The part I love the most is that the machine doesn't only create order out of chaos, it then flushes all the order down the drain and does it all over again, every time.
As much as i like your interpretation, I am shocked by this iterative example of first doing then thinking which results in a huge waste of money and resources…
Yes, it is a clever solution. The challenge however is that it appears to take over 2 minutes to sort a new number (based on the last 2 minutes of the video).
It works, but it creates a problem. The time it will take to arrange the next minute display is variable. Because the process is relying on randomization (did I get the correct coin flip of a black marble? Nope another white one, kick it off the elevator and try again), there's going to be a statistically expected amount of time to "print" the time. And looking at those timelapse clips there's a good chance of it taking more than 60 seconds.
@@Satarack Yeah, I think he needs to sort the returning marbles by color and then place the correct color into the proper elevator position to make the process less random. White and black can be done on separate levels of the elevator.
Very clever way to identify the marble colors! The sorting “algorithm” is very inefficient because it’s randomly waiting for a white marble to land in the right spot, and has to reject all of the black marbles, which occur more frequently. The efficiency is also hurt by dumping the rejections back into the starting pile because they have to be sorted multiple times. Instead if you sort the marbles into white/black baskets, you’ll sort each marble once per rotation, which will be more efficient and allow for a faster clock refresh rate. Very nice project!
A simple improvement would be to equalize the proportion of white and black marbles closer to 50:50. Yes, the display uses many more black marbles, but the limiting factor to complete the image is the time for the slowest row. The rows second from the top and bottom have about as many white as black, so targeting that ratio will optimize the total fill time.
Вы правы! Эффективность с помощью сброса шариков снижается. Они должны сортироваться внизу или наверху и с помощью сепаратора переключателя заполнять индикатор. Да и лифта достаточно одного или двух. Если сортировка внизу - то нужно два лифта. Наверху нужны два бункера-накопителя.
@@niknik5616 Translation: "You're right! The efficiency with the use of dropping balls decreases. They should be sorted at the bottom or the top and with the help of a switch separator fill the indicator. Yes, and one or two elevators are enough. If the sorting is at the bottom, then two elevators are needed. Two storage bins are needed at the top."
He needs to use software to determine if the 'rejected' ball will be needed next on that same channel, if so have it go down a return feed that automatically makes it next in line again. This could hugely drop the time to display a time without much more hardware complexity. It's too slow rn, will never work. That or just hugely speed up the lift.
I still recommend bearing balls. Even if you’re not pursuing the magnetic separation approach, they’re machined to much smaller tolerances and you can blue the black marbles and leave the white ones silver.
For your gate problem, if the left 180-degree track twisted like a mobius strip, I think that would keep your slopes for each marble path more consistent and avoid the problem with the low slope stopping the marbles along the top rows of the time display.
Getting it to work is the first step. Getting it to work Quickly is the second step. And having enough marbles to actually have the next number in place when the change is ready is still part of Step 1.
There's no way he can't crank it up to go faster. Probably just wanting to make it accurate and reliable first. It's clear to see the marble lift is just going really slow. He can just crank the speed up on it and probably get it at least going twice as fast.
As a long time viewer of the Wintergarten Marble Machine I look forward to the many episodes it will take to get this right. Enjoy, Ivan - I intend to.
You could always try some sort of ball bearing. They would probably be more precise then heat 50% of them up and quench them in oil. Or brass and stainless could be cool if it could handle the weight. This is awesome and just keeps getting better
Great demonstration of rapid prototyping and lots of satisfying components. I love the solenoids and, as always, your enthusiasm! :D Having sorted many a marble in the past few years, I've found glass marbles always have a fair amount of variance because of the manufacturing process, so whether it's you or the marble supplier, someone will just have to test and sort out the best marbles. For absolute precision, you could consider pairing some super precise balls like White POM bearing balls and dark steel bearing balls. That might compromise your vision though... glass marbles look and sound really nice! The audible difference between 'white' and 'black' balls would be an interesting element, and POM vs steel would introduce the ability to sort by density. Just spitballing haha, I'm excited to see the rest of the journey!
@@ivanmirandawastaken Theres a pocket83 video about sorting marbles, that may be helpful so you don't have to wait for a marble to jam before you remove it.
3:52 when we did lime followers for Uni no one taught us how, so I used these light-detecting resistors. The tip I got much later is: make your LEDs shine at a specific frequency, then put a band filter on your readings from resistors. Pretty sneaky, and AFAIK should work when flashing LEDs at 5Hz
If you separate the marbles upstream, you can then speed up the row constructions a lot. As marbles are discharged from the clock, separate them in black and white trays, with the black at a lower level. As the conveyor goes up, it will fill with all black, then you eject some marbles to create voids and move the row to be level with the 2nd tray, where white marbles fill only the voids. This way you'll end up with complete rows every time.
Then the elevator needs to switch functions: First run it and punch out all marbles, either in the white or black tray. Then swap to the function as you describe it. After displaying the next time, switch again to the filling function. Otherwise, filling the white tray will be filling all the rows with black marbles.
Nice idea but the separation process would be more error prone than building rows as he is. Time isn’t an issue at this point, so it’s adding complexity for no benefit. Accuracy is the key here. Classic engineering.
@@genehenson8851 Don't get me wrong, the clock is beautiful and I love the ingenuity on the engineering side, but right now it is non-deterministic. I know it's kind of pedantic, but getting the right configuration for each minute could take way more than one minute, as it depends on the probability of black and white balls falling in the right spots in the first place. You can see this partially happening as some rows are lagging behind the others and in the earlier designs it was more obvious as rejected balls fell closer to the conveyor. Of course as paul8731 highlighted, preselection is not a perfect cure, as it requires that a row is filled without voids (and that adds even more complexity).
Amazing. Too late now, but you could have the display using vertical columns instead of rows, this would allow a single digit to be changed instead of all 4.
You should consider a vertical layout for the marbles, so that you can update every number individually instead of having a collective - that way you can speed up the process to work automatically in real time without pauses or human intervention. It would be really cool!
I was hoping he would just use an efficient algorithm to update the display. Instead of flushing the whole thing every time it would just let out enough marbles from each track to reuse whatever it could. And then to make updates a bit faster he could add another set of gates on the upper part of the track and release the whole update and flush the bottom marbles at the same time.
I was thinking the same thing! Or maybe in this configuration have the next minute already loaded in the upper feed so when the minute changes it flushes the old time and gravity feeds the new time and the elevator is already building the next minute. Just has to be able to load the next minute in the upper feed portion within 50 seconds to keep pace.
Можно рассмотреть ещё вариант: разделить на два табло - часы отдельно и минуты отдельно. Цифры означающие часы меняются редко и поэтому корректироваться будут не часто. Для эффектности показания сбрасываются, все шарики съезжают, но часовые шарики делают кольцевой пробег и быстро становятся на свои места.
Ever since the last video, I've been wondering what mechanism you will use to select the right marbles for the right track. I would never have come up with this "subtractive" approach where you remove the marbles you don't want. What a cool idea!! 😁👍
This is really cool, but I would've used a slightly different approach. I'd have 2 layers of marbles where they load into the conveyer - the bottom layer black, then punch out holes where you want white marbles, and then an upper layer that loads any open spots with white. That way you don't have to depend on the randomness of getting the right color into the right tracks, which takes an indeterminate amount of time. You can load up the marbles you need in each row in constant time every time. You also need to sort the "discard" marbles back into the lower tray (black) and upper tray (white).
