EKG 50/60hz noise, hardware solution?

[saand]

Hi All,
There are a few posts on noise at 50/60hz in the EKG signal. I have also tested the shield-EKG-EMG system and have observed this noise on the analog output.

In developing the 3 signal board I am interested to know if anyone has resolved this noise with a hardware solution. Either with shielded probes or filters on the shield board.
I have seen a software solution to this however it would be nice to resolve in hardware for this design.

Thanks
Scott

[LubOlimex]

Hey Scott,

As I've mentioned it a few times before - the only way to completely remove the 50Hz/60Hz noise is to power from a device that is not connected to the electrical utility network of your country. Any laptop operating on batteries that powers SHIELD-EKG-EMG would exclude this hum.

Best regards,
Lub/OLIMEX

[peterisza]

Hey Scott,

I think your project is awesome.

Many ECG systems use notch filters or more advanced software solutions to eliminate the 50 Hz / 60 Hz noise from the ECG signal.

The noise can be present even when the system is wireless due to capacitive coupling.

Imagine that you put 0V and 230V at 50Hz between the plates of a huge capacitor. If the material between the plates is homogeneous, the complex voltage will have a linear gradient between the plates. Now imagine that there is a human between the plates, and you are trying to measure differential voltages on the skin.

Real life scenarios can be pretty much like the described situation.

The good news is that the human body is pretty conductive inside, only the skin is an insulator (see http://niremf.ifac.cnr.it/tissprop/htmlclie/htmlclie.htm).

One part of the capacitive coupling causes
  1. real noise in the differential voltage,
and the other part causes
  2. huge (300 mV) 50 Hz voltage swings in the common mode signal.

You can't really deal with the real differential noise other than filtering the record with software. You will have to pay careful attention to the common-mode rejection ratio. For example if the CMRR is only 60 dB, the error signal's amplitude will be 0.3 mV if the common mode signal is 300 mV.

The good news is that nowadays it's pretty easy to achieve a high CMRR in the hardware with the ADS1298 chip. The bad news is that a bad CMRR can come from differences in the input impedance of the leads too. The impedance of the skin and the electrodes can vary depending on a lot of factors (hair, age of electrodes, skin type, etc.). The electrode impedance together with the input impedance of the lead forms a voltage divider.

Let's consider a case where you are using a 10M pull-up resistor for lead-off detection, and one of the electrodes has a 10 kOhm impedance, while the other has 20 kOhm, and the common mode signal is 300 mV. 300 mV * (10MOhm / (10 MOhm + 10kOhm) - 10MOhm / (10 MOhm + 20kOhm)) = 299 uV, which will screw up the whole ECG signal. I almost forgot to mention that the capacitance of the shielded cables also counts.

You also want to reduce the common-mode signal itself. You need a good right-leg driver circuit to achieve this. The current limits described in the medical standards shouldn't be exceeded at any frequency, but it's a good idea to get the maximum out of it.

To sum things up, this is what you can do:

• Make the system wireless.
• Good right-leg drive circuit.
• Pay attention to CMRR.
• Use shielded patient cables.
• Don't use pull-up resistors for lead-off detection.
• Use a digital filter to remove the remaining 50 Hz / 60 Hz noise from the signal.

And here's an idea that requires a bit more research:

• Measure the common-mode signal with a low resolution ADC.
• For each device, do a calibration process to determine what kind of error signal is generated from the common mode signal. If you fit a second-order IIR filter, I think it'll be fine. After you have the filters, you can reproduce the error signal in the software and subtract it from the measured signal.
• Based on the electrical properties of the human body, one can even come up with a model that uses the common mode signal and the ECG signals to deal with the capacitive coupling problem with less distortion of the original ECG curve.

Peter