MagicShifter 3000: An Over-Engineered POV Stick With A 15-Year Journey

3 hackers, 16 LEDs, 15 years of development, one goal: A persistence of vision display stick that fits into your pocket. That’s the magicShifter 3000. When waved, the little, 10 cm (4 inches) long handheld device draws stable images in midair using the persistence of vision effect. Now, the project has reached another milestone: production.

The design has evolved since it started with a green LED bargraph around 2002. The current version features 16 APA102 (aka DotStar) RGB LEDs, an ESP-12E WiFi module, an NXP accelerometer/magnetometer, the mandatory Silabs USB interface, as well as a LiPo battery and charger with an impressive portion of power management. An Arduino-friendly firmware implements image stabilization as well as a React-based web interface for uploading and drawing images.

After experimenting with Seeedstudio for their previous prototypes, the team manufactured 500 units in Bulgaria. Their project took them on a roundtrip through hardware manufacturing. From ironing out minuscule flaws for a rock-solid design, over building test rigs and writing test procedures, to yield management. All magicShifter enclosures are — traditionally — 3D printed, so [Overflo] and [Martin] are working in shifts to start the 500 prints, which take about 50 minutes each to complete. The printers are still buzzing, but assembled units can be obtained in their shop.

Over all the years, the magicShifter has earned fame and funding as the over-engineered open hardware pocket POV stick. If you’re living in Europe, chances are that you either already saw one of the numerous prototype units or ran into [Phillip Tiefenbacher] aka [wizard23] on a random hacker event to be given a brief demo of the magicShifter. The project always documented the status quo of hardware hacking: Every year, it got a bit smaller, better, and reflected what parts happened to be en vogue.


The firmware and 3D-printable enclosure are still open source and the schematics for the latest design can be found on GitHub. Although, you will search in vain for layout or Gerber files. The risk of manufacturing large batches and then being put out of business by cheap clones put its mark on the project, letting the magicShifter reflect the current, globalized status of hardware hacking once more. Nevertheless, we’re glad the bedrock of POV projects still persists. Check out the catchy explanatory video below.

64 thoughts on “MagicShifter 3000: An Over-Engineered POV Stick With A 15-Year Journey

    1. I dunno if 70 Euro is “VERY VERY Expensive”. Kinda pricey, sure, but there are plenty of things on the market that cost more for a lot less gadget. And seriously, ‘overengineered’ looks like an understatement.

      Anyway, the article says “The risk of manufacturing large batches and then being put out of business by cheap clones put its mark on the project” and your comment just underscores the problem: If you invest in developing quality open-source hardware, you will always be undercut by someone making dodgy units with cheap parts using your design.

      1. TL;DR: Hardware is hard.

        @Tane: I see that you know what you are talking about. :)
        We had a wild ride over the last years and learned A LOT about manufacturing and calculation.
        We are going to update the page with some stories that we encountered on our way here..

        1. It is surprising what things cost without labour subsidized under communism.

          Don’t spend 1 minute worrying about the bottom feeders – as they are poor, and can’t afford things anyway.

          Without moving 20k units a month, most people just aren’t going to get a cheap gadget.

      2. In my opinion around 70 Euro is a lot of money.
        But with what I saw it is still a good bang for the buck!

        Apart from this. With 500 Units and 5% defects, tax deducted the margin is just below 16 Euro. For the lead time this is nothing. If all 500 units sold and there weren’t any additional costs. This would be at most 7935 Euro. 2635 per engineer and 14,63 per engineer per month developing time.

        It is not even calculated by a big factor. A cup of starbucks is way more expensive. ;)
        A pizza is made for about 1$ and sold for 10-12$ that is way more expensive.

        69 Euro is a very reasonable price.

    2. Yes, we know and agree.
      But this is how much we have to charge when we are producing in small numbers.
      Getting the price down is one of the top priorities right now ..

      This is something that can be achieved by production in large quantities and using injection molds instead of 3D-printing.
      How to get there?
      Maybe kickstarter or something like this .. but we need to get the word out before that.
      How about Hackaday? :)

        1. Yup, it’d benefit from a redesign, doesn’t need high-end components like that. Dirtest cheapest RGB LEDs rather than smart ones, cheapest controller that will do the job. Bluetooth + MCU over ESP8266 Wifi. Doesn’t need rechargable lithium, most people aren’t going to use it every day, AA cells will be fine. Etc. Change the IMU for a weight on a spring, hitting a contact. Or a mercury switch or similar.

