EMG design questions: capacitor in Rg, virtual ground noise

[Arti G]

Hello

I am designing EMG amplifier with INA333 and OPA4376. I use 3 dry electrodes made of stainless steel. I saw many designs and read a lot, but after making a prototype it wasn't working. If there is anyone with experience with EMGs it would be very valuable if could answer some of my questions.



1. I couldn't get any signal from first stage of amplifier - after INA333. I tried Rg values from 300R to 4,7k Ohms. I saw 2 different designs than most. One with 47uF capacitor in place of Rg (no resistor) and another with 300Ohm + 22uF in series in Rg. I couldn't find any explanation to it in datasheets or any other designs. Do you know purpose of it? Is there any article explaining using a capacitor in place of Rg resistor or adding capacitor in series to Rg? Should I care more about input bias? Any equation for choosing capacitor and resistor for proper gain?
A: 47uF https://hackaday.io/project/8823-super-simple-muscle-emg-sensor



"The super simple EMG circuit uses a capacitors reactive impedance attached to the IA gain pins, which sets a variable frequency dependant gain, thereby only amplifying higher frequencies, and thus creating a similar affect to a high pass filter. The capacitors impedance (which is inversely proportional to the IA amplification) is calculated from below equation

Impedance = 1/(2 ∗ PI ∗ Frequncy ∗ Capactance)"


B: 300R + 22uF https://www.dfrobot.com/product-1661.html

When I tried 300R + 10uF it worked! But I don't know why and if it is correct way to design. There is not enough explanation around on the Internet and many designs just use a resistor in Rg (no idea if these designs were checked in real).

2. Is voltage divider with buffer a good solution to supply power to INA and OPA? I saw that I have like ~200kHz noise on Vref and it brings noise to signal output. Any solution to this noise? When I removed OPA and used only divider without buffer then noise was gone, but divider without buffer is generally not a good idea.

I know that opams don't like high capacitance on output, but does it apply to buffered divider? I have to use decoupling capacitors on INA and OPA supply so it will add capacitance to buffer. Seems like no one cares about output capacitance when using opamp as a buffer to get virtual ground. Can input capacitance (two 10uF) influence working of the buffer?

Maybe I should change to low noise negative voltage generator?

[moffy]

1. If you can't get any signal after the INA333, why is that? You should at least be getting amplified noise. Are the supplies to the INA333 correct, is the output saturated at either 0v or Vcc?

IMPORTANT: I hope you are using batteries to power this circuit as it is attaching directly to yours or someones skin. You need to provide safe isolation and for a home project that means batteries without the charger attached.

How are you measuring the output of the INA333, with a multimeter? or do you have access to an oscilloscope. Before trying a number of random variations make sure the basics are working first otherwise you are likely to confuse yourself.

2. Puting a 22uF cap in series with a 300 ohm resistor for the INA333 turns it into a high pass filter. At very low frequencies the 22uF cap dominates and attenuates the signal (simplified) while at higher frequencies the 300 ohm resistor dominates and sets the maximum gain.

[jonpaul]

1. eliminate power supply, use two 9v batteries.

2. remove filter and rectifier, no lopass " antialiaising"

3. follow good low noise cable, earth and Shielding practice

Jon

[Terry Bites]

Electrical isolation of the amplifier and the measurment device is vital for good CMRR and safety. The recording device must not be mains powered. A total fail on both counts. The proper solution is to use an isolation amp between the amplfier and the measuring device. You need to use isolated power (or batteries) for the amplifer as well. The isolation also massively boosts CMRR. www.biosemi.com/publications/artikel5.htm (A Van Rijn's articles on this site are a good read) These days you digitise on the patient side and use iCouplers or similar to get the isolation. Loads of options on serial ADCs. You can get sigma delta ADCs which are very easy to use. Some even have isolation built in such as the AMC1303x from TI, that will lower the sytem costs. Using sigma delta lets you recover the analog voltage on the measuring side with out a DAC.  A simple lowpass filter works fine.

DC potentials wil ruin your day. Poor cabling and shielding arrangements will overload the amp with AC and/ or RF sending it off to the rails. All the CMRR in the world will not protect you from this.

So some options for the amplifier:

Keep the IA gain low. Say Av=5. Electrode materal and skin moisture create electrolytic cells. These may generate 10s of mV in cell potentials: www.pulseai.io/blog/why-electrodes-matter-electrode-electrolyte-interface
If your gain is high the differences in cell potentials may saturate the IA. That's likely your problem.

Note that connecting the output to a grounded device may (will) mess up the CMRR. You should use protection resistors in series with the IA inputs to limit fault currents to a value of 50uA or less. So Vss/50uA

The IA should be AC coupled to a gain stage- Av ~100 for rthe low gain option. You will get much better performance with a driven ref electrode. Note reduced 90% feedback to shield. 100% may casue instabilities.

There is a better way that allows for a high gain IA. A charge balancing RC network kills the cell potential effects. Integrator feedback provides the AC coupling function. Use the same shield and Ref drive in both cases.
The opamps can be anything, nothing special.  Look into contactless electrode systems.