Wake up delay difference if working w/ or w/o satellites

Hi everyone.
I’m planning on setting up Rhasspy and while I would like to have multiple satellites in different rooms, I’m concerned about the wake up delay because for me the responsiveness at that point is the most important thing…

If you would compare a single Raspberry Pi 4 with Rhasspy on it with a setup of a Raspberry Pi 4 as base and several satellites (e.g. Raspberry Pi Zero W): how much would that impact the hot word detection/wake up responsiveness in your experience?

(you can assume that all systems have solid Wi-Fi connections but feel free to enlighten me about the positive impact of 5GHz Wi-Fi or wired LAN connections.)

Thank you so much in advance!

When you run a Pi Zero W and have the wakeword set, the only factor is the speed of the device.
Disable the wake audio feedback for the quickest repsonse possible.

Ohh I don’t know why but somehow I thought, the satellite would only work as a remote mic and speaker. After checking the docs again I feel stupid to have assumed, the satellite would nonstop stream its audio input to the base for the base to run wake word detection. Thank you for clearing that up!
Do you think a raspberry pi zero w is a good choice for a satellite or would a “real” raspberry pi bring advantages in responsiveness?

Not if you install Rhasspy on it :slight_smile:
I think a zero is fast enough, but to be sure you would have to try it on your specific location. Hard to tell.

I am using a PiZeroW satellite with a Pi4 base and the delay is really low. The system as a whole can be slow if you chose Larynx TTS high quality for example, but that’s unrelated to satellite/base delay.

Thank you so much for your response! I’m excited to set it up soon. :slight_smile:

After reading many comments I decided that RasPi Zero + reSpeaker 2-mic HAT seems a more affordable combination in a tidy little package. It is.

I have currently got 2 satellites running on my desk. they are:

  • RasPi 3B I already had, with the reSpeaker 4-mic HAT.
  • RasPi Zero WH with Adafruit Voice Bonnet (equivalent of reSpeaker 2-mic HAT).

Both are running Raspberry Pi OS Lite and Rhasspy configured as a satellite with identical configuration (Audio Recording, Audio Playing and Porcupine WakeWord detection with the real work done on a RasPi4 base station), and both do the job well.

Of course the RasPi 3 is faster than the RasPi Zero - most notably in booting up and starting Rhasspy. My RasPi Zero just took 0:55 to boot to command line, and 5:48 to start Rhasspy to the console message advising that the web server is started. This delay got me looking at another add-on to use the LEDs to indicate when Rhasspy is ready.

Once running, Rhasspy on the RasPi Zero is only fractionally slower to recognise the wakeword and respond. I judge that it is still very acceptable.

If I build another satellite, I would however use a RasPi 3A since it has the same RasPi 3 processor and Wi-Fi, but saves cost by having RasPi Zero’s reduced I/O.

I would also suggest making a dedicated “RasPi Zero Rhasspy satellite” image with trimmed down OS and limiting the Rhasspy options.

Yes, boot to webinterface takes some time. Around 3 minutes on my setup. Maybe your sd card is slower.