@@gorak9000100%. Separate all black and white at time of discard, and build what you need from the two color sources. His method sure looks cool though
@@haileycollet4147 His method basically has all the parts needed to do the other way, they just need to be re-organized. Move the color detection to the discard chutes, and he needs a switchable gate to direct discard marbles to the white tray or let them continue down to the black tray. Then he needs to 3d print the upper tray for the white marbles and it should work. Maybe he also needs a way to work with discard marbles one row at a time too. Much easier to do this kind of re-organization for optimization in code where you can just copy/paste lines of code to move them around!
Maybe you could alternate the colors: Minute 1 is white text on black, and Minute 2 is black text on white, etc. That way, the number of marbles should be the same (50% of each) and the likelyhood of finding the right color for the next minute goes up since the previous minute's background is using a lot of marbles of that color.
sounds like an idea that could work. I thought about some way to "mix" the marbles, as some rows out of pure statistical bad luck took way longer to get the nextright marble than others. Regarding the marble quality, I wonder if appropiatesized plastic balls instead of glas marbles would work better, although the sound would be way different.
The plastic ones would be way more consistent but the weight vs coefficient of friction will be greater and slower in the channels and may even get stuck easier
@@erikpip3024 Yeah, good point, the weight might indeed be an issue. Although the difference is less significant than i thought. Glas is about 2.5 g/cm³ while hard PVC is about 1.4 g/cm³. Other plastics like ABS are even lighter, around 1 g/cm³, others may be heavier (but way to soft in comparison). To be honest I somewhat expected glas to be multiple times as dense (and thus heavier). But in conjunction with any plastic being "softer" as glas marbles, friction indeed may add up on top.
You'd do well using a barcode optic to detect ball color. They are more resistant to ambient light that affects discrete LED sensors. The only way around that is to clock the LED optic so you know whether it is reading room light or reflecting off of a ball
I like that you talk about issues you are having with the clock design right away, as soon as they happen. It's interesting as you get to immediately see the trouble shooting steps you took to try and over come it.
I truly admire the skills, the abilities, to complete this project. Personally, I can see practical usefulness in several design elements. The reality is, that I live in a world where the sheer number of real-world challenges and desperate needs, prevents me from embarking on a project like this one. As Mike Patey (see his YT channel) says, 'Machine it, Draft it, Build it Test it, Break it. Every time something new gets built, the entire world advances'. My regards to you, Ivan.
*@Ivan Miranda* - Improvements: 8:24 If you reject the marbles to the left, so they get into the very back of the queue, you would get even more circulation of marbles. - -:- - It would also be useful with some kind of "buffer system", so you can store a rejected color & then swap it in later when needed, it could make a LOT more elevator utilization. 10:22 You could copy the lower straight piece into the upper piece, and build the next time in the upper part, to remove the pressure from the marbles in the bend.
Yeah I don't know why it has to roll down that long straight away to get to the bend I'm not sure why he doesn't just set the elevator on the left side and have it drop down into where the bend is... I guess it's more for show to watch them roll all the way over
I had the exact same idea. BUT, i just made another comment saying that if i was to have made this, i would have build it differently with less marbles. All you need is a track which holds the combinations needed to form a side of a digit. Split tracks. So for every number you have a tracklayout that can build up the number. And the marbles are stored in a tube or something like that. And so every number of the clock, 4 in total basically, you have a different track. All you need to do then is have controllers for those tracks that put a new set of marbles in it when the number needs to change. You only need to have 0 to 9, so not that many marbles per track. Sorry if it does not make sense. In my head it does. The only problem i have is: I can draw it but it still would not make sense without the electronics drawn in and the circuits. O, and the second problem is: I have no time to build it myself. But the idea i got is almost like the flap down clocks, where a card flips. @@kmktruthserum9328 PS: btw, the long ramp does have a purpose. I just thought of it. He needs to be able to change from 23:01 to 23:02 in 1 minute time. That is not going to happen if he does not prepare it in that minute, to instantly change over basically. Therefor the longer ramp. (or so i think. I litterally came up with it just now but it does make sense. Otherwise he can not change from 23:01 to 23:02. Because that requires time to set up.
the longer ramp does have a purpose, it is to set up the next minute. How else can he set it up? It has to change from 23:01 to 23:02 in an instant. That can not take more than a minute. Otherwise it wont run ontime. So it has to be prepared. So with this setup he needs to have a long ramp to set up the next minute. Heck, the ramp may even be too short to keep the clock running ontime. Let us see how he will manage to keep it running ontime. @@kmktruthserum9328
@@kmktruthserum9328 It's so it can queue the next minute up or else you'll have to wait and sort the next minute on the elevator itself, which is too much.
To keep the marbles from getting stuck in the curve, you should have the next time queued up on the straight section on top, not on the curve. The gravity will work better.
something that might help with the speed at which the numbers are constructed is switching the background and foreground numbers each time(assuming about a 50/50 split)
@@SVSQZ yes, but at the risk of having waaaay too many blacks and not being able to populate a white 'pixel' if they can't make it to the front. If the mix is roughly 50:50, and alternating front and back colors it would kinda solve a lot of issues as the majority of the next color you need would be the most available to move up the chain for the next set.
@@SVSQZ Maybe but wouldn't reducing the number of white marbles also result in the collection thing taking longer to pick up a white marble in the correct lane?
@@SVSQZ Hrm, if p1 is the probability the display needs a white marble, and p2 is the probability a marble in the tray is white, then the chance of a correct marble is p1*p2 (white) + (1-p1)*(1-p2) (black). But that's unhelpful to optimize because the highest chance of a correct marble is achieved by having 100% of the kind that's more commonly in the image. So instead we probably want to figure out the average number of marbles pulled before getting a correct one. Luckily this is a geometric distribution and fairly simple - when a white marble is needed it's 1/p2, when a black marble is needed it's 1/(1-p2). so actually we want to minimize f(p2) = p1/p2 + (1-p1)/(1-p2). Looking at the way the image is constructed, there are the 7 segment display parts (4 marbles each) and the black background (48 marbles per digit) so average about 4 white to 15 black marbles, so p1=0.2 approximately. Putting this in the function gets p2 ~= 33% There is a catch here though, that's only the marbles in the tray, there are also marbles in the displayed time and the number about to be displayed. So really you want about 80% of the marbles to be in that 20% white proportion, the remainder in the 33% proportion, making it overall about 22.7% white marbles. If you alternate displaying white and black numbers this gets more complicated of course, since the marbles in the in-tray are the opposite distribution to the displayed number.
Love the work you are doing. Keep it up! I just don't understand one thing. You could install infrared couples a row below solenoids and only then take a reading when the lift stops. I think it will work as well.
This would make for a really cool interactive art piece! 🎨 Imagine a museum with a computer containing something like a 200x30 pixel grid where you can draw in white with a black background, and on submit, it can "print" your picture.
The Precision Plastic Ball people based in the UK are my usual go-to for balls/ marbles. The Phenolic resin ones might be a good shout for this project as they have a pretty consistent diameter. Looks cool, can't wait for part #3! :)
This is sweet! Though I had a thought: If you had the time on the right of the elevator instead of the left, you could then fill the minutes in last. This could be helpful if you had a way to clear the minutes without clearing the hours (perhaps where the : is?) Then only the minutes get reset each minute, which is faster. But then every hour you'd get a really satisfying sound of all the marbles returning. Edit: Just thought this would be easier perhaps, if the numbers are assembled vertically (in columns) instead of horizontally (in rows). This could also solve your issue with marbles getting stuck.