          Some redesigns would make it harder to construct, would need extra labour, so in that case don’t do it, stick with stuff you can make easily as you do already.

          I’d start from scratch and design it based on cheapness over everything else. 40 Euros isn’t pocket money. For 40, I’d want something actually useful to my life. I might pay 20 just for a silly toy like this. Not that being a silly toy is a bad thing! Might make a nice fun gift.

          If there’s enough demand, sure the Chinese will be able to knock them out cheaply, but no need to worry about that. As long as you haven’t invested a lot in production, you can stop whenever you like. DON’T invest thousands! You’re doing it the right way, one at a time, and see if it takes off.

          You might sell some at 40EUR, to nerds who have the money, but it’s pretty much a niche thing. Other problem being, a nerd can likely make something like this himself, and would have fun doing so, especially since it’s open source. Still I think there’s enough who don’t have the time or skills, who would enjoy playing with it, but that price is upper-end, and frankly for rich men who don’t have families or big bills, but do have a toy budget.

          This needs to be cheap enough to be essentially disposable, to be money you don’t mind wasting, because it has no practical use, and you can only play with it so much before you get bored. Doesn’t mean it’s bad, or doesn’t deserve to exist, but you need to know it’s limits.

          I actually quite like it, looks like fun.

    3. I can understand that. what would be a good price for you?
      Just to explain it: because we only made 500 units the cost of one unit was ~40euros. Also we had to account for 5% defective units. How much would you charge?

      1. This looks very much like the previous version with Atmega CPU and TLC5947 (?) chips. This was essentially more expensive than the current version. But you have less LEDs with very wide spacing.

        1. The driver ICs are the TLC5940. I’m not sure how it would be more expensive, as I put together the BOM from Newark (Farnell) 100 qty prices and came up with much less than $50 for a fully assembled design. My LEDs are only $0.39 each – perhaps that’s the difference?

          I know you are on to another supplier now, but Seeed is extremely expensive. If you’re interested, I can offer you the name of a much more reasonable supplier I have used and you could quote your assembly there – just private message me. I don’t think I am going to continue my POV light work, so I don’t intend to compete with you.

  1. Awesome, I have no doubt that if I purchased one I’d become irresistible to other humans, but I have just one request, can it let you select Unicode emoji from a menu?

      1. Maybe allow for longer sweeps and have it display lines of text like ‘don’t shoot’ for protesters and messages by people stuck on roofs during flooding and all such situations.

        Maybe someone should make one which connects to the car’s data port then use the speed taken from there to display POV lines while driving, would sell even when illegal to use in many places I bet. Or would it?

        1. Thanks for seeing this democratic crisis on the planet with the needed Wit and Humor :)

          Long pictures only work out when you the device on the end of a long stick. that allows you to swing it smoothly without straining your arm. There are no hardware limitations on the width of the images. It’s just that the POV effect breaks down when you use more than lets say 64pixel with. As I said: a long stick allows you to make the same distance faster so this works out :)

    1. wow! cool this case would indeed fit nicely and you could even use a larger LiPo to let it run for a week or even more :)

      Did I mention that there is a LED strip connector on the PCB for hackers? so you could easily fit a 64LED high density pixel strip and make a “light sword” the CPU has enough muscle to handle this but you definitely need a larger LiPo since 64LEDs would draw 3840mA when all are white and thats beyond the capabilities of the current 400mAH LiPo.

    1. Cool! Thanks for reminding me: We have to port the Bike Light Mode form the old 8bit firmware to the new ESP8266 32 bit core ;) the old firmware source code can be found here:

      Yes we used bitbucket back then because they were the first to offer free private repos. The new Firmware is of course hostet at github (which gets open source and now does the same (offer free private repos :)

  2. stick to your price guys, let the deep(er) pocket innovators buy the first lots. I know a couple of mates who would think twice dropping 69eur on this. then work your price down…..btw 3d printing the cases production stylee seems a bit crazy! wouldnt it be cheaper to get a company to CNC mill them?

    1. There might be some suitable plastic cases you can buy in bulk, even if it means you need to cut the slot in them yourself. Be much cheaper, and probably sturdier, than 3D printing, since it’s such a simple shape.

