Hi,
we are currently trying to integrate Artiq into our experiment and have the following situation/question:

What we need (in parallel):

  • Continuous readout of the sampler in a loop (always running)
  • Submit experiments which do some scan of parameters (DDS frequency).
  • Perform different simple tasks like e.g. send a TTL signal via our TTL IO or change a DC voltage via our DAC.

Our approach (not working)

  • For all communications with our modules (Sampler, TTL, Zotino,...) we currently submit experiments following the online examples (like the LED switch example). They all use the 'core' device, like in the examples.

The problem

  • Submitted experiments can not be run in parallel, because they interfere on the core.

Our question

  • How do we realize simple 'single request' tasks in parallel to a submitted experiment running on the core?
  • Can we run not-time-critical experiments without involving the core?

Further information

  • We are using Artiq version 6.7
  • For the single calls we don't have high requirements on timing.
  • We don't use the Artiq Dashboard, but use our own custom software to submit experiments to artiq. That works so far.
  • On the Artiq Dashboard we see that while a submitted experiment is running, we can actually send single calls to Artiq modules using the TTL applet. If I understand correctly there is a function involved which is called 'inject()'. Is this what we need to use? If so, how can we use it?
6 days later

Concurrent experiments are currently not supported.

Some devices/channels (e.g. TTLs, old non-Urukul DDS, Zotino) support monitoring and/or injection. This allows the dashboard to interact (read with monitoring and set with injection) directly with those channels, bypassing experiments running on the core device.

Implementing monitoring for sampler would be possible. The downside would be regular and not arbitrary conversions.
Implementing injection for Zotino or Urukul would be more involved.
For Fastino and Phaser monitoring and injection would be doable again.

The monitoring/injection api can also be accessed from the host (like the dashboard does). It's unclear how well that scales to many clients.

A Stabilizer board would be a good place to run not-time-critical experiments without involving the core. If it doesn't have enough analog channels, the EEM connector on it could be used to read a Sampler (though as far as I know, nobody has tested this or written the required software yet).