Maybe by having the time display as a shallow ‘V’ with the ‘:’ at the bottom, so the hours are put in from the left and the minutes come in their own track from the left. Both of them empty inwards toward the ‘:’ so you can control either side individually
@@billkeithchannelwell atleast you was watching something that related to marbles, i was a qomen playing with baby jaguar 😂 , so i have no idea why im here lol But i liked the vid and definitely subbed
I really enjoyed your video! Some thoughts: your clock "refresh rate" is bound to the randomness (mainly) of a white marble going to the right place and the speed of the lift. I would suggest moving the actuators up, use them to short the marbles in two "buffer baskets" (white/black) and from there rapidly feed your "dispaly". (Actuators to each color letting marbles drop in place). Again, super nice project and great video!
Yeah, i feel this random approach will just be too slow to have the next number ready in 60 seconds. When he started talking about differenciating black and white marbles, i thought he was gonna implement a sorting method.
You should drop the marbles a bit higher and have a ramp going left to move the rejected marbles back to the back row. This way you get a better color mix of marbles and will fill faster.
Love it! It would require a huge redesign, but if you made the display out of vertical runs instead of horizontal runs the marbles would position themselves much more quickly … possibly even fast enough to display seconds as well as minutes a long way down the design timeline. Great concept, looking forward to the finished article!
I would like to suggest to use the effect of rejected marbles stack up in front of the elevator to create black rows as you actually need much more black marbles than white ones. If you actively dumped a row of either white or black marbles from the rejected section when your system needs those you might get to cut down on your wait time until a certain type shows up at a certain position
I was thinking about this problem too and ypur idea sounds like it could work! What if he has an only black elevator with completely seperate white feed in. Then when white is needed, the appropraite blacks are dumped out and the whites are released to fill the gaps. You should be able to create full clock faces within the elevator already!
I'm thinking two rows of solenoids. The bottom row only rejects black marbles from where white marbles should be. The other row, much higher up the chain, rejects unwanted white marbles. The whites are pushed onto a tray that ends up filled with only whites. The tray leads back to the elevator, slightly lower, between the two solenoid rows. If the fill rate at the bottom is perfect for the black marbles (i.e. no empty slots), and there are enough white marbles on top to fill every gap, you might not even need controllable gates at the white tray. The whites would just fill all gaps left by the solenoids kicking out black marbles where they're not wanted. The result could be really efficient. If the fill rate at the bottom isn't perfect, the whites will fill too many holes, but they could still be detected and re-rejected by the white solenoid row above the tray.
@@arantala unfortunately, if the whites are not sorted at the bottom and brought up to the white tray, i think there would be the chance of all the whites being in the 'system' and the tray getting empty.. the whites could het stuck in the black mass at the bottom and never come up again. Otherwise a very good idea! (Also if i am wrong, please tell me haha)
You should consider putting a mixer of some sort between the ball return and the lift... this would prevent a build-up of 'rejected' pieces and make the very slow marble picking time at least more predictable!
Here’s an idea… alternate the background from black to white for AM and PM. Black Digits for the AM and White Digits for the PM. Not saying there’s anything wrong with the 24hr format, there’s just potential to take it an extra step further with the 12hr format. Just Imagine the transition of a white background to black from 11:59 am to 12:00 pm.
That would require having double the marbles in the system and a great many more rejects. At the moment say 90% of the display is black so 10% of the marbles need to be white. Put 20% white marbles in the system and on any channel you'll get black 8/10 times and white the other two you'll fill up the blacks quicker and have whites available when you need them. 50/50 the rate you can fill the blacks is reduced by 30%
@@zyebormAdding more marbles would be the easy part. The problem is he is not sorting the marbles efficiently enough. If he could separate the marbles before they get onto the elevator, or hold the rejects in a buffer for later, he could easily switch between black and white backgrounds.
Epic build! Just one little thing, how about building the display with marbles that come down vertically. That way you can purge each digit individually?
This is awesome! May I suggest using two separate feeders, one for black marbles (do that first, punch out the spots that should have white marbles), then a second one that only has white marbles to fill in the spots that should have those (punch out any white marbles that got into spots the black marbles didn't fill before, but should have.) Can't wait to see the finished clock.
This is so creative and so different and so cool!! I love watching the process of producing parts and then making the necessary modifications to fix the issues -- learning and designing on-the-fly! And some of the intentional (or maybe NOT intentional) comedy makes it fun to watch.
@@bertjesklotepino I don't remember exactly off the top of my head, but I think I remember seeing steel balls rather thanks traditional marbles. It would make sense that they would have to be steel balls so they don't break or chip like marbles would. I just called it a musical marble machine because it sounds better and gives a better mental picture than calling it a musical ball bearing machine.
Well, if i was to build anything like it, i would prefer the steel balls because they are uniform in size and indeed don't break or chip. And understandable why you called it a musical marble machine because it basically is. Who cares what the marbles are made out of? And in the case of this clock, i would for sure have gone with steel balls because a clock needs to be accurate to keep time. No way you gonna achieve that with marbles. They roll great sometimes, but sometimes they get stuck. Won't happen if the balls are uniform. @@davefenton102
@@sgsax yes, true, but by that measure, Babbage's Analytical Engine didn't technically "work" until someone actually built it. However the idea worked and was still pretty awesome, even though it was just an idea on paper for a long time. I reckon they should collaborate on a musical marble clock
Very cool. I like the relative simplicity of the rejection mechanism! I did notice that it took about two minutes to build up the time though. Looking forward to seeing how you account for that!
Yeah, the randomness of "waiting for a white marble in this exact track" is going to inevitably cause issues. There's always going to be a possibility of never seeing a white. Might be easier to have the sorting done *before* the elevator and put into two separate buckets that are pulled from as needed.
A clock that is right every five minutes is better than one that is right twice a day. In the days of analog clocks, we usually said it was 20 minutes to 3, and not 21 minutes to 3 or two thirty-nine..
I feel like using the marble rejection thing to sort the marbles into colors and then placing them would be a lot more efficient in terms of building the current time before it’s wrong
Agreed, the current system feels like it's dependent on RNG, a healthy filtered system and smart marble reservoir could resolve so many potential problems
While you're right... I think it kind of takes the fun out of the whole thing if it's just sorting them into two separate color pools like that. Really if he can just speed up the whole thing by maybe 20% then it could just work as is. I feel like it could definitely move faster.
you can optimize the slope by using a bradistochromic curve! it would make all the marbles take the same amount of time to fall regardless of placement (i know it’s misspelled)
Why not just use Collumns instead of rows? The marbles would just fall down, when flushed And it would allow to keep the higher digits therefore leaving more marbles in the tray for the smaller digits to update Edit: Maybe if all the balls would roll into those Collumns from right to left, allways filling the rightmost before filling the next one, the leftmost digit could slowly fill within ten hours, while the rightmost digit could update much much faster I am tempted to implement such an update-algorithm
@@lorenzumhofer566Remember that each lane comes from a sorting mechanism. If you want one lane on each column, that would mean a lot of sorting device. On the opposite if you want only one lane delivering all the columns from right to left, that sorting lane will be too busy.
This is amazing! So many problem solved in different ways. Designing, 3D modelling & printing, coding, electronics, general physics, and probably more I can't think of. If it wasn't so loud I would want to build one for my home.
