Can the corner frequency of thermal noise change with change in source impedance?
In EMG there is a certain level of thermal noise created at the electrode-gel-skin interface which depends on the skin-electrode impedance (also known as source impedance).
I have been able get the thermal noise data for only dry electrodes of different materials. Dry electrodes are the same electrodes in this case but placed without any Gel. At 100kOhms impedance the white noise is the same as any 100KOhms resistor but the different electrodes have different corner frequency and the exact curve of the 1/f noise on the graph is different as well. (Noise density per square root Hz on the y-axis and frequency on the x-axis, like in op-amp data sheets)
Will/can the corner frequency change if the source impedance is reduced to <5kOhms using a gel or the corner frequency and curve will be the same as 100kOhms data except that the noise density for white and pink noise will be reduced to a <5kOhms resistor level?
https://pubs.acs.org/cms/10.1021/acsomega.2c00282/asset/images/medium/ao2c00282_0008.gif The electrode-skin interface forms a galvanic half-cell (Geddes, 1972) It is significantly more difficult (and perhaps much more expensive) to collect the noise data at low impedance.
2 answers
The following users marked this post as Works for me:
| User | Comment | Date |
|---|---|---|
| tilly | (no comment) | Aug 31, 2023 at 10:12 |
Any time you have a dielectric interface ( ___ pF ) source resistance will always have an error in signal response until the electrode impedance is eqal or less than the sub-cutaneous impedance. This will affect the expected 1/f noise response.
Your goal is to reduce Rs and increase Cs with gel and physical restraints to reduce galvanic skin response(GSR) noise from modulation of electrode pressure which affects Cs mainly from the surface gap.
When the electrode is dry, it is the highest impedance and thus most sensitive to surface pressure (GSR) error and thus reduces SNR of muscle signals and makes 1/f noise more obvious.
EMG signals are far easier to detect than EEG signals where the problem exists. Your signal conditioner BW ought to match the signal spectrum for ideal transfer function and Rx sensitivity.
I CAN MODEL THIS ERROR IN FALSTAD'S FILTER SITE OR YOU CAN TAKE MY WORD FOR IT.
The thermal noise of a resistance depends only on that resistance. However, there is always some parasitic capacitance. That capacitance against the resistance causes a low pass filter.
So in a sense, yes, the frequency content of thermal noise does vary with resistance in the real world due to the inevitable parasitic capacitance.
0 comment threads