Building 45 Part 2: Props and Programs
Quick and Dirty Mockup
I wanted an idea of how realistic my concept art was
As an engineer I hate to admit it, but this project is mostly form over function. I want the tile to look nice for the demos as this project is meant to be a proof of concept. I made this mockup to see how close my design schematic matched my concept art.
It’s a 1:1 mockup but I oversized the edge vents and face plate a little bit. Overall I’m happy with how it looks and I’m sure the final design will look even better!
Tile Emotion Matrix Design
This project’s main appeal is its personality and there’ll be plenty of it!
The eye will be the tile’s main form of expression so it is vital that I get this right. I’m using the MAX7219 IC to control this custom 8*8 LED matrix I made. The LEDs will be covered by a light diffusor from an LCD so it looks more natural.
I’m going to get this board fabricated and I’ll need to solder the 64 SMD LEDs myself. I hope to get a stencil and use a hot air gun to make it easy.
I designed the board with the programming in mind. The 8 rows are curved but connected similarly to the square version. This will hopefully make it easier to program the expressions in later on.
Aside on my Noise Research
The tile’s noise output can hurt its usability so that is where I focused this weeks research
I’ve been thinking about the noise issue a lot and after this demo I might try a few new designs. A balloon biased design seems to be the most promising option as it would be quieter and less disruptive. Gasses are difficult to deal with though. Both Hydrogen and Helium permeate through thin materials so it would need to be refilled every 1-2 days. Helium needs to be bought and is very difficult to store. I would need to buy a pressure vessel and swap it out every month. Not to mention the required fittings and the dangers of high pressure gasses.
Thankfully I have experience working with hydrogen from another project. I’ve built multiple electrolysis reactors (sadly undocumented other then my sketches) that spit water into hydrogen and oxygen. Back then I was only floating vegetable bags but this could be a better option if the drone design doesn’t work out.
I would need to refill the balloon every 1-2 days with a full refresh every 1-2 weeks (Due to oxygen intrusion) but this design might be more pleasant to be around (If I want to start bringing this around “pleasantness” is vital). I would also need to reduce the weight of the design (Base station is already required). I would love to learn more about PCB design so combining all my separate sensor boards onto one (flexible?) PCB would be a great challenge.
Anyway, I planned the drone design so that is what I’m moving forward with for the demo in April.
Propeller Noise Experiment
I measured the sound output of three props to test the viability of the drone design
I go over the test setup in the video below. After the test I found out that the batteries were running low on the scale which could’ve been causing the instability of the thrust result. This was the same for all the tests though, so the results give the correct overall picture. More test information is in the results sheet shown below.
Experiment Results
The 4″ prop I was originally planning on using turned out to be the best
The running theme with this project seems to be that I doubt part of my original design, I go through extensive research and testing… just to find out that the original plan was correct. The 4″ 6 blade propeller performed the best in my testing and sadly my custom single blade prop didn’t work. I was hoping that the Toroidal propellers would be quieter but the numbers don’t lie. I got the 5″ 6 blade props just after I finished the testing so I’ll test those next week. I also ordered some 3.5″ (largest size available) 8 bade propellers to see if they are any quieter.
The trend I’m seeing is that larger props and more blades make a prop quieter. More tests to come!
Speech to Speech Working!*
Using the Text Gen Interface to talk to the AI
On a more positive note, I can talk to the AI now! I should’ve gotten this working for the initial demo because it was very easy to pull off! This simplicity comes at a cost though. The responses are too slow to be useful in a conversation. This is because they use the original version of whisper which is very slow.
The transcription of my speech takes over half the compute time which I’ll solve by using “Faster Whisper” (The model I need a newer GPU to run). I’ll also use the built in streaming functions to avoid needing to complete each step sequentially. Implementing these methods will allow for near instantaneous conversation.
Next Steps:
My plans for next week’s post
I’ve bought all the materials I need to build the Tile but the server needs some more work. I have the speech recognition working but I will need the newer GPU to make it useful. The 4080 I want to get cannot fit in my server nor would I really want to put it in there. My server’s two CPUs can be beat by any semi-modern low end CPU. I will start phasing out my R720 to get something more capable of handling Machine learning. I’ll do more research and get a parts list together.
I’m still waiting on some parts to ship but after I do the server research, I plan to start the wiring. The goal for spring break was to have the Tile built but I’ll be lucky to get to the CAD next week. I need to get the wiring done and make sure the drone is behaving correctly before I put it in its shell.
TLDR:
I’m going to research and order all the required server parts before finishing the wiring next week. I’d also like to have a drone control program in place if possible. CAD model coming the following week.
This is all great, Ranger. I agree that your noise concerns are valid. One thing to consider is that the tile doesn’t necessarily always need to hover. If the plan was for it to only use blades for relocation from fixed location to fixed location, then the noise isn’t as much of a concern. Since blades for movement are needed in either case, you may just want to go with the gas option first and decide if it gets eliminated later. ,,,or inside applications and outside applications have different rules. Balloons outdoors (in wind) are more of a concern. Blades indoors (with noise) are more of a concern. Glad to read that the Speech-to-Speech is working!