Maybe for version 2 you could have 2 chain lifts. 1 for white, 1 for black. Allow the ones you want up the chain and pop out the rest like you are now. When the time changes and the mix of marbles drops, use something to sort them back into their separate colours to be sent off the the correct hopper.
Bravo this is such a great video. I don't know if you do all the production yourself or if you have a big team behind you but I love how this whole story was presented. Somehow it made a device as ludicrous as that contraption seem like something anybody could whip up on a weekend!
You need to swap it around so that the minutes fill in LAST and can be "dumped" first. (keeping the Hour in place). ALL the digits can be dumped on the hour . and the entire display "recharged." OR you can have 24 columns of marbles in groups of seven that drop down from ABOVE into the display area, then, after a minute, drop down to be recycled up the back to be dropped down again. Hummmm. . . Now YOU have inspired ME! ;)
I think part 4 will be to fix the situation where the clock fails because it cannot get all the marbles in a channel in time. There is, after all, a non-zero chance of no white marbles entering a channel for a minute, and with 1440 minute changes a day, that will happen... My bet is on 08:58 :D You can prevent this from happening by keeping the marbles within their channel and ensuring there are at least as much marbles of each color that you need for two timeframes (but a few spares will speed up finding the right color.)
Based on the multiple gate setup, I wouldn't be too surprised if the end goal is to queue up multiple minutes in advance, thus smoothing out the variance in slow and fast cycles.
just using the elevator to sort white and black, and having them going into 2 hoppers at the top which feed white and black queues for each row with selection gates that will select for a single white or black marble at a time for each row would be the reliable option, that relieves the elevator speed from affecting the selection speed making the actual time selection real time predictable.
I'm loving this project! I know it's not cheap but maybe you could get metal balls coated somehow if you need better tolerances. I'm not sure about the durability of the coating with the marbles being dropped constantly. I'm starting to think that switching to a vertical layout might make more sense since you could put the conveyor directly behind the clock, it will make filling the channels faster, and overall shrink the footprint significantly. A vertical layout also might make the marble tolerance less important since there will be more falling than rolling.
Hi Ivan, I think that you can eliminate the top and bottom lines because they will always be black. I don't know if that gives you any other benefit apart from saving marbles, but I needed to mention it.
Yeah they are, that’s a really good idea too! A display visible even in the dark would be awesomeThe only issue is that most glow marbles are plastic, and I think he’s relying on the weight of glass marbles to keep them moving smoothly through the system. He could always use uranium glass marbles, which shine very brightly under UV light. They’re only a *little* radioactive :)
This is not an engineering project, this is ART! I love it - the pt 1, I was like "what a crap". Now I am fascinated. I looks amazing! Sorry for doubting you.
YES! After Pt.1 I thought, "Yeah, but you're going to have to tear out a huge section for the sorter." I thought about this "rejector" idea but didn't think it could be done so simply. This is quite brilliant!
Cool work! While it works eventually, it feels like a super bruteforce way to achieve this result. May be better to store marble in two separate reservoir based on their color, and dispatch the right marble to the right channel. If you end up having two elevators in parallel, each with their marble color, you just need something at the top to push either a right or a while one. Or you could have two levels to "load" a while or a black marble into the elevator.
Having a second level for the white marbles seems like a really good idea! If it's above the black ones, the rejection mechanism can stay the same, and gaps created by it will be filled with white marbles automatically. Then the only problem to solve is how to separate the white marbles into their own reservoir.
Have you tried putting the 'staging' section on the top straight, instead of on the curve? I think the issue you're having is that the marbles stacked on the tight curves don't release as fast, but it might go fine if you stack them on the straight so they're moving when they hit the curves.
This is a very interesting build I’m looking forward to seeing the final result. It would be a very expensive alternative but you could try searching for silicon nitride and zirconium oxide ceramic bearings instead of glass marbles. they’re black and white, precisely round and would be less brittle than glass. Good luck 🤞
3d printing combined with some basic electronics, make one of the coolest things ever, people are so incredibly creative with use of things at their disposal. can't wait to get my 3d printer
This is awesome. Need to make 2.0 where the marbles build the numbers from right to left, so you can create a way to only change the minute digits as the time changes.
@@TheArachnoBot Since filling top-down would require almost entirely reworking the marble track anyway, a slightly smaller version of the release gate already being used to reset the display could be put under each digit. When it’s time to change that number, the gate(s) opens and dumps the marbles, quickly closing afterwards to allow the next one to build.
Great content! One recommendation: i would rather hear how you plan to solve a problem and then why it didnt work instead of this was my solution and it didnt work. Even if you film afterwards please frame it as a possible success until its known to the watcher
@@notahotshot my comment wasn't a criticism, it was a recommendation of improvement to someone who may or may not listen to my comment. How is your perspective helping anyone?
@@vindrmasuta as long as it doesn't take 3 minutes to make either display. Don't forget that with his design there is no way to predict how long this will take because of how rejected balls are returned. Even putting them to the very back would be undesirable for some numbers.
How would you determine this? The 'rejected' pieces might be the desired pieces for the next number (and they are somewhat randomly distributed back into the mix) and with the return mechanism there is no way to guarantee it doesn't have to cycle through every waiting marble to get the correct time. Whose to say he doesn't have bad luck and one time it takes 5 minutes to build the display? or 10? @@notahotshot
Yup, cool gimmick unfortunately this will never work as a real clock with his current design. You'd need a faster way to pick marbles, or just speed up the lift A TON.
Que lindo y que fácil es trabajar con presupuesto ilimitado. Se ahorra muchísimo tiempo en diseño. Porque es cuestión de probar y probar y probar hardware hasta que funciona como uno quiere. Sin tener que darle cuenta de los gastos a nadie.
The amount of creativity, ingenuity and technical expertise that went into making this project is commendable. This video is a complete pack of thrill and is a delight to watch. The wise use of different electronic components and the genius way of putting it all together to put up a wholesome show is really worth appreciating. Keep up the great work! Eagerly waiting to learn something new from the next part.
You have to be loved for the creations, edification, construction and portrayal in the video. Thank you very much for this type of technology and entertainment. You should exhibit the machine in Germany at the Mathematicum in Giessen. Thanks so much…
The part I love the most is that the machine doesn't only create order out of chaos, it then flushes all the order down the drain and does it all over again, every time.
"every time" pun intended?
Has to be intentional 😂
A clock for showing what time it was 3-8 minutes ago! very handy!
the bogosort clock
As much as i like your interpretation, I am shocked by this iterative example of first doing then thinking which results in a huge waste of money and resources…
I was so interested on how he was going to do the marble sorting. I didn't expect him to make it in the elevator chain. Super clever.
yeah. I kept saying....uh... big part needed next. great work!!
Yes, it is a clever solution. The challenge however is that it appears to take over 2 minutes to sort a new number (based on the last 2 minutes of the video).
me too i was like how tf he gonna fit a sorting mechanism into that tiny space where the rack is loaded
It works, but it creates a problem. The time it will take to arrange the next minute display is variable. Because the process is relying on randomization (did I get the correct coin flip of a black marble? Nope another white one, kick it off the elevator and try again), there's going to be a statistically expected amount of time to "print" the time. And looking at those timelapse clips there's a good chance of it taking more than 60 seconds.
@@Satarack Yeah, I think he needs to sort the returning marbles by color and then place the correct color into the proper elevator position to make the process less random. White and black can be done on separate levels of the elevator.
