5
\$\begingroup\$

Please note: Although this question mentions an electronic circuit depicting an EEG, I believe this is purely a bio-electrical question that is appropriate for this site, and should be answered by any battle-weary Biomedical Engineers that have had experience with EEG.


I stumbled across this DIY EEG Instructable, which is interesting and novel, but which has a number of drawbacks (from an EE perspective).

What surprises me most about this circuit is the constant amount of correction/filtering that the author is doing at each step:

  • Two 60Hz notch filters one at each "end" of the circuit, to reduce noise at a particular frequency (do we really need two?)
  • High pass filter for circumventing galvanic skin response
  • Low pass filter for waves > 30Hz
  • Yet another high pass filter
  • Op-Amp

To me, it makes sense to digitize the circuit upfront (by running the analog brain waves through an analog-to-digital converter, or ADC).

This leaves me feeling that perhaps there is some intrinsic/inherent value to EEG evaluation of analog (as opposed to digital) waves. So I ask: is there a compelling reason to tune/correct/filter raw analog brainwave/analog signals for EEG, or is it perfectly fine to digitize the waves and then process them?

\$\endgroup\$
  • 3
    \$\begingroup\$ Possible duplicate of To what extent does an ADC solve noise along an analog circuit? \$\endgroup\$ – Brian Drummond Feb 17 '16 at 11:21
  • \$\begingroup\$ For those looking for the schematic: cdn.instructables.com/FR5/NQTR/H3QFL7UP/… or the whole page: instructables.com/id/DIY-EEG-and-ECG-Circuit/step2/… \$\endgroup\$ – Jan Dorniak Feb 17 '16 at 11:22
  • \$\begingroup\$ You've asked this before, and your answer is the same. You MUST do analog processing to make the signal suitable for sampling by an ADC. Do the math. What happens when you have a 16 bit ADC covering 3 volts? Is there sufficient resolution to handle 0.1 mv signals? \$\endgroup\$ – Scott Seidman Feb 17 '16 at 11:43
  • 1
    \$\begingroup\$ This is sbolutly an ee quesion \$\endgroup\$ – Scott Seidman Feb 17 '16 at 11:43
  • \$\begingroup\$ Matlab can't "magic away" frequency aliasing or quantization error from digitized data. And the signals of interest are many, many orders of magnitude lower than the 50/60Hz line noise induced by the patient monitor leads (wires)... I'm under NDA so can't say much more, but real medical applications circuits use carefully designed printed circuit boards (not solderless breadboards!), several stages of analog front end processing before the ADC, followed by digital signal processing algorithms (prototyped in Matlab). \$\endgroup\$ – MarkU Feb 17 '16 at 21:01
6
\$\begingroup\$

You MUST do analog processing to make the signal suitable for sampling by an ADC. Do the math. What happens when you have a 12 bit ADC covering 3 volts? Is there sufficient resolution to handle 0.1 mv signals? – No. The LSB in such a case is about 0.7mV. At 16 bits, your signal would span a bit more than 2 LSBs. There must be some amplification, probably by a factor of about 500, to give you enough LSB's to sample effectively, and you need to condition the signal to remove offsets to be able to get a gain that high.

So, attempting to sample without analog preconditioning is probably misguided here.

There is some argument to be made that sampling unconditioned with 24-bit or higher ADC's is OK, but I think that an instrumentation amplifier input stage with modest gain, which is well optimized for high common mode rejection ratio, is a fine thing.

\$\endgroup\$
1
\$\begingroup\$

As a short, and somewhat opinion-based answer the following;

1) Is there a bio-electrical advantage to processing EEG as analog (vs digitized) signals?

No there isn't; at present everything is digitized so that's the way to go.

2) Two 60Hz notch filters do we really need two?!?`

A circuit would be helpful here; the link is kind of popular sciency. But as far as I can see, one is more than enough. Notch filters are typically used to filter out the noise from the mains. In Europe and Australia, the mains has a frequency of 50 Hz. In the US notch filters at 60 Hz are commonly used for the same reason.

High pass filter

Often used to get rid of baseline distortions (drift) and DC effects. Very commonplace.

Low pass filter for waves > 30Hz

Often used to get rid of anything uninteresting; a 30 Hz cut-off is OK for standard EEG recordings, but if you are interested in beta (16 - 31 hz) or gamma (>30 Hz) bands, you may want to use a higher cut-off.

Yet another high pass filter

Please post the circuit. I don't know why yet another one.

Op-Amp

Without it, no signal. Amplification is paramount.

is there a compelling reason to tune/correct/filter raw analog brainwave/analog signals for EEG, or is it perfectly fine to digitize the waves and then process them?

My preference is to record and store digitized raw (unfiltered) data. Software like Matlab are perfectly suitable to do the filtering offline. Filtering means cleaner data, but also loosing data. To prevent throwing away the baby with the bathwater, I would recommend to filter offline.

\$\endgroup\$
  • 2
    \$\begingroup\$ +1 For the record (I know you know), the obvious caveat to the final comment is that you need some form of hardware anti-aliasing (low-pass) filters in front of the ADC to prevent out of band noise/interfering signals from folding into your digital data stream. Don't just sample the raw full-bandwidth signal unless your amplifier or other component is going to guarantee low pass cutoff inside your sampled signal bandwidth (at least < 1/2 sampling frequency). \$\endgroup\$ – DrFriedParts Feb 17 '16 at 16:51
  • \$\begingroup\$ @DrFriedParts - thanks for the upvote (and that at Physics! wow, I'm proud ^__^). It's a good point. Anti-aliasing filters is definitely where my knowledge ends. I'll have a look into it and I am happy this question ended up here. \$\endgroup\$ – AliceD Feb 17 '16 at 19:34
  • \$\begingroup\$ I've rolled back the last edit to this answer as it was right the first time - Europe/Australia = 50Hz, USA = 60Hz. \$\endgroup\$ – Tom Carpenter Feb 22 '16 at 5:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.