Causes of Noise in Electrophysiological Recordings

 

A poor ground or reference connection will cause large amplitude noise, which will completely drown out the electrophysiological signal. The noise is wide-band, but is usually particularly strong at 50/60Hz and its harmonics

 

Trouble-shooting noise problems in electrophysiological recordings can seem like black magic; it often seems like an endless series of trial-and-error until you find a solution, and even then the solution may not have an obvious reason for working, or on another day may not work at all. However, there is cause-and-effect at work in any noise problems you encounter, and the illusion of randomness is caused by a multitude of possible causes of noise. Having an understanding of the possible causes of noise allows you to approach trouble-shooting problems in a more systematic and efficient way.

This series will discuss some of the more common causes of noise problems, as well as strategies to identify and correct them.

Using Spectral Analysis to Identify Noise Problems

Many sources of noise create signals of specific ranges of frequency, so identifying the spectral components of the noise signal can be very useful for identifying the source, and in some cases where the noise problem cannot be eliminated, allows you to create filters to remove the noise. The PlexControl software includes the ability to create a spectrograph of recorded data in real-time, allowing you to test different solutions without stopping the recording, making trouble-shooting far quicker and easier. See sections 12.3 and 12.4 of the OmniPlex User’s Guide for more information on creating two and three-dimensional spectrographs in real-time.

When trouble-shooting noise, it is a good idea to look at the raw, unfiltered data. Although many sources of noise produce primary frequencies well below your frequencies of interest (particularly when recording just spike data), their harmonics are still likely to cause problems. Identifying the primary frequency of the problematic noise makes it easier to identify the source.

Floating Ground

By far the most common cause of noise problems is related to a poor ground connection and should be the first thing checked whenever trouble shooting noise problems. A floating ground will result in large-amplitude wide-band noise across all channels, since all channels will be acting as an “antennae”, although particularly strong signals can present more significantly, such as 50/60Hz and their harmonics from nearby electrical lines and devices. Common causes of a floating ground can be a broken ground wire, the wire coming loose from the grounding site (or the screw coming loose if using a skull screw), using a ground site too far from the recording, a broken wire/pin on the electrode or headstage, or a malfunctioning headstage.

A floating reference will look similar to a floating ground since the noise picked up by the floating reference will be added to all channels. By contrast, any floating channels will be limited to the affected channel. Like a floating ground, floating references are likewise caused by a loose or broken connection in the electrical path. Floating channels are usually caused by physical damage to the electrode itself, or a damaged input or blown channel on the headstage.

Possible solutions:

If another headstage is available, try swapping it out to rule out a problem with the headstage. If the problem persists, check all connections. If you still have access to the ground screw or wire, or reference site, a continuity tester can be used to ensure that a stable and low impedance path exists for the reference and ground; most Plexon headstages have two exposed solder pads next to the connector for the headstage cable which are tied to ground and can be used for testing for continuity. These solder pads can be used to provide an alternative path to ground the subject if needed. If the reference is floating and cannot be easily fixed, the reference can be shorted to ground to remove the introduced noise. The reference can easily be tied to ground internal to the headstage by setting the headstage to “grounded referencing” in the Plexon Digital Headstage Processor Device Settings menu; see Appendix D of the OmniPlex User’s Guide for more information. If you are forced to ground out the reference, but would still need to use referencing to address other noise issues, the PlexControl software offers options for using a recording channel as a digital reference, or the averaged signal or median signal from all channels or a subset of channels as a reference. See section 14.2 of the OmniPlex User’s Guide for more information on setting up digital referencing.

Intermittent noise caused by a wireless device communicating through WiFi. In this case the noise created artifacts similar in amplitude to the recorded spikes.

RF and EMI Interference

Strong sources of electromagnetic noise can also cause more than enough noise to ruin your recording. Overhead lighting can cause strong 50/60Hz noise (the ballasts from fluorescent lighting are particularly likely to cause problems), as can the fields created by transformers and the electrical lines themselves. The power supplies of most electronics (lab computers are a common culprit) usually generate stronger noise at twice the power frequency (ie 100/120Hz) since the power signal is rectified when converted to DC. Electric motors create particularly far-reaching electric fields (coolant pumps in refrigerators and freezers are a common source of noise in wet labs). Cell phones and other communication devices commonly cause intermittent, high frequency noise signals when pinging cell towers or other devices. Also be aware of other devices in the building that can have far ranging effects, such as MRI scanners, microwave communication equipment, radio transmitters, etc.

Here the recording area was unwittingly placed too close to power conduit in the wall. Here the spike data is minimally affected thanks to the high-pass filter, but local field potential recordings will be affected.

Possible solutions:

Try turning out the lights to see if that reduces 50/60Hz noise. Move the recording area away from any active electronics, as well as transformers, electrical outlets and walls (where power lines may be located). Keep reference and grounding cables as short as is possible, and if using analog headstages also try to use headstage cables that are as short as possible, and avoid looping excess cable (looping creates inductance and increases the antennae effect). Upgrading to digital headstages will remove any noise picked up by the headstage cable, since the signal is immune to noise once digitized. Turn off cell phones, as well as all other unnecessary electronics while recording. In cases where the recording area can not be moved away from sources of noise, a Faraday cage may be necessary. Remember that Faraday cages should be properly grounded to be most effective.

