SimenZhor

The main cost driver is the sheer number of individual components. In total, the 252 electromagnets add up to more than $1000, and the components on the PCBs add up to roughly $400 plus the PCBs themselves at around $100. Then there’s the ferrofluid itself and the power supply, which both are around a hundred bucks each, as well as connectors, cables, a microcontroller, nuts, bolts, plywood, etc. that adds the rest. Since we’ve made multiple revisions of each part the total we’ve spent is closer to $3000 for the development of this entire project.

That's really cool to hear, Andy! Please keep us posted, and let us know if you find a source for PCB mounted magnets. We would love something like that and have been talking about it for a long time, to be honest.The new PCBs are tested and pretty much ready to go. We just want to do a full integration test and finish the video about that before releasing the new design so that we know it works with 12 PCBs at the same time (so far we've only tested one at a time). Hopefully, we can complete the full integration test before November, and release the video shortly after that. CheersSimen

The material costs are around $2000 for a single unit of this display.

I'll let you know why Assembly is often overlooked. It's often just not necessary to go to that length. For most programmers, Assembly code will barely be readable (if at all), and avoiding it whenever possible seems a smart choice to me. For our specific application, I'm sure we could easily see a massive performance boost by further optimizing the code (and we do state this in the video about performance) - but for us it was a question of "where can we get the best performance boost for the least amount of coding time". We decreased the runtime of the main loop by a factor of 450, and that was more than enough to do what we wanted.Additionally, we have always planned to revise our control hardware, and most of the optimization techniques we could see would have to be redone wh…

I'll let you know why Assembly is often overlooked. It's often just not necessary to go to that length. For most programmers, Assembly code will barely be readable (if at all), and avoiding it whenever possible seems a smart choice to me. For our specific application, I'm sure we could easily see a massive performance boost by further optimizing the code (and we do state this in the video about performance) - but for us it was a question of "where can we get the best performance boost for the least amount of coding time". We decreased the runtime of the main loop by a factor of 450, and that was more than enough to do what we wanted.Additionally, we have always planned to revise our control hardware, and most of the optimization techniques we could see would have to be redone when changing out the hardware. Actually, we're planning on completely replacing the software PWM with hardware PWM (and will hopefully test this in the coming weeks), meaning further optimization of the shift out sequence would be a massive waste of time.Finally, please keep in mind that we are relatively inexperienced with embedded systems, we are all still studying, and this is purely a hobby project done on the side. It's just not meant to be "perfect". We are more than happy to get it working and to a state where we're proud to show it off.

We have a theory that it will be possible for the player to remember which blob they're controlling because it will respond to their commands. It's obviously not the best UX, and it will have a learning curve, but with the technical limitations of this display it's the best we can do.

I think the voltage range is just written that way to indicate that they will make magnets according to customer specs, but the most common voltages for these magnets are 12V and 24V. We have never bothered to actually measure the average current draw, because we are more concerned about the peak current draw and didn't have equipment suited for measuring that. Therefore we instead went for a solid overspecification of the power supply (1200W to the theoretical ~400W needed). This "trade-off" wasn't really a trade-off because it literally had zero downsides compared to for example a 500W ATX PSU for a "regular" desktop computer. The formfactor of the 1200W PSU was better for our purpose, the price was better, the power rating was better, and it was more "hackable&…

I think the voltage range is just written that way to indicate that they will make magnets according to customer specs, but the most common voltages for these magnets are 12V and 24V. We have never bothered to actually measure the average current draw, because we are more concerned about the peak current draw and didn't have equipment suited for measuring that. Therefore we instead went for a solid overspecification of the power supply (1200W to the theoretical ~400W needed). This "trade-off" wasn't really a trade-off because it literally had zero downsides compared to for example a 500W ATX PSU for a "regular" desktop computer. The formfactor of the 1200W PSU was better for our purpose, the price was better, the power rating was better, and it was more "hackable". With all this said, if you already have a 500W PSU lying about and want to use it for powering magnets, it will probably work for this specific use case. Keep in mind that there will never be a real use-case where all magnets are being used at once. The only thing that could make that happen would be a programming or user error (so, I'm not saying it won't ever happen).

We bought 2 x 250ml bottles of EF-H1 by Ferrotec from this supplier: https://www.czferro.com/products/ferrofluid-bulk but that has proven to be WAY too much :) I think we have used roughly half of one 250ml bottle (maybe a bit more) but that includes a lot of experimentation and goofing around. 60ml would probably have been more than enough for the display alone.I completely forgot to mention that we use EF-H1, although it is mentioned in the Instructable we link to about handling the ferrofluid itself. I've added a link to the supplier we used in the "supply" section now, thanks for letting me know!

Great project! Love the aestethic

That depends on how much you have from before, of course, in addition to the prices in your country. Personally I didn't have any of the painting stuff, which summed up to around $60 when bought from Norwegian retailers (I made two pieces, and still have leftovers that can be used for other projects). The paint can easily be skipped though, and with just a plain 3D-printed surface the price shouldn't be more than around $10.