    2. For limited mass production. I’d be thinking of a solution where the 3D printing part is minimized, maybe vacuum forming the box with a printed front panel. Or using some of the old project box standbys like sections of electrical conduit, with 3D printed panel and ends to finish.

    1. I know, but I experienced that the more “stupid” I looked when delivering a demo, the people that watched felt less stupid and allowed them self to really try it out ;)

  3. Hello wizard23, overflo and Philipp, I have a bit of experience creating and selling devices and here are my 2 cents.

    Try to reduce the amount of unique components in your BOM; you currently have 9 different resistors and 7 different capacitors. The components themselves barely cost any money but storing them and loading them into a pick and place machine do. More unique components means more money, number of components less so.

    Personally I would ditch the pin protection; It drives the cost up just to protect the 1~5% of the people who *might* want to tinker with the software, that doesn’t seem like a good tradeoff IMHO.

    USB and WiFi seems redundant and seems to cater to the tinker people; I would remove the serial chip and tell people to use the existing FTDI header.

    I would ditch the LiPo and at the very least replace it with a LiFePO4 for fire safety reasons. Preferably I would replace the entire LiPo + charge circuit with an AA(A) battery holder to drive down cost and improve on safety / not getting sued.
    This would however change the battery voltage and would affect the rest of the circuit design…

    All your components are SMD with the exception of the buttons, these need to be wave soldered which means they need an extra pass which costs $$$ and depending on the order they place the components that could also mean some of your components are getting heated more than once which could explain your 5% defective units.
    Yes through hole buttons are sturdier but they are driving the cost up, try to replace them with SMD buttons and find a way to make the enclosure give them support.

    Consider placing pads on your PCB and creating a pogo pin bed to quickly test and program circuit boards, preferably the whole panel. Might want to consider lending or giving the tool to the fab house.

    Try to make your components as big as possible to make rework as easy as possible though looking at your board 0603 seems like a fine choice.

    Consider an alternative for the enclosure, 500 3d printed does not sound cheaper / a better time investment than using injection molding.

    And now the elephant in the room…
    Have you spent time on FCC / CE certification? This is not a course to be taken lightly. I am not trying to discourage you but selling 500 devices with WiFi and rechargeable LiPo batteries without certification is problematic.

    I wish you the best of luck with your product and I hope you find my advice useful.

    1. One comment about AA, since I used it on my POV light. You pull a hell of a lot of current from an AA when all the LEDs are white. It’s doable, but requires the right regulator IC. The work in finding the right regulator IC morphed into my BooSTick project. On a side note, I finally shipped the last BooSTicks from the Kickstarter campaign…whew!

          1. Well, you’d need two AAs in series to get the voltage high enough to turn on the LEDs, or you’d need a boost regulator, and that’s going to draw more current from that single AA than 1 A. The modulation of the LEDs helps a lot, and capacitors on the LEDs help too. However, I’ve come to the conclusion that they are driving their LEDs extremely hard because they want incredible brightness (which should be obvious from their video given how bright it is outdoors). That’s great, but it does force them to make certain power decisions that preclude the use of a single AA.

      1. True, with 16 LEDs pulling 60mA you’re close to pulling 1A which is a lot. That could be solved by adjusting the max brightness and perhaps some big caps close to the LEDs to catch the peak currents. Choosing the right (switching) regulator would also be important. Regardless, discarding the LiPo would amount in lower light brightness which is a compromise which may or may not be acceptable.

        Congratulations with your kickstarter campaign by the way! :D

    2. You have small LiPos in a variety of consumer devices: Cameras, Phones, Bluetooth headsets, Camera-Goggles. LiFePo is good for E-Vehicle batteries with kWh of energy, but in this case it is quite small. AA batteries are also very big and heavy. I personally hate devices which do not function with rechargeable batteries and really prefer Li-based batteries as having no memory effect or self discharge.

      A test adapter with pogo pins is in the range of 10.000 euros, especially if it should be that big to test a whole panel. That would be another range of devices/production numbers.

      1. The biggest problem with LiPo batteries and charge circuits is that it’s a hassle to get them certified, while seemingly perfect in the design phase for their characteristics I would avoid them for devices where price is an issue. (Certifying your product costs a lot)

        LiFePO4 is a good compromise which offers greater safety but in all honesty I don’t know how much they effect they have on getting your product CE/FCC certified and would be glad to know if someone knows the answer.

