Data Elements

These pages contain information about data types produced by Data Source Operators.

DataFrames

Data Source Operators produce a sequence of DataFrame or BufferedDataFrame elements. Both types carry the same fields; the difference is that a DataFrame holds a single device sample per field, while a BufferedDataFrame holds an array of samples. Each frame contains:

• Clock: a system-wide synchronized clock count that can be related to the Clock of any other data frame collected within the same ONIX Context.
• HubClock: a hardware timestamp generated by the local clock on the device's host hardware (for instance the oscillator on a headstage).
• The device data payload, e.g. a set of analog voltage samples like AnalogData within the AnalogInputDataFrame.

Using DataFrame type information

The data element pages also document the type and size of each property. This is useful in a few ways:

• Estimating data rates. Multiply the per-sample byte count of all properties by the device sample rate to get throughput in bytes per second. For example, a single NeuropixelsV2eData stream produces 384 channels of ushort (2 bytes each) plus two 8-byte clock fields at 30 kHz, giving roughly 23.5 MB/s.

  \[ \begin{equation} \frac{2*384+8+8\,bytes}{sample}*\frac{30,000\,samples}{s}*\frac{1\,MB}{10^6bytes} = 23.52\,MB/s \label{eq:1x_npx2_bw} \end{equation} \]

• Operator compatibility in Bonsai. Bonsai operators accept only specific types as inputs. If a connection between two nodes produces a type mismatch, Bonsai will indicate an error. Knowing a property's type helps diagnose these errors and choose the correct downstream operator.

• Loading raw binary files. If data was saved with MatrixWriter, the property type determines the correct NumPy dtype to use when reading the file back. For example, a Clock property saved as ulong requires dtype=np.uint64 when loading.

  Tip

  The DataFrameWriter includes all frame type information within the file it writes, so the user does not need to worry about knowing the exact type of each property. See the tutorial for details on its use.

Other data elements

Aside from DataFrames, OpenEphys.Onix1 operators produce:

• ContextTask: an object passed through the configuration chain for writing to and reading from the ONIX hardware.
• OutputClockParameters: the parameters used to configure the precise hardware output clock when the workflow starts.