Pi 5 and NVMe base, can we ditch SD cards!?

Hi all.

I am finally trying to move some things onto the Pi5 (home things, not :openflexure: things), and I just realised that PiMoroni sell an base that connects to the new PCIe port which allows NVMe hard drives. As it sits underneath this doesn’t stop you using a Sangaboard hat.

It seems people have got the pi to boot from the NVMe hard drives. These are so much faster and more reliable than SD cards. The reliance on SD cards was always the thing was most worried about with the Pi.

As the base costs about £13, and NVMe SSDs are similar price (£25 ish for a 256GB) drive… this seems ideal. They also have dual bases allowing separate OS and data drives.

Has anyone played with one of the bases?

1 Like

That absolutely should work, and seems to be a major reason for including the PCIe interface on the Pi5.

The current Openflexure operating system/server will not run on a Pi5 because it is tied to an old version of Rasbian. I think the coming v3.0 does not yet run on a Pi5, although that should be more a matter of getting round to it, because it is based on the new Pi OS.

Thanks. I shall report back on my NVMe progress for other projects.

1 Like

That’s good to hear! Buster is desperately outdated and won’t get any security fixes…

Yes. We were stuck on Buster because of breaking changes to the way the camera is implemented in the later operating systems, which meant that it requires a lot of re-writing in the server. That is currently in progress.

Yes - the new server is working, and in use on some of our medical projects - but due to some server migration pain I’ve not yet been able to properly share the disk image. Also, the code needs some serious tidying up before it’s really releasable (e.g. python packages rather than cloned repositories).

The new code will be based on Bookworm, i.e. up to date again. However, Bookworm did introduce breaking changes to the camera stack this January that caused headaches, so the most recent working version is actually patched into a Bookworm image from November IIRC. I have not yet had time to test if it has been fixed in a subsequent release, but it’s on the list!

I know Pi 5 changes a few things in the camera stack, so it will probably not work out of the box even with the new Bookworm-based softwrae, but hopefully the fixes will not be too tricky. The extra power is very tempting - but dealing with a fan in the base (my understanding was that a fan is basically required now) may require more than just a taller electronics drawer…

Either way, I love the idea of an NVMe drive replacing the SD card, this would eliminate a major headache :slight_smile:

2 Likes

We can just update the specs to say the microscope resolution is “vibration limited” rather than diffraction limited?

Would be worth experimenting with a heavier base and some rubber dampers between the pi and the base. But yes quite an engineering project to get everything perfect and confirmed.

yes - I’ve long wanted a nice robust box, where the electronics are fixed to the case and the microscope is “floating”, ideally with a big fat power bank under it for added mass. That would solve a few problems, including this one. Probably that’s something not-3D-printed for robustness and cleanability - a colleague here has a nice acrylic bender that would do, though there are simpler ways to make a decent box!

One day we’ll find the time…

We need to try the Pi5 and see how much of an issue heat is in this application. For running the actual microscope function it should be fine with no cooling or with some passive cooling, the computational need is not high. For things like analysis and stitching that could be more of an issue, although Pi4 is used for many things without even passive cooling and I don’t think Pi5 can be more power hungry per operation - it just lets you do a lot more operations in a reasonably time.

In the end, surely trying to do everything on the embedded machine is not sensible? Put a basic and headless Pi in the microscope to minimise heat and have no vibration. Then use a wired network connection to an adjacent beefier Pi5 or other computer for the heavy lifting.

Vibration control in a box would be useful anyway.

A beefy wide heat sink will do a lot of work with cooling. Could even try and pipe the heat to a big heatsink in the void space.

I’m not too sure about bent perspex. Perspex is actually quite an awful material really. Brittle and scratches.

I actually have been looking at peli cases. I think they are very standard, very strong, and actually sell frames designed to lock in an instrument. I think there should be a way to have the microscope mounted in a pelicase. May not be idea for all uses but would work well for a portable microscope.

I think working out what easy to specify boxes are available is a better way forward than trying to design a box.

1 Like