The Raspberry Pi Compute Module hasn’t seen as much attention as it should have in our community, probably because the equivalents from the familiar consumer range can be so much cheaper. When a Raspberry Pi Zero is a similar size to a Compute Module and costs so much less, we can’t blame you for asking what would be the point of using the industrial version.
It’s interesting then to see an Instructables piece from [Manolis Agkopian] in which he takes the reader through the process of creating their own Compute Module project. Following hot on the heels of the recent launch of the latest in the range it’s come to us at an appropriate moment to take a fresh look at the fruity computer’s more obscure incarnation. He starts with a description of the Compute Module and its official development board, before taking us through setting up a module and putting an OS on it. Finally he shows us his board design, which he offers us as a jumping-off point for our own projects.
So given that it’s piqued your interest, why might you want to design a Compute Module project? The answer’s simple enough: the consumer boards only provide the subset of features the Pi foundation people deemed appropriate for their mission. A Compute Module project is the equivalent of designing a Raspberry Pi that does it your way, tailored exactly for your needs. If you want an example, look no further than this stereoscopic camera.
Via Hacker News.
Bought a Raspberry Pi. What now?
Put it in the drawer.
:o)
One is my home’s “WLANoT” server/gateway and feed scraper + MQTT broker.
One is our set-top box.
One runs an IR camera and motion detection — was looking for a hedgehog in the back yard, but he’s hibernating and we’ve since gotten a mouse in the basement.
One is permanently installed in an ancient synthesizer, spoofing a SCSI hard-drive over the network, but sounding a lot less like a helicopter in flight.
(Not all are actually RPis — some are OrangePi, one is a BB Black.)
I could use a couple more if you’ve got spares.
The original audio output of the Pi is not very good, so coupling it to a synthesizer is a good idea, especially when you want a good reproduction of a helicopter-in-flight sound. But you write, the combination makes less sound. :-)
Except he’s not using the Pi to output audio, he’s using it to spoof a SCSI drive to the synth. Its just allowing access to files that the synth can then use natively.
Yeah. This: https://hackaday.com/2017/05/01/the-raspberry-pi-becomes-a-scsi-device/
That helicopter-in-flight sound was previously made by scsi drive itself, without any synthesizer necessary :D
That’s the one! Trading it out for a Pi + software was well worth the cost of admission.
Downside: now I can hear the backlight inverter on the old SY99 that sits next to the sampler that used to contain the offending SCSI drive. Upside: the backlight on a 24-yr old synth still works…
idk, as a huge fan of black metal i find that the raspi audio outputs are very kvlt.
trve
also there is i2s
Totally answered a question I had for a way to update the HP 16601A. Any issues to consider you found?
Build a SatNOGS amateur satellite groundstation
This guy came out with a new carrier board that exposes both CSI camera ports for stereo vision, and is nice and compact! http://stereopi.com/
Yes, that looks awesome.
And I’m (we’re) actually working on another carrier board for the Pi which allows a Myriad X to in-between the stereo cameras and the Raspberry Pi Compute Module. It’s also stereo-camera but for a completely different purpose.
We haven’t published much yet (so bear with the hackaday.io page), but here’s the sort of stuff it would allow you to do (but at a higher framerate than that):
https://hackaday.io/project/163679-aipi/log/158979-example-object-detection-and-absolute-xyz-postion
The sweet part is the computer vision/ML stuff enables really-simple Python logic now on stuff that used to be SO hard. Like now you can do code that’s like ‘follow the closest person’. Or per a recent Hackaday (below), ‘follow the bottle’:
https://hackaday.com/2019/01/25/robot-cant-take-its-eyes-off-the-bottle/
So in short I think we’re going to see Raspberry Pi compute modules used like this, particularly with AI/ML accelerators which can then feed them results to work with.
wow that’s pretty advanced stuff. Movidius made the Google Tango chips IIRC
Yes, the Google AIY Kit is effectively the same idea, but with a Movidius 2 (the older version) and a Raspberry Pi Zero:
https://aiyprojects.withgoogle.com/vision/
And it’s monocular vision instead of stereo vision, as the Movidius 2 couldn’t do neural networks and stereo at the same time, where the Movidius X can.
Since Google is making their Edge TPU now, it doesn’t seem they’re going to make a Movidius X version:
https://cloud.google.com/edge-tpu/
So that’s where the AiPi fits in. We’re making it because we need it for a slew of things, and figured others would want/need it as well, given the gap that ended up here between the Myriad 2 and the Edge TPU.
Best,
Brandon
Interesting. How do these perform compared to a calibration grid over time (or each time testing) as an optical coordinate measuring system?
That picture is not a Compute Module. That is the IO board *for* the Compute Module.
There’s a compute module attached to the board.
Are any other interesting I/O boards for the compute module out there?
The main problem.with PI module is that you can’t build any commercial product because he don’t have any cerification needed to be sold on a production/industrial environment.
Is.good for in house products (but watch out insurance probably will no cover any lose due a Hw failure build/drive with a RasPi)
For commercial product is much easier build you own PCB or.use a industrial board
https://www.raspberrypi.org/documentation/hardware/raspberrypi/conformity.md
“The Raspberry Pi Compute Module hasn’t seen as much attention as it should have in our community, probably because the equivalents from the familiar consumer range can be so much cheaper. ”
Probably because very few of us can solder a SODIMM and paying for both the pi and the IO board when you can just pay for a Pi doesn’t really make any sense!
I really don’t understand why people keep trying to market computers with SODIMM or SD form factors to the hobbyist market. Did nobody notice that it took maybe a day for everyone to forget about Gumstix once the first Pi arrived?