The HIL system performs the following operational steps:
A system MATLAB model running on the host computer generates a set of synthetic snapshots from a simulation model where \(S\) targets are configured to move at constant velocities while emitting an EM signal resulting in incident signals towards the ULA with given angles with respect to the boresight direction.
The host computer runs the MUSIC model on the generated snapshots and generates reference results for the detected DOAs. These are compared to values produced by the AI Engine implementation. Steps 1 and 2 can be performed once again after Step 5.
The VC1902 PS application initializes the hardware data path including memory buffers, PL data movers, and AI Engine configuration.
The VC1902 PS application starts a TCP server to accept incoming TCP connection requests on a port specified via a command line parameter.
The host computer starts a TCP client on the port number on which the TCP server accepts incoming TCP/IP packets.
The TCP client on the host computer sends the configured number of snapshots in a single batch carried in TCP/IP packets. Each snapshot in the batch is encapsulated in a frame that includes a header that provides start of frame, a sequence number, snapshot type as well as several other system and MUSIC configuration parameters as shown in the Appendix.
PS client data extracts the payload from incoming TCP/IP packets and collects data in the memory buffers in DDR4 before initiating the hardware pipeline.
Upon the collection of the last (or a unique) snapshot from TCP client, the fabric
mm2s()
DMA starts transferring the snapshots to the AI engine as a single DMA operation.The AI Engine receives data from two 64-bit PLIO inputs running @ 625 MHz, performs MUSIC processing steps through one or more AI Engine graph iterations, depending on the number of snapshots.
The AI Engine outputs the spectral bins and DOA vector transferred by the
s2mm()
PL DMA into the memory buffer located in DDR4 of the VCK190 board.The PS application waits for all hardware pipeline stages to complete their operations and transfers the expected amounts of data to/from memory.
The TCP server sends the output produced by the AI Engine back to the TCP client. Each snapshot output including its spectrum bins and DOA vector is encapsulated into a frame with a proper header.
The host computer receives the spectrum and DOA vector computed by the AI Engine and plots the results of both the expected and reference data.
The TCP/IP server remains in listening mode for a fixed amount of time, while the client is performing data visualization.
Go to Step 12 if no new data is received from the TCP/IP client during the wait time window. Go to Step 6 if the TCP/IP client sends a new input batch.