Viewing the frequency spectrum of a wideband or LFP signal is important to detect noise issues, whether line noise, monitor refresh noise, reward artifacts or other anomalies. Removing artifacts is critical not only to obtain clean LFPs and wideband signals, but also to sort spikes in low SNR scenarios, for example with multi-electrode arrays.

We use a modified version of the continuous LFP client that comes with the Plexon Matlab SDK to weed out the largest of these artifacts. Nevertheless, it’s important to periodically do a test recording for a short task and examine it offline since some artifacts are intermittent and only occur during actual tasks (reward artifacts for example).

I use Baudline to examine wideband signals offline in Linux. Baudline can read any signal consisting of a stream of bytes, for example the Plexon .ddt file format. You can convert .plx to .ddt using PlexUtil on Windows, or write your own code on other platforms; I use my own Python code that uses the .plx reading facilities of OpenElectrophy to do this.

The default settings of Baudline will read a .ddt file as is; you only need to change the sampling rate so that it agrees with that of your system. The interface is highly responsive, which is not the case if you try to visualize long wideband recordings in Matlab. The default view shows the windows power spectrum as a function of time. Here’s an example of a bad recording:

As you can see, there’s both vertical streaks corresponding to harmonics of line noise (the 180Hz line is especially strong) as well as horizontal streaks which are probably reward artifacts. These are subtle enough that you might not notice them with the online monitor, but they make spike sorting that much more difficult. Here’s a clean(er) recording for comparison: