- Real time neural network inference directly on FPGA hardware for ultra low latency processing.
- Supports custom neural network models trained offline (e.g., with Python) for direct deployment to Moku devices.
- Up to 5 fully connected layers with up to 100 neurons per layer and multiple activation function options.
- Multi channel signal handling in both parallel and serial processing modes.
- Seamless integration with other Moku software defined instruments in a single device.
- Ideal for closed loop control, real time classification, noise reduction, and smart signal conditioning.
- Designed for accessibility enables ML in test and measurement without requiring deep expertise in AI.
Moku Neural Network
Moku Neural Network brings machine learning directly into the physical test and measurement environment. By leveraging FPGA-based hardware, it enables real time inference of neural networks in line with other measurement instruments making ML driven signal processing, control, and analysis part of your lab’s standard workflow. With Moku Neural Network, you can design and train neural networks offline (e.g. using Python), then upload the trained model to a compatible Moku device (such as Moku:Pro or Moku:Delta) to perform low latency, real time processing of incoming signals. This lets you automatically classify signals, filter noise, condition sensor inputs, or implement closed loop feedback all with minimal latency and high reliability.
Features
Key capabilities and advantages
Specification
Technical specifications and parameters
| Moku Delta | AI 8 channels, up to 2 GHz, AO 8 channels, up to 2 GHz, Digital I/O 32 channels, Sampling rate up to 5 GSa/s, FPGA-enabled, supports Neural Network, Multi-Instrument Mode up to 8 instruments |
| Moku Pro | AI 4 channels, up to 600 MHz, AO 4 channels, up to 500 MHz, Sampling rate 5 GSa/s (1ch) / 1.25 GSa/s (4ch), FPGA-enabled, supports Neural Network, Multi-Instrument Mode up to 4 instruments |
| Moku Lab | AI 2 channels, up to 200 MHz, AO 2 channels, up to 300 MHz, Sampling rate 500 MSa/s (inputs), 1 GSa/s (outputs), Supports moderate Multi-Instrument Mode, Suitable for prototyping / testing |
| Moku Go | AI 2 channels, up to 30 MHz, AO 2 channels, up to 20 MHz, Sampling rate 125 MSa/s, 16-channel digital I/O, Portable, entry-level, limited Neural Network support |
Chat Us