Very clever way to identify the marble colors! The sorting “algorithm” is very inefficient because it’s randomly waiting for a white marble to land in the right spot, and has to reject all of the black marbles, which occur more frequently. The efficiency is also hurt by dumping the rejections back into the starting pile because they have to be sorted multiple times. Instead if you sort the marbles into white/black baskets, you’ll sort each marble once per rotation, which will be more efficient and allow for a faster clock refresh rate. Very nice project!
👍
A simple improvement would be to equalize the proportion of white and black marbles closer to 50:50. Yes, the display uses many more black marbles, but the limiting factor to complete the image is the time for the slowest row.
The rows second from the top and bottom have about as many white as black, so targeting that ratio will optimize the total fill time.
Вы правы! Эффективность с помощью сброса шариков снижается. Они должны сортироваться внизу или наверху и с помощью сепаратора переключателя заполнять индикатор. Да и лифта достаточно одного или двух. Если сортировка внизу - то нужно два лифта. Наверху нужны два бункера-накопителя.
@@BenAlternate-zf9nr : right, but actually you cannot predict where marbles are going unless there're sorted before
@@niknik5616 Translation: "You're right! The efficiency with the use of dropping balls decreases. They should be sorted at the bottom or the top and with the help of a switch separator fill the indicator. Yes, and one or two elevators are enough. If the sorting is at the bottom, then two elevators are needed. Two storage bins are needed at the top."
I hope this ends up in a museum and is maintained for decades! This is such an amazing and unnecessarily complicated clock! I LOVE IT
Agreed, same thought.
He needs to use software to determine if the 'rejected' ball will be needed next on that same channel, if so have it go down a return feed that automatically makes it next in line again. This could hugely drop the time to display a time without much more hardware complexity. It's too slow rn, will never work. That or just hugely speed up the lift.
If its in a museum I think it should be a aluminum version, as 3D printed filament will eventually warp or go bad
Good idea.
Needs to be done.
The must be a museum of chronometry somewhere that would give it a good home.
I think it'd fit in better in an art gallery!
I still recommend bearing balls. Even if you’re not pursuing the magnetic separation approach, they’re machined to much smaller tolerances and you can blue the black marbles and leave the white ones silver.
Brilliant
That was going to be my second recommendation, after doing something to stop those _glass_ marbles from repeatedly landing on the floor.
Will his structure handle that kind of weight? Steel is MUCH heavier than glass.
...but then we miss out of him saying the word marbles :p
@@tomvanbroekhoven1821are we pedantic enough to not call them marbles if he uses ball bearings?
For your gate problem, if the left 180-degree track twisted like a mobius strip, I think that would keep your slopes for each marble path more consistent and avoid the problem with the low slope stopping the marbles along the top rows of the time display.
Took the clock 132 seconds to display one minute.
I absolutely love the creativity, ingenuity, brilliance, and awesomeness of this clock anyway.
Getting it to work is the first step. Getting it to work Quickly is the second step. And having enough marbles to actually have the next number in place when the change is ready is still part of Step 1.
lmaooo
There's no way he can't crank it up to go faster. Probably just wanting to make it accurate and reliable first. It's clear to see the marble lift is just going really slow. He can just crank the speed up on it and probably get it at least going twice as fast.
As a long time viewer of the Wintergarten Marble Machine I look forward to the many episodes it will take to get this right. Enjoy, Ivan - I intend to.
He could program the clock to only display every third minute.
One of the coolest things I’ve ever seen built on YT
For real!!
I agree
ikr? can't wait for pt 3
thats true@@davefenton102
Right ❤
You could always try some sort of ball bearing. They would probably be more precise then heat 50% of them up and quench them in oil. Or brass and stainless could be cool if it could handle the weight. This is awesome and just keeps getting better
Great demonstration of rapid prototyping and lots of satisfying components. I love the solenoids and, as always, your enthusiasm! :D
Having sorted many a marble in the past few years, I've found glass marbles always have a fair amount of variance because of the manufacturing process, so whether it's you or the marble supplier, someone will just have to test and sort out the best marbles. For absolute precision, you could consider pairing some super precise balls like White POM bearing balls and dark steel bearing balls. That might compromise your vision though... glass marbles look and sound really nice! The audible difference between 'white' and 'black' balls would be an interesting element, and POM vs steel would introduce the ability to sort by density. Just spitballing haha, I'm excited to see the rest of the journey!
White POM bearing balls would be a great contrast against steel marbles. I’ll definitely check them out. Cheers Devin!!
@@ivanmirandawastaken Theres a pocket83 video about sorting marbles, that may be helpful so you don't have to wait for a marble to jam before you remove it.
I mean if Make Anything is already in the chat, what about ABS no hole beads?
Easier sorting with an inductive sensor.
3:52 when we did lime followers for Uni no one taught us how, so I used these light-detecting resistors. The tip I got much later is: make your LEDs shine at a specific frequency, then put a band filter on your readings from resistors. Pretty sneaky, and AFAIK should work when flashing LEDs at 5Hz
I'm still checking back for part 3 every few days - haven't forgotten! I hope everything is OK. Look forward to seeing what happened next...
Check now 🎉
@@Wookie632 Woo hoo! Thanks.
If you separate the marbles upstream, you can then speed up the row constructions a lot. As marbles are discharged from the clock, separate them in black and white trays, with the black at a lower level. As the conveyor goes up, it will fill with all black, then you eject some marbles to create voids and move the row to be level with the 2nd tray, where white marbles fill only the voids. This way you'll end up with complete rows every time.
Nice idea. Only issue I see is it assumes all voids should be white, but maybe the blacks didn't fill a whole row for some reason.
Then the elevator needs to switch functions: First run it and punch out all marbles, either in the white or black tray. Then swap to the function as you describe it. After displaying the next time, switch again to the filling function. Otherwise, filling the white tray will be filling all the rows with black marbles.
@@paul8731In the design every row must be filled for the marble based display to work. You need a white or black marble in every pixel.
Nice idea but the separation process would be more error prone than building rows as he is. Time isn’t an issue at this point, so it’s adding complexity for no benefit. Accuracy is the key here. Classic engineering.
@@genehenson8851 Don't get me wrong, the clock is beautiful and I love the ingenuity on the engineering side, but right now it is non-deterministic. I know it's kind of pedantic, but getting the right configuration for each minute could take way more than one minute, as it depends on the probability of black and white balls falling in the right spots in the first place. You can see this partially happening as some rows are lagging behind the others and in the earlier designs it was more obvious as rejected balls fell closer to the conveyor. Of course as paul8731 highlighted, preselection is not a perfect cure, as it requires that a row is filled without voids (and that adds even more complexity).
Amazing. Too late now, but you could have the display using vertical columns instead of rows, this would allow a single digit to be changed instead of all 4.
Not really too late. Best part about 3d printing.
It would be faster too. But then, there would be a lot more marbles required. A lot more headache especially with inconsistent quality marbles.
Idk, there’s something cool about the entire display flushing at once and the next one being stacked in line.
Unnecessarily efficient and simple.
It would work better, but this is more fun!
You should consider a vertical layout for the marbles, so that you can update every number individually instead of having a collective - that way you can speed up the process to work automatically in real time without pauses or human intervention. It would be really cool!
I was hoping he would just use an efficient algorithm to update the display. Instead of flushing the whole thing every time it would just let out enough marbles from each track to reuse whatever it could.
And then to make updates a bit faster he could add another set of gates on the upper part of the track and release the whole update and flush the bottom marbles at the same time.
Might also help resolve the issue with marbles getting stuck when flushing.
