29
\$\begingroup\$

When recording electrical signals from cells (in a dish or inside a living human or animal body), one major problem is to increase the signal to noise ratio.

These signals are usually in the 10uV to 100mV range and are generated by very low power sources that can yield currents in the order of nanoAmps.

Often signals of interest fall within 1Hz-10KHz range (most often 10Hz-10KHz).

To make the matters worse usually there are lots of noise generating tools that are necessary to be around (in the clinic these are other monitoring, diagnostic and therapeutic devices in the lab these are other monitoring, scientific devices).

To reduce the impact of noise and increase the signal to noise ratio, there are a few generally applied rules like:

  • If possible use a current amplifier (often called head-stage), an amplifier with very high input impedance and rather low voltage amplification or even no voltage amplification. very near to the signal source (body).
  • To connect the source (recording electrodes) to the first stage amplifier (head-stage) use wires that don't have shields (to avoid capacitative distortions of the signal).
  • Avoid ground loops
  • When possible use differential amplifiers (to cancel the induction noise from the electromagnetic sources around).
  • Always use Faraday cages and grounded shields (usually Aluminium foils) to cover the signal source and anything connected to it (body, equipment ...).
  • You can't do this without proper filters (usually a 10KHz high cut and a low cut that depending on the signal may be anywhere from 1Hz to 300Hz )
  • If you can't get ride of the mains noise (50Hz or 60Hz in different countries) and only if your signal covers that range you can use active filters like Humbug http://www.autom8.com/hum_bug.html

My question is: Are there any other suggestions that I missed? Is any of these suggestions flowed or wrong?

Usually people in this fields (like me) do not have formal education in electrical engineering and sometimes there are myths passing from a teacher to student generation after generation without proper evidence. This is an attempt to correct this.

EDIT:
- if possible use batteries or very well regulated power supplies in all your devices, including any pumps, microdrives, monitoring devices, even you can put filters on the mains of your computers (although this usually is not a serious issue).

\$\endgroup\$
  • 4
    \$\begingroup\$ Someone talk about driving an active shield, such as those used in EKG cables, and line termination. I'm tired. This comment will self-destruct within 48h. \$\endgroup\$ – tyblu Jan 18 '12 at 8:12
  • \$\begingroup\$ related: electronics.stackexchange.com/questions/933/… \$\endgroup\$ – davidcary Jan 20 '12 at 12:12
  • \$\begingroup\$ "usually Aluminium foils" I would think copper foil because you need to solder it to ground, no? \$\endgroup\$ – endolith Dec 9 '15 at 15:26
13
\$\begingroup\$

Driven shield

It is possible to use shielded wires between the electrodes and the pre-amp without a lot of influence from the shield's added parasitic capacitance (your 2nd dot). The signal itself won't be hurt much because it is very small compared to the common-mode component. To understand this, imagine a tiny differential signal on top of a much, much larger common-mode signal (mostly caused by 50 Hz or 60 Hz mains voltage) and a DC-to-low-frequency component caused by the interaction of the tissue with the electrodes and the body itself. As far as I understand the issue, the interference coupled onto the signal via the cable's capacitance is much worse than having the signal itself fed through the cable capacity.

The trick is to actively drive the cable's shield with the common-mode part of the signal instead of connecting the shield to the pre-amp's ground. Some years ago, I've built such pre-amp with an active guard and was able to use shielded wires as long as 2 m between the electrodes and the first stage of the amp. The schematics can be found in this thesis (not mine, but conveniently includes the most interesting schematics of my EMG amp). Please see fig. 8.7, 8.8 and 8.9 and all the stuff around them in chapter 8. Fig. 8.12 discusses how interference is capacitively coupled onto the signal of interest. Sorry, the thesis is in German, but I hope the images and schematics are international.

A good place to pick up the common mode signal is the "middle" of the gain setting resistor of the initial InAmp (again, see the thesis linked above).

Driven right leg

The right leg is used as a reference to measure signal on left leg, left arm and right arm.

The concept of a driven shield can be extended to actively drive the patient, and the connection is made at the location used as a reference for the signals to be measunred, which is the right leg. This is known as a driven right leg (DRL); there's a good discussion about DRL amps in this article by EDN.

