Sync offset at SMA 2ms

[ddvoraknyu]

Hi,
I noticed that there is an offset between TTL series from SMA connector and the data. The offset is about 2ms (60 samples at 30k). I used signal generator with both TTL and sinewave output. I fed TTL into SMA and sinewave into saline bath with probe submerged.

This is what signals look on scope:

This is what I get using SpikeGLX:

This is what I get using Open Ephys:

The offset can be corrected by retreating (moving backward) sync series by 2ms (55 samples at 30k).

I am using NI PXIe-1071 with Imec card, v1 probes, latest versions of software.

Could someone replicate this?

Dino

[nsteinme]

Thanks for making this kind of careful measurement, which actually I had not made myself before. I passed the issue on to some people who might be able to help test/diagnose - hopefully we'll hear back here from them.

[ddvoraknyu]

Bit more info for the record - exactly the same experiment but done with OpenEphys with Neuronexus 1x32 probe results in correct sync between TTL and data.

[jenniferColonell]

There is no time offset between the sync and neural channels. It is not a bug. It is an effect of the filtering of the AP band. High pass filtering of a low frequency since wave adds a phase shift.

By default, the Neuropixels 1.0 probes output data in two bands: the AP band, to which a 300 Hz high pass filter has been applied, and the LF band, which has been through a 0.5-1000 Hz bandpass filter. This strategy allows more efficient use of the dynamic range of the AP band by eliminating offsets and low frequency oscillations. For measurement of signals below ~600 Hz, the LF band should be analyzed, rather than the AP band.

The filter can be disabled. In SpikeGLX edit the imro table. In OpenEphys use the “AP Filter Cut” selection box.

Attached is a set of figures showing the phase shifts of low frequency sine signals and the alignment between neural channels and the sync channel.

NP1.0_filter_effect.pptx