Although it is better to remove the source of line noise when possible, the PlexControl software offers an adaptive Line Noise Filter to remove 50/60Hz noise and its harmonics with as little affect to the electrophysiological signal as possible. See section 12.5 of the OmniPlex User’s Guide for information on how to enable the Line Noise Filter in the Filter Control Panel.

Here the recording area was unwittingly placed too close to power conduit in the wall. Here the spike data is minimally affected thanks to the high-pass filter, but local field potential recordings will be affected.
An example of movement artifacts created by a subject pulling against the headstage cable strong enough to cause movement in the Omnetics connector

Movement Artifacts

Recording from awake and freely behaving animals can introduce its own challenges in the form of movement artifacts. These artifacts have three main causes; artifacts from a swinging cable, movement in the connectors between the electrode and headstage, and EMG or myographic artifacts created by contracting muscles in the subject. Swinging cable artifacts will only be seen when using analog headstages, since the signal is already digitized when it reaches the cable of a digital headstage. Unfortunately artifacts from both movement in the electrode/headstage connectors and myographic artifacts occur before digitization even with a digital headstage, and so can still be a problem. Myographic artifacts with rodents is especially common in rodents, both because of their tendency store food in cheek pouches and randomly start chewing it at anytime, and because of the size of their powerful jaw muscles in comparison to the rest of their heads. Food or liquid rewards during a recording will also cause myogenic artifacts, which can also be compounded by artifacts caused by coming in contact any conductive parts from a feeder, liquid delivery device or water bottle.

Possible solutions:

A commutator is strongly recommended for behaving subjects. If using a passive commutator, and you are seeing artifacts due to movement between the headstage and connector, consider switching to an active commutator, which will reduce drag on the animal’s head and result in less tugging and pulling from the animal.

If using analog headstages and artifacts due to a swinging cable are a problem, try to use headstage cables that are as short as possible. The cable should be just long enough to comfortably reach the subject when it is at the furthest possible location from the commutator. If there is too much slack in the cable, you can try looping it at the commutator to take up slack. As mentioned in the section on RF and EMI interference, looping will increase the antennae effect, but it is generally preferrable to excess swinging. If the swinging cable problem persists, consider upgrading to digital headstages.

Myographic artifacts can be reduced by a good reference, but can also be made worse if the reference wire is moving during jaw movement; you can experiment with turning referencing on and off to see if it is improving the signal or making it worse. If the experiment permits, try to avoid feeding the subject for several hours before recording. If using water bottles or feeders with metal parts coming into contact with the subject, try using a grounding wire to ground those parts.

Even with the headstage heavily shielded, and all channels tied to ground, thermal noise remains. Note the fluctuating “red noise” distribution. Barring severe overheating, the spectagraph needs to be zoomed into lower frequencies and very low amplitude in order to visualize the thermal noise.

Physiological Noise

A subject’s beating heart and respiratory muscles can also create problematic signals. Although unlikely to affect single unit recording, they can be problematic for local field potentials and EEG. Additionally, even in acute, anesthetized preps, myogenic artifacts can be caused by muscle twitches from an under-sedated subject.

Possible solutions:

A properly placed reference should remove the majority of cardiac and respiratory noise. If it does persists, the cardiac and respiratory signals can be recorded independently (a differential headstage would be ideal for this), so that with the appropriate filtering and scaling, the signal can be subtracted in post-processing. Finally, a notch filter can be used at the animal’s cardiac or respiratory rate, although this is not ideal for several reasons: it results in a loss of LFP or EEG data arMovementound those frequencies, variations in heart beat and respiration can cause the signal to intermittently leak through the filter, and the notch filter itself can cause ringing or phase distortions that can be detrimental to some types of analysis. Information on setting up notch filters can be found in the section titled “Global Filters” in appendix B of the OmniPlex User’s Guide.

If any visible twitching can be seen from the subject, or if large, intermittent artifacts can be seen in the scrolling data, consider delivering additional anesthesia to the subject, although be careful not to cause an overdose.

Thermal Noise

In the unlikely event that you are able to eliminate all other sources of noise, your signal will still have thermal noise. This is caused by Brownian motion in conductors or semiconductors in the entire signal path before the signal is digitized (including the electrode itself), and since Brownian motion is independent in every location, it cannot be removed or reduced by a reference. Luckily though thermal noise is unlikely to drastically interfere with the electrophysiological signal as long as none of your devices are overheating.

Possible solutions:

Make sure that electronics are properly ventilated and kept in a cool environment.

Plexon’s Sales & Support Team is Available to Help

If you encounter a noise problem that you cannot figure out, do not fret, Plexon’s sales and support team are available to help! All Plexon systems come with a lifetime of support; we can be reached at info@plexon.com, or through the online support request feature on our website. It is helpful to have a recorded example of the noise problem ready to share; for continuous noise a minute or so of recording should be sufficient, and for intermittent noise a recording with several examples of the problem is preferable. Including a picture of your recording environment is also often helpful.

 

Written by Nafi Yasar