your results.
Best Regards,Collin WellsPrecision Linear Applications
Regards,Collin WellsPrecision Linear Applications
Sorry for my question inside another thread but is about the presented circuit of the OT. All the filters are referenced on GND and not on the 2.5V Vref. Is that correct? Or should the filters use the 2.5V as reference? Thank you!
Are you referring to R9, R11, C11, and C13 in 0143.EMG_circuit.pdf?
Precision Analog Applications
Yeah, i am referring exactly to these components.
Any comment on that?
Luis Filipe, those components are grounded. The reference is connected only on the + pin of the amplifiers.Thanks for the reply Collin. I haven't tried that solution yet, but I'm pretty sure it's that. I'll try it in the next version of the circuit. The current prototype is usable if I set the gain so that the reference it's at 2V.
Yea, and thats why i am asking if that is correct. As your filter is using Vref as a reference, you should threat it as your "ground" when designing the filters. As voltages are always relatives, if we consider your Vref as 0V, the GND is -2.5V. So lets assume that you got a 2.5V, a -2.5V and a center reference instead of yours 5V, 2.5V and 0V. What you are doing is placing a -2.5V at your filter what makes your poles different than you calculated. Or i am missing something here?
I simulated this in TINA-TI and found no difference (ideally). However, including an impedance in series with the 2.5V reference degrades the filter performance. Additional complexities come into play depending on whether or not the reference sources and sinks current. If the reference doesn't sink current the output impedance may not be well defined which could vary the filter's response. One possible solution is to place a large capacitor on the output of the reference, but viability will depend on the reference device. Therefore, if you find the circuit to work satisfactorily, it is my recommendation to use the circuit as shown.
I hope this helps.
Makes sense. But still they do not have the same transfer function for sure. Take a look at the TINA-TI steady state response of both filters. They have the same bode curve, but not same transfer. Anyway, if i cant create a low-impedance reference it will still be a problem. On my design i am using a voltage follower but with not output filter, just the decoupling capacitor. Is that enough as a low impedance reference?
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.