I was thinking the same thing! Or maybe in this configuration have the next minute already loaded in the upper feed so when the minute changes it flushes the old time and gravity feeds the new time and the elevator is already building the next minute. Just has to be able to load the next minute in the upper feed portion within 50 seconds to keep pace.
Можно рассмотреть ещё вариант: разделить на два табло - часы отдельно и минуты отдельно. Цифры означающие часы меняются редко и поэтому корректироваться будут не часто. Для эффектности показания сбрасываются, все шарики съезжают, но часовые шарики делают кольцевой пробег и быстро становятся на свои места.
That was I also thinking
Ever since the last video, I've been wondering what mechanism you will use to select the right marbles for the right track. I would never have come up with this "subtractive" approach where you remove the marbles you don't want. What a cool idea!! 😁👍
This is really cool, but I would've used a slightly different approach. I'd have 2 layers of marbles where they load into the conveyer - the bottom layer black, then punch out holes where you want white marbles, and then an upper layer that loads any open spots with white. That way you don't have to depend on the randomness of getting the right color into the right tracks, which takes an indeterminate amount of time. You can load up the marbles you need in each row in constant time every time. You also need to sort the "discard" marbles back into the lower tray (black) and upper tray (white).
@@gorak9000 I love that idea! Sounds very efficient, and not too difficult!
IR sensor
@@gorak9000100%. Separate all black and white at time of discard, and build what you need from the two color sources. His method sure looks cool though
@@haileycollet4147 His method basically has all the parts needed to do the other way, they just need to be re-organized. Move the color detection to the discard chutes, and he needs a switchable gate to direct discard marbles to the white tray or let them continue down to the black tray. Then he needs to 3d print the upper tray for the white marbles and it should work. Maybe he also needs a way to work with discard marbles one row at a time too. Much easier to do this kind of re-organization for optimization in code where you can just copy/paste lines of code to move them around!
Maybe you could alternate the colors: Minute 1 is white text on black, and Minute 2 is black text on white, etc. That way, the number of marbles should be the same (50% of each) and the likelyhood of finding the right color for the next minute goes up since the previous minute's background is using a lot of marbles of that color.
Or just switch the ratio of the marbles overall?
sounds like an idea that could work. I thought about some way to "mix" the marbles, as some rows out of pure statistical bad luck took way longer to get the nextright marble than others.
Regarding the marble quality, I wonder if appropiatesized plastic balls instead of glas marbles would work better, although the sound would be way different.
The plastic ones would be way more consistent but the weight vs coefficient of friction will be greater and slower in the channels and may even get stuck easier
Good idea. It would also add (another) aspect of visual interest.
@@erikpip3024 Yeah, good point, the weight might indeed be an issue. Although the difference is less significant than i thought. Glas is about 2.5 g/cm³ while hard PVC is about 1.4 g/cm³. Other plastics like ABS are even lighter, around 1 g/cm³, others may be heavier (but way to soft in comparison).
To be honest I somewhat expected glas to be multiple times as dense (and thus heavier). But in conjunction with any plastic being "softer" as glas marbles, friction indeed may add up on top.
You'd do well using a barcode optic to detect ball color. They are more resistant to ambient light that affects discrete LED sensors. The only way around that is to clock the LED optic so you know whether it is reading room light or reflecting off of a ball
This is one of the coolest 3D Printing projects I’ve ever seen!
That means a lot coming from you 👊👊👊
serious respect for putting out a part 2 so fast, I can see that you put a lot of work into this.
skyblock pfp?
@@blazeyreal7777 I think maybe? Made it when i was 8 and I'm 22 now, so i dont fully remember lol
@@NateMiner8 Definitly not, the server im thinking of started in 2020
@@blazeyreal7777 then who knows, mysterious minecraft pfp, i had a whole intro with it too haha
I like that you talk about issues you are having with the clock design right away, as soon as they happen. It's interesting as you get to immediately see the trouble shooting steps you took to try and over come it.
I love this. When it's done, please live stream the clock running for an entire day.
this!!
great idea
No, forever.
24/7 live stream lol
I really wasn’t expecting a part 2 this quick. It’s crazy how you figured out most of the kinks in just two videos
Especially nailing the clock-display code seemingly very fast!
fr thought it was gonna be 2 months before a new vid so hype that it’s so soon bro rlly committed
Lol
judging by the number of videos , that's very illogical
@@guotyr2502 i’m new here and don’t check every new channels upload frequency
I truly admire the skills, the abilities, to complete this project. Personally, I can see practical usefulness in several design elements. The reality is, that I live in a world where the sheer number of real-world challenges and desperate needs, prevents me from embarking on a project like this one. As Mike Patey (see his YT channel) says, 'Machine it, Draft it, Build it Test it, Break it. Every time something new gets built, the entire world advances'. My regards to you, Ivan.
This is awesome. Great to see your design process and how seemingly quickly you iterate
*@Ivan Miranda* - Improvements:
8:24 If you reject the marbles to the left, so they get into the very back of the queue, you would get even more circulation of marbles.
- -:- - It would also be useful with some kind of "buffer system", so you can store a rejected color & then swap it in later when needed, it could make a LOT more elevator utilization.
10:22 You could copy the lower straight piece into the upper piece, and build the next time in the upper part, to remove the pressure from the marbles in the bend.
Yeah I don't know why it has to roll down that long straight away to get to the bend I'm not sure why he doesn't just set the elevator on the left side and have it drop down into where the bend is... I guess it's more for show to watch them roll all the way over
I had the exact same idea.
BUT, i just made another comment saying that if i was to have made this, i would have build it differently with less marbles.
All you need is a track which holds the combinations needed to form a side of a digit. Split tracks.
So for every number you have a tracklayout that can build up the number.
And the marbles are stored in a tube or something like that.
And so every number of the clock, 4 in total basically, you have a different track.
All you need to do then is have controllers for those tracks that put a new set of marbles in it when the number needs to change.
You only need to have 0 to 9, so not that many marbles per track.
Sorry if it does not make sense. In my head it does.
The only problem i have is: I can draw it but it still would not make sense without the electronics drawn in and the circuits.
O, and the second problem is: I have no time to build it myself.
But the idea i got is almost like the flap down clocks, where a card flips.
@@kmktruthserum9328
PS: btw, the long ramp does have a purpose. I just thought of it.
He needs to be able to change from 23:01 to 23:02 in 1 minute time.
That is not going to happen if he does not prepare it in that minute, to instantly change over basically.
Therefor the longer ramp. (or so i think. I litterally came up with it just now but it does make sense. Otherwise he can not change from 23:01 to 23:02.
Because that requires time to set up.
the longer ramp does have a purpose, it is to set up the next minute.
How else can he set it up?
It has to change from 23:01 to 23:02 in an instant.
That can not take more than a minute.
Otherwise it wont run ontime.
So it has to be prepared.
So with this setup he needs to have a long ramp to set up the next minute.
Heck, the ramp may even be too short to keep the clock running ontime.
Let us see how he will manage to keep it running ontime.
@@kmktruthserum9328
I thought the design was meant to also buffer the next time
@@kmktruthserum9328 It's so it can queue the next minute up or else you'll have to wait and sort the next minute on the elevator itself, which is too much.
I like that when it finishes to display 23:01, it's already 23:04 😆
Great project! Can't wait for pt 3
That's exactly what I was thinking, I was looking for a comment that thought like me 😊
this is so akward
Just set the clock 3 minutes ahead. Next question.
