Roach Lab Spectrometer
The old lab spectrometer is dead. Or at least it's pining for the Fjjords. We need something for the receiver guys to use.
This spectrometer is meant for testing receivers. Since that's the case, we'll start off with a minimalist machine.
- Mode 1 -- 2 inputs
- 4096 channels, 1 second minimum integration
- 1 second minimum integration time
- Lots of FIR filter taps if possible
- 32 bit integer output spectra
So the data rate for this would be: (4096 channels * 2 pols * 4 bytes/channel) / 1 second = 32K bytes/second. Heck, we can dump faster than 1 second if we need to...
Other modes can be added as the mood strikes.
We will start out with one ADC connected to one ROACH. This ROACH will do one channel of the spectrometer. We'll duplicate this for the second channel. There is no need for full-stokes, so we don't need to sync up the 2 boards.
Use a Valon fixed-frequency synthesizer to clock the ADC. This should be locked to the 10 MHz in the spectrometer rack.
Start with Srikanth's gateware for the 20 meter. It should build fine and run with a clock of 800 MHz once one channel is split out and the stokes stuff is removed. Hmm. Maybe just remove the stokes and the accumulators for the stokes and leave the 2nd ADC in there. It might just build fine, then. If so, just keep it all on one ROACH.
Use KATCP for control and reading out the data, starting with Srikanth's existing stuff. We'll need to query Roger Norrod about the format for the output files and any metadata we want to collect. I suppose that they'd want something similar to the existing cleo manager screen. We will make the software run on castor, connected to the ROACHs by 100 Mb Ethernet.