If your measurements are not taken from a human body but from some cells in a dish, you can probably put the DRL electrode onto the bottom or into the jelly / growth medium, close to where your reference electrode sits. This way, you use the same strategy as you would in the sense of a DRL setup.

Notch filter

Also, If the hum is really bad, you can put a notch filter at 50 Hz or 60 Hz into the signal path, but this will also hurt the signal of interest.

Very important safety note: The electrodes must not have any direct galvanic connection to protective earth (PE). This is necessary because once the patient gets connected to a potentially lethal voltage by a fault in another device around the lab, the fault current will have a very good path through the patient and via the electrodes to ground. When talking about a ground reference around the electrodes or the pre-amp, be sure to make this a ground referenced to the pre-amp only and not to the real ground usually known as PE! This usually requires an isolation amp somewhere around or just past the pre-amp, or a digital isolator if you wish to have the ADC close to the pre-amp. More about this in DIN EN 60601-1 and other relevant standards.

\$\endgroup\$
9
\$\begingroup\$

1. Use an Instrumentation Amplifier as the pre-amplifier (with right-leg drive)

An instrumentation amplifier, among other things, has a very high input impedance. This is ideal for measuring small currents. See the datasheet for the INA128. Page 11 has a reference schematic (attached below) which is similar to what you are looking for.

Reference schematic from INA128 datasheet page 11.

2. ALWAYS use power supply isolation for biomedical instrumentation!

Use a power supply isolation IC. See some examples from Maxim.

3. Use an active filter

Use TI's free FilterPro software to easily design an active amplifier for your desired frequency range. A Sallen-key band pass filter is easy to implement.

4. Digitize the signal and use DSP for additional filtering.

Use and ADC or an oscilloscope or a digitizer to take the signal to the digital domain where you can try out a variety of DSP techniques. A mains noise band reject filter can be easily done in software, for example. A book on the topic might be helpful. Also, don't forget to use digital isolators on the ADC outputs. ADUM1100 is an example.

\$\endgroup\$
  • 3
    \$\begingroup\$ As well as instrumentation amplifiers, look at electrometer amplifiers-they are designed for exactly this kind of thing \$\endgroup\$ – D_Weight Jan 19 '12 at 8:59
  • 2
    \$\begingroup\$ @D_Weight, searching for that made me come across this chip. LMC6001. Great find- thanks for pointing that out. \$\endgroup\$ – jeep9911 Jan 19 '12 at 17:33
  • \$\begingroup\$ I'm looking at this in detail now, and that's actually not a great driven right leg circuit. The idea of a driven leg circuit is to decrease the effective impedance of the right leg (ground) electrode, and the 390k resistor just blows that away. heartrhythmuk.org.uk/files/image/Case%20Reports/… shows a better arrangement, with that resistor (which is for patient protection if the amp is saturated) being inside the feedback of Rf. That said, the circuit you use pops up all over the place, and I suspect they all come from the same (wrong) source. \$\endgroup\$ – Scott Seidman Mar 19 '14 at 16:52
  • \$\begingroup\$ @ScottSeidman If the purpose is to decrease the impedance of the person relative to ground, to swamp out the Cbody capacitance, why not just ground the patient? Also, the common-mode voltages could be 100 VAC, no? How does an op-amp that can only drive ±15 V cancel it out? \$\endgroup\$ – endolith Mar 4 '15 at 21:53
  • 1
    \$\begingroup\$ @endolith - common mode voltage we're talking about is well within rails. The amp inputs are usually diode clamped to the rails, and common mode input ranges for the In-Amps are relatively small. Larger transients are dealt with by spark-gaps (ever wonder how an ECG can survive defibrillators?). The purpose isn't to lower impedance wrt ground, its to lower the effective impedance of all electrode-skin interfaces, and thus make them closer in value (it's the electrode contact mismatch that's the issue). See elastyc.unimore.it/fonda/ELBIOM/… \$\endgroup\$ – Scott Seidman Mar 5 '15 at 13:33
5
\$\begingroup\$

You might be able to use a lock-in amplifier.

It is not a general method you can apply in any case, but if you can, it gives you unsurpassable results. It requires you to modulate the original signal (e.g. if it is an optical signal, by a chopper wheel). Due to the modulation of the signal it is only useful for signals that change much slower than the modulation.

The benefits, however, are stunning. Using Lock-in amplification you can recover signals whose amplitude is orders of magnitude BELOW the noise.