To keep the marbles from getting stuck in the curve, you should have the next time queued up on the straight section on top, not on the curve. The gravity will work better.
something that might help with the speed at which the numbers are constructed is switching the background and foreground numbers each time(assuming about a 50/50 split)
Good idea. This will also reduce the total number of marbles required to fit in the holding tray.
Would adjusting the ratio of white to black marbles also accomplish something similar?
@@SVSQZ yes, but at the risk of having waaaay too many blacks and not being able to populate a white 'pixel' if they can't make it to the front. If the mix is roughly 50:50, and alternating front and back colors it would kinda solve a lot of issues as the majority of the next color you need would be the most available to move up the chain for the next set.
@@SVSQZ Maybe but wouldn't reducing the number of white marbles also result in the collection thing taking longer to pick up a white marble in the correct lane?
@@SVSQZ Hrm, if p1 is the probability the display needs a white marble, and p2 is the probability a marble in the tray is white, then the chance of a correct marble is p1*p2 (white) + (1-p1)*(1-p2) (black). But that's unhelpful to optimize because the highest chance of a correct marble is achieved by having 100% of the kind that's more commonly in the image. So instead we probably want to figure out the average number of marbles pulled before getting a correct one. Luckily this is a geometric distribution and fairly simple - when a white marble is needed it's 1/p2, when a black marble is needed it's 1/(1-p2). so actually we want to minimize f(p2) = p1/p2 + (1-p1)/(1-p2).
Looking at the way the image is constructed, there are the 7 segment display parts (4 marbles each) and the black background (48 marbles per digit) so average about 4 white to 15 black marbles, so p1=0.2 approximately. Putting this in the function gets p2 ~= 33%
There is a catch here though, that's only the marbles in the tray, there are also marbles in the displayed time and the number about to be displayed. So really you want about 80% of the marbles to be in that 20% white proportion, the remainder in the 33% proportion, making it overall about 22.7% white marbles.
If you alternate displaying white and black numbers this gets more complicated of course, since the marbles in the in-tray are the opposite distribution to the displayed number.
Love the work you are doing. Keep it up!
I just don't understand one thing. You could install infrared couples a row below solenoids and only then take a reading when the lift stops. I think it will work as well.
Hmmmmmmm
He would have thought of that
This would make for a really cool interactive art piece! 🎨
Imagine a museum with a computer containing something like a 200x30 pixel grid where you can draw in white with a black background, and on submit, it can "print" your picture.
Every other image would be Kilroy, rude fingers or body bits, but still, yes.
@@tactileslut lol i love this comment
The Precision Plastic Ball people based in the UK are my usual go-to for balls/ marbles.
The Phenolic resin ones might be a good shout for this project as they have a pretty consistent diameter.
Looks cool, can't wait for part #3! :)
Very impressive problem solving, and determination. And everything looked so neat in your build. Two thumbs up!
This is sweet!
Though I had a thought: If you had the time on the right of the elevator instead of the left, you could then fill the minutes in last. This could be helpful if you had a way to clear the minutes without clearing the hours (perhaps where the : is?) Then only the minutes get reset each minute, which is faster. But then every hour you'd get a really satisfying sound of all the marbles returning.
Edit:
Just thought this would be easier perhaps, if the numbers are assembled vertically (in columns) instead of horizontally (in rows). This could also solve your issue with marbles getting stuck.
falling marbles isn't as satisfying as the visual and auditory effect of rolling marbles
How would you remove the minute marbles without removing the hour marbles even if the time was on the right?
@@TheArachnoBot maybe another array of solenoids?
@@pravus9769 You mean an array behind the minutes on the board? That would ruin the entire charm of the clock. Plus that's a LOT of solenoids.
Maybe by having the time display as a shallow ‘V’ with the ‘:’ at the bottom, so the hours are put in from the left and the minutes come in their own track from the left. Both of them empty inwards toward the ‘:’ so you can control either side individually
Cant lie...have ABSOLUTELY no idea how this series was recommended to me, but I'm hooked for more...subbed and looking forward to pt. 3
Ditto
I know how for me. I was watching the latest Wintergatan Wednesday episodes of his Marble Machine design and prototype and this was in the sidebar.
@@billkeithchannelwell atleast you was watching something that related to marbles, i was a qomen playing with baby jaguar 😂 , so i have no idea why im here lol
But i liked the vid and definitely subbed
The final form of this should be placed in a museum or a public place. Its great
I really enjoyed your video! Some thoughts: your clock "refresh rate" is bound to the randomness (mainly) of a white marble going to the right place and the speed of the lift. I would suggest moving the actuators up, use them to short the marbles in two "buffer baskets" (white/black) and from there rapidly feed your "dispaly". (Actuators to each color letting marbles drop in place). Again, super nice project and great video!
Yeah, i feel this random approach will just be too slow to have the next number ready in 60 seconds. When he started talking about differenciating black and white marbles, i thought he was gonna implement a sorting method.
You should drop the marbles a bit higher and have a ramp going left to move the rejected marbles back to the back row. This way you get a better color mix of marbles and will fill faster.
Love it!
It would require a huge redesign, but if you made the display out of vertical runs instead of horizontal runs the marbles would position themselves much more quickly … possibly even fast enough to display seconds as well as minutes a long way down the design timeline. Great concept, looking forward to the finished article!
This is really cool, can't wait to see the finished product on this!!
I would like to suggest to use the effect of rejected marbles stack up in front of the elevator to create black rows as you actually need much more black marbles than white ones. If you actively dumped a row of either white or black marbles from the rejected section when your system needs those you might get to cut down on your wait time until a certain type shows up at a certain position
I was thinking about this problem too and ypur idea sounds like it could work! What if he has an only black elevator with completely seperate white feed in. Then when white is needed, the appropraite blacks are dumped out and the whites are released to fill the gaps. You should be able to create full clock faces within the elevator already!
I'm thinking two rows of solenoids. The bottom row only rejects black marbles from where white marbles should be. The other row, much higher up the chain, rejects unwanted white marbles. The whites are pushed onto a tray that ends up filled with only whites. The tray leads back to the elevator, slightly lower, between the two solenoid rows.
If the fill rate at the bottom is perfect for the black marbles (i.e. no empty slots), and there are enough white marbles on top to fill every gap, you might not even need controllable gates at the white tray. The whites would just fill all gaps left by the solenoids kicking out black marbles where they're not wanted. The result could be really efficient.
If the fill rate at the bottom isn't perfect, the whites will fill too many holes, but they could still be detected and re-rejected by the white solenoid row above the tray.
@@arantala unfortunately, if the whites are not sorted at the bottom and brought up to the white tray, i think there would be the chance of all the whites being in the 'system' and the tray getting empty.. the whites could het stuck in the black mass at the bottom and never come up again. Otherwise a very good idea! (Also if i am wrong, please tell me haha)
Ok so now I want to see more marbles and the date on top with time below.
Just incredible glad I clicked on this.
You should consider putting a mixer of some sort between the ball return and the lift... this would prevent a build-up of 'rejected' pieces and make the very slow marble picking time at least more predictable!
Here’s an idea… alternate the background from black to white for AM and PM. Black Digits for the AM and White Digits for the PM. Not saying there’s anything wrong with the 24hr format, there’s just potential to take it an extra step further with the 12hr format. Just Imagine the transition of a white background to black from 11:59 am to 12:00 pm.
That would require having double the marbles in the system and a great many more rejects.
At the moment say 90% of the display is black so 10% of the marbles need to be white.
