A signal averaging package from Spectrum Instrumentation offers new capabilities for acquiring and averaging signals, particularly low-level signals or signals hidden in random noise. Spectrum’s CUDA Access for Parallel Processing (SCAPP), used with the company’s digitizers, harnesses the power of any CUDA-based GPU card. SCAPP enables porting data directly to the GPU, using remote direct memory access (RDMA), where high speed time and frequency signal averaging is performed without length limitations. Since the data is ported directly to the GPU card without intervention by the host processor, averaging can be performed on signals of almost any length.
SCAPP works with Spectrum’s fast M4i and mid-range M2p series PCIe digitizers, so users can select the performance best matching the signal acquisition requirements. The M4i series samples up to 5 GSPS with 8-bit resolution, 500 MSPS with 14-bit resolution or 250 MSPS with 16-bit resolution. The M2p cards sample from 20 to 125 MSPS with 16-bit resolution, with up to eight channels per card. The digitizer cards include flexible trigger, acquisition and readout modes, enabling them to average signals when trigger rates are extremely high.
The averaging package is part of SCAPP’s driver package and includes the RDMA extension for direct data transfer from the digitizer to the GPU, either directly for PCs running LINUX or via the CPU for systems running Windows. The package contains examples for interacting with the digitizer and a set of CUDA parallel processing routines for averaging. All software is written in C/C++, which can easily be modified for users wishing to develop unique algorithms.
Spectrum Instrumentation GmbH