The principle:

  • The original signal is modulated with known frequency and phase.
  • The detected signal (plus lots of noise) is amplified and multiplied with a rectangular signal of same frequency and phase and then integrated (phase sensitive detection). Almost all the noise is canceled.

I think searching the web for "lock-in amplifier" gives you enough more detailed descriptions.

\$\endgroup\$
  • 1
    \$\begingroup\$ Several kinds of optical imaging and brain imaging techniques have signals well below noise, but give pretty good results after averaging many trials. Showing the part of the brain active milliseconds before a person pushes the "A" button or the "B" button, etc. \$\endgroup\$ – davidcary Jan 20 '12 at 12:05
  • \$\begingroup\$ Sure, that averaging over repetitions works nicely in cases that you can afford it. In some situations (and of course in the ideal conditions) you want to be able to see the signal in each and every trial, though. \$\endgroup\$ – Ali Jan 20 '12 at 12:50
  • \$\begingroup\$ @davidcary: a lock-in amplifier can bee seen as a kind of averaging over many trials, but its more than just that: It subtracts also the "offset" (low frequency noise) that does not belong to the signal of interest. Especially effective if there is 1/f-noise. \$\endgroup\$ – Curd Jan 21 '12 at 0:13
3
\$\begingroup\$

I would change 2nd bullet to: "If using shielded wires, make sure the shields are grounded correctly. Ungrounded shield may introduce additional capacitively coupled noise."

Consider running experiments outside normal business hours when HVAC and other EMI-producing equipment may be off.

EDIT: In response to comments about DC power. Running electrophysiology rigs off 12V lead-acid batteries is an old and not uncommon practice. As a result some specialized equipment used for and around electrophysiology is designed to run off 12Vdc. Labs even build "quiet" sheds away from buildings and power lines. Rigs inside these sheds are powered with 12V battery banks, AC cords which are used for charging are retracted during experiments.

\$\endgroup\$
2
\$\begingroup\$

If mains noise is still a problem, run the circuits from a DC source like a battery.

\$\endgroup\$
  • 1
    \$\begingroup\$ Sure, this often helps a lot. Also adding a "50 Hz band pass" to before you feed the mains to your amplifiers or D/A or other devices is very helpful too (it sounds counter-intuitive but eliminating high frequency noise and spikes from the mains usually helps with the introduction of mains noise). \$\endgroup\$ – Ali Jan 17 '12 at 18:54
  • 2
    \$\begingroup\$ @spearson, most devices do use DC power, it is just the way of the digital world. I do not see how this puts a significant dent in what needs to be accomplished. \$\endgroup\$ – Kortuk Jan 17 '12 at 20:57
  • 2
    \$\begingroup\$ I think @spearson meant from batteries not a mains line transformed into DC. \$\endgroup\$ – Ali Jan 18 '12 at 1:11
  • 1
    \$\begingroup\$ @Ali: Did you mean "50 Hz band filter" (a.k.a. notch filter), maybe? \$\endgroup\$ – zebonaut Jan 18 '12 at 8:41
  • 2
    \$\begingroup\$ @Ali: Now I get it. Sorry, I was not careful enough while reading your comment. You're talking about the supply line; I thought you meant the measured signal. Yes, I agree, filtering harmonics from the supply might help a bit. BTW: Your profile says you're from the USA. You might need to filter 60 Hz instead of 50 Hz. \$\endgroup\$ – zebonaut Jan 18 '12 at 11:53
1
\$\begingroup\$

It's also quite important to try to get the connection of a surface electrode to be as good as possible-- and all the electrodes as identically fixed to the surface as possible. Two reasons.

  1. If the electrodes aren't nearly identically attached, there are likely to be fairly sizable junction potentential differences between electrodes, which can actually saturate high gain input stages if the inputs aren't high-passed. I don't particularly like to high-pass my inputs if I can avoid it, as it can mess up your input impedance if you're not careful. I like to get by small differential signals into a brick-wall impedance amp with high CMRR as soon as I can.

  2. Nicely attached 'trodes reduce movement artifact

  3. If the resistance in the electrode attachments differs too much, all that EM noise on the body through capacitive coupling to the world won't arrive at the amp as a common mode signal, but there will be a substantial noise component to the differential signal as well.

\$\endgroup\$

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.