Non-python Seabreeze question - do open spectrometers vanish after a while?

[jetmonk]

Apologies in advance from digressing from python-seabreeze, but it's hard to find any community using SeaBreeze.

I'm writing an interface to SeaBreeze 3.0.11 for Common Lisp, currently for a USB4000.

Everything works fine for a while, except that open devices seem to vanish after some timeout, or system sleep.

Specifically, the USB4000 still shows up on the OSX/Darwin usb bus. The function sbapi_get_device_ids() shows that there is a device at device_id=2. But any attempt to access or reopen the device, like sbapi_open_device(2), gives Error 2 (No device found).

The only way to get the device back is to unplug it from the usb, and plug it back in.

Has anyone encountered this, and, if so, has python-seabreeze found a workaround?

[ap--]

Hi @jetmonk,

Can you reproduce this behavior with python-seabreeze?

I remember during development that unplugging and reconnecting the spectrometer was needed every now and then when an incorrect command was sent to the spectrometer or a read buffer wasn't fully emptied. The spectrometer basically ended up in an undefined state. But this happened more often when developing the pyusb based backend.

Can you reproduce the connection loss reliably? And if so: what is the minimal set of API calls that reproduce your error?

Cheers,
Andreas

[jetmonk]

Hi @ap--

Thank you for replying so quickly. I now have a theory what happens. My USB4000 is connected to a monitor (USB hub) that can go to sleep. I think that a monitor sleep can put the spectrometer into a "undefined state" - I don't see why a sleep/wake isn't the same as unplugging/plugging, though.

Today, I changed the sleep settings of my monitors so that monitor sleep and system sleep didn't happen at the same moment, and my spectrometer hasn't gone off-line since.

So I consider this to be tentatively solved.

[ap--]

I don't see why a sleep/wake isn't the same as unplugging/plugging, though.

I remember looking into this a few years ago, and I tried various tools on linux that claimed to reset usb devices without unplugging. (Mainly to have a reliably running test suite)
But in certain cases I was just never able to reconnect to the spectrometer and had to unplug.

At some point when the pyusb backend became more reliable this became a non-issue. So I never investigated further. My explanation was just, that the firmware running on the spectrometer isn't able to recover from certain fail states, and that a software USB reset just doesn't reboot the spectrometer in those cases. But that's me guessing.