        I understand that not being able to recharge your device is a hassle and that LiPo’s offer the lowest self discharge of all rechargeable (consumer) batteries. I should note that non rechargeable alkaline batteries have lower self discharge though. You could put in rechargeable batteries but that does increase the hassle of charging the device up.

        As for the weight it would be possible to opt for smaller batteries than AA’s such as AAA’s or even AAAA’s.

        Perhaps it would cost 10.000 euros if you let an external company make one, but making your own is MUCH cheaper.
        It is quite possible to manufacture such a test adapter yourself which would drastically reduce the price to the 100~350 euro range depending on your wishes. Take a look at sparkfun, they are quite proficient in building their own cheap test adapters. You don’t even need a very big PCB to test/program several devices, just make multiple small one’s and hook them together with whatever interface you find easiest; SPI, uart etc.

        I know it’s really difficult to push the price down because I’ve been there and I know the pain of making compromises but compromises are the name of the game when building products because people don’t want to spend a lot of money. (1st post of this article being the prime example)

    3. Sorry to say it so bluntly but you are really mistaken.

      Using a AA battery???! This is the worst for our environment! I know that LiPos have an ecological footprint as well but at least they can be recharged 1001x and then they still have 200mAh capacity left which is enough to get you through a night of low power POV action.

      We also print with PLA instead of ABS for exctly the same reason.

      Did you ever fiddle around when exchanging empty AAA (or AA it does not matter) batteries? Well you wont fiddle with the MS3000 ;) The problem just does not come up.

      I’m sure our customers know that a good LipO battery is worth 5$ in the long run (5$/1001 == 0.04 cents per charge:) One charge gets you through a weekend of POV action ;)

      Ditching the pin protection, to save costs???! Well try explaining that to a 6 year old kid who just bricked their birthday present when writing their first custom firmware :| Not cool…

      1. The down side to PLA, is the glass transition temperature. Have you ever left a PLA piece in your car on a hot summer day? It is very likely to deform from the heat.

        I agree with you on the pin protection. That is a must, and something that is usually missing from most hobbyist level designs. To me the few pennies that it costs is insignificant compared to the support troubles that can come from not having the protection.

        1. Yup. PLA is not perfect, but then again: What Is? We are currently working with an Industrial Designer to create an injection molded case. This will be a much shinier and sturdier case :)

  4. €69,- is a lot of money for something for a handful of bright lights. And although I fully understand your standpoint of costs, why on earth would anybody pay this amount for a bunch of LED’s that can build an image with a very poor resolution and require you to wave your arm pretty fast in order to produce an image. I know what I’m talking about, please let me explain:

    These devices are not new, I’ve build one in 1998 and the idea wasn’t new at at all at that time. The user programmed the texts over RS-232 serial. I used my Amiga 1200. I’ve build 2 (one for myself, one for a friend), they were based on the trusty 16F84. Both devices are still functioning.. my Amiga is not. The sensor that syncs the image to the motion of the waving arm was just a simple weight that closed a switch and nothing more. It could store 8 texts of max 9 characters each, no images. It used a single 9 volt battery that lasted for months if used in standby or longer if switched off.
    I thought it could be useful to attract attention in bars. For instance the attention of the bartender to signal my order… but it was considered nerd-like at the time and perhaps therefore it never has rewarded me with any attention of any girl. That was the time that I began to realize that being a hacker/maker/nerd isn’t as cool as Hollywood sometimes shows us ;-)
    Therefore I find the promo of a girl waving it to a guy very misleading (that’s a joke, the video was very informative).

    Anyway, today we have phones that can produce a very bright image and have very large screens and a build in battery. They can do anything and more that the magicshifter3000 can without looking like and idiot waving your arm in a much to unnatural way. So in other words, it’s too expensive for what it does in a time where it’s functions is/are superseded by other devices. Development of 15 years shows great determination but in the end… it is just 15 years too late.

    Then again… I could be wrong (it wouldn’t be the first time)

    1. > And although I fully understand your standpoint of costs,
      > why on earth would anybody pay this amount for a bunch
      > of LED’s that can build an image with a very poor resolution
      > and require you to wave your arm pretty fast in order
      > to produce an image.

      It’s a mystery to me as well :)

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.