Noise: Reference and Ground

Overview

Imec probes have two solder connections: reference and ground. This note covers what they are and how to use them.

The Ground Line

Always Connect To CSF.

This is an extension of the basestation's wall ground. However, because the ground runs through a 5-meter, thin wire, the probe ground is about +650 mV higher than wall ground. This sets the ground level for the probe's base circuits that sample, amplify, filter, digitize and transmit neural signals.

Generally the subject should be grounded, whether an animal or buffer, for these reasons:

  • Damage: If the subject is floating and holding a static charge, the probe could get zapped and damaged.

  • Saturation: If subject potential is far from that of the electronics, the amplifiers can saturate. The reading are then unusable. Moreover, the amplifiers can sometimes lock into saturation until restarted, which wastes time.

  • Noise reduction: Connecting the subject to ground reduces its ability to act as an antenna. Although the differential nature of the measurements acts to cancel common noise, it's better to drain the noise to ground before it ever gets to the amplifiers.

In summary, always ground the animal: run a wire from the ground terminal of the probe flex to the animal CSF, either via a pin through the skull into CSF, or placed in the saline in the craniotomy.

IMPORTANT: The probe ground should be the ONLY ground connection to the animal. Otherwise you'll make ground loops which are noise antennas.

For self-referenced noise measurements in a bath, we also recommend connecting probe ground to the bath to act as a noise drain. This is a better measurement of the probe's own internal electronics noise. It's also more similar to an animal ground configuration. It is highly recommended to do self-referenced noise measurements one-probe-at-a-time to prevent any possibility of multiple probes interacting with each other.

The Reference Line

Flexibility: Depends On Needs.

All imec probes are designed for differential measurements. The electrode sites connect to (+) amplifier inputs, and the selected reference level connects to the (-) inputs. The benefit of differential measurement is common mode noise rejection. This only works if the reference is connected to the animal.

The system supports a variety of reference configurations to meet your needs:

  • External: Select External in software (in the IMRO editor), and connect the probe's reference terminal to any desired point. That could be the bath in a noise measurement setup, the animal CSF, or a second wire in the brain, far from the imec probe. BTW, external is the default option.

  • Ground: Some probes (mainly 2.0) offer a software option for Ground referencing. This internally connects the (-) amplifier inputs to circuit ground. Using this mode, you can make measurements relative to ground without soldering a wire to the probe's reference terminal. If your probe doesn't support the software Ground option, you can select External and solder a wire from the reference terminal to the animal as above. BTW, because the ground line is connected to the animal, this mode is still a differential measurement retaining a useful degree of common mode rejection.

  • Tip: This internally connects the given shank tip electrode to the (-) inputs. As with the Ground option, this does not require any wire to be connected to the reference terminal on the flex.

Pro tip: A tiny internal switch is used to connect the (-) amplifier input to the selected reference. Only the External mode requires a wire from the reference terminal on the flex to some point in the brain. In any other mode, having a reference wire connected is OK, but it isn't doing anything valuable. The ideal practice is to omit the reference wire entirely to minimize crosstalk from that terminal, across the switch. The worst thing is to connect a long free wire to the reference terminal, making a noise antenna.

Standard Practice For Spikes

The most important thing for referencing in spike measurement is low noise. There isn't a rule about what reference to use or where to put wires.

For AP band it probably doesn't make a lot of difference where you reference. Different choices will differ from each other primarily in DC offset, and hopefully lowish frequencies, and that will be removed by the band-pass and CAR filters you'll use in postprocessing.

So what does everyone else do? It's helpful to allow a main and a backup option, and have the convenience of a single extra wire coming off the probe. The wiring part of the strategy depends on probe model:

  • 2.0 with Ground Option: Connect one wire from the flex ground terminal to the CSF.

  • {1.0, NHP, Ultra} w/o Ground Option: Solder the reference and ground terminals together at the flex and run a single wire from there to the CSF.

This way you can start with ground referencing, and if the noise is small, you're all set. If it's too noisy, you may have a contaminated ground, so just switch to Tip referencing in software without changing any wires. That may look better. So with this setup you get two chances to get data today, and you can try to get rid of the noise in your lab tomorrow.

LFP Measurements

Use External or Ground Modes.

LFP tends to vary slowly over large areas. Tip referencing may be an especially bad choice: for sites near the tip the differential measurement will cancel signal as well as background, and at the other end of the probe measurements contain (local LFP - tip LFP), so the referencing is inconsistent. Moreover, you'd rather have a larger area electrode for better averaging.

Noise

To minimize noise use low impedance paths for reference and ground:

  • Do a good job with soldering any and all the wires.
  • Use beefy wires and ground straps where you can.
  • Don't use hacks like alligator clips to connect things.

fin