Put 20% white marbles in the system and on any channel you'll get black 8/10 times and white the other two you'll fill up the blacks quicker and have whites available when you need them.
50/50 the rate you can fill the blacks is reduced by 30%
Yes, complexity will be incresed. But 5his is very cool idea.
@@zyebormAdding more marbles would be the easy part. The problem is he is not sorting the marbles efficiently enough. If he could separate the marbles before they get onto the elevator, or hold the rejects in a buffer for later, he could easily switch between black and white backgrounds.
It would be very cool!!
Genius, sorting colors and forming the rows on the conveyor. Great mechanical engineering and coding. 👍👍
Epic build! Just one little thing, how about building the display with marbles that come down vertically. That way you can purge each digit individually?
That will also help alleviate some of the feed issues, going more vertical and less horizontal zig zag
This is awesome! May I suggest using two separate feeders, one for black marbles (do that first, punch out the spots that should have white marbles), then a second one that only has white marbles to fill in the spots that should have those (punch out any white marbles that got into spots the black marbles didn't fill before, but should have.) Can't wait to see the finished clock.
This is so creative and so different and so cool!! I love watching the process of producing parts and then making the necessary modifications to fix the issues -- learning and designing on-the-fly! And some of the intentional (or maybe NOT intentional) comedy makes it fun to watch.
This is the coolest passion project I've ever seen; watching your engineering process is fascinating! Can't wait to see the results of the next video!
You should look for wintergatan than
This reminds me of Wintergaten's musical marble machines. So many similar marble machine sounds and visuals.
is wintergaten using marbles or bearings?
Except this one actually works...
@@bertjesklotepino I don't remember exactly off the top of my head, but I think I remember seeing steel balls rather thanks traditional marbles. It would make sense that they would have to be steel balls so they don't break or chip like marbles would.
I just called it a musical marble machine because it sounds better and gives a better mental picture than calling it a musical ball bearing machine.
Well, if i was to build anything like it, i would prefer the steel balls because they are uniform in size and indeed don't break or chip.
And understandable why you called it a musical marble machine because it basically is.
Who cares what the marbles are made out of?
And in the case of this clock, i would for sure have gone with steel balls because a clock needs to be accurate to keep time.
No way you gonna achieve that with marbles.
They roll great sometimes, but sometimes they get stuck. Won't happen if the balls are uniform.
@@davefenton102
@@sgsax yes, true, but by that measure, Babbage's Analytical Engine didn't technically "work" until someone actually built it. However the idea worked and was still pretty awesome, even though it was just an idea on paper for a long time.
I reckon they should collaborate on a musical marble clock
Very cool. I like the relative simplicity of the rejection mechanism! I did notice that it took about two minutes to build up the time though. Looking forward to seeing how you account for that!
Yeah, the randomness of "waiting for a white marble in this exact track" is going to inevitably cause issues. There's always going to be a possibility of never seeing a white. Might be easier to have the sorting done *before* the elevator and put into two separate buckets that are pulled from as needed.
A clock that is right every five minutes is better than one that is right twice a day. In the days of analog clocks, we usually said it was 20 minutes to 3, and not 21 minutes to 3 or two thirty-nine..
This is a really cool idea, and I think it’s well executed, too. Can’t wait for part 3!
Part 23 is where it be at. ;-b
発想を 実現した事に 敬服です😂 自分は今まで机上の空論でしたから 凄いって事しか言えないです😊
I feel like using the marble rejection thing to sort the marbles into colors and then placing them would be a lot more efficient in terms of building the current time before it’s wrong
Agreed, the current system feels like it's dependent on RNG, a healthy filtered system and smart marble reservoir could resolve so many potential problems
While you're right... I think it kind of takes the fun out of the whole thing if it's just sorting them into two separate color pools like that. Really if he can just speed up the whole thing by maybe 20% then it could just work as is. I feel like it could definitely move faster.
Guys dont say I feel like well if your gay anyway women day this feminine habit
@@perrycoffey5410FFS, get a life.
Agreed, but the randomness of how the numbers appear in the clock face is a part of its charm (for me)
you can optimize the slope by using a bradistochromic curve! it would make all the marbles take the same amount of time to fall regardless of placement (i know it’s misspelled)
If you can't remember the word, store bought cycloids are fine.
*brachistochrone ;)
Why not just use Collumns instead of rows?
The marbles would just fall down, when flushed
And it would allow to keep the higher digits therefore leaving more marbles in the tray for the smaller digits to update
Edit:
Maybe if all the balls would roll into those Collumns from right to left, allways filling the rightmost before filling the next one, the leftmost digit could slowly fill within ten hours, while the rightmost digit could update much much faster
I am tempted to implement such an update-algorithm
@@lorenzumhofer566Remember that each lane comes from a sorting mechanism. If you want one lane on each column, that would mean a lot of sorting device.
On the opposite if you want only one lane delivering all the columns from right to left, that sorting lane will be too busy.
@@lorenzumhofer566your solution looks like the one used on the marble machine made by wintergatan ;)
This is amazing! So many problem solved in different ways. Designing, 3D modelling & printing, coding, electronics, general physics, and probably more I can't think of. If it wasn't so loud I would want to build one for my home.
Maybe for version 2 you could have 2 chain lifts. 1 for white, 1 for black. Allow the ones you want up the chain and pop out the rest like you are now. When the time changes and the mix of marbles drops, use something to sort them back into their separate colours to be sent off the the correct hopper.
You design and build the most amazing projects!
Most important words in thei video are "How good is sleeping!"
Amazing work, Ivan. Bravo!
love this project! 😎
Subbed :)
yoooo 3d printing guy
Cheers Adrian!!
Bravo this is such a great video. I don't know if you do all the production yourself or if you have a big team behind you but I love how this whole story was presented. Somehow it made a device as ludicrous as that contraption seem like something anybody could whip up on a weekend!
Feeding the display vertically by columns would save you a lot of trouble :) Amazing determination!
You need to swap it around so that the minutes fill in LAST and can be "dumped" first. (keeping the Hour in place). ALL the digits can be dumped on the hour . and the entire display "recharged." OR you can have 24 columns of marbles in groups of seven that drop down from ABOVE into the display area, then, after a minute, drop down to be recycled up the back to be dropped down again. Hummmm. . . Now YOU have inspired ME! ;)
I think part 4 will be to fix the situation where the clock fails because it cannot get all the marbles in a channel in time. There is, after all, a non-zero chance of no white marbles entering a channel for a minute, and with 1440 minute changes a day, that will happen... My bet is on 08:58 :D
You can prevent this from happening by keeping the marbles within their channel and ensuring there are at least as much marbles of each color that you need for two timeframes (but a few spares will speed up finding the right color.)
Based on the multiple gate setup, I wouldn't be too surprised if the end goal is to queue up multiple minutes in advance, thus smoothing out the variance in slow and fast cycles.
just using the elevator to sort white and black, and having them going into 2 hoppers at the top which feed white and black queues for each row with selection gates that will select for a single white or black marble at a time for each row would be the reliable option, that relieves the elevator speed from affecting the selection speed making the actual time selection real time predictable.
Very nice construction and good to have in the living-room while listening music, watching TV or sleeping for a while 😄
I'm loving this project! I know it's not cheap but maybe you could get metal balls coated somehow if you need better tolerances. I'm not sure about the durability of the coating with the marbles being dropped constantly. I'm starting to think that switching to a vertical layout might make more sense since you could put the conveyor directly behind the clock, it will make filling the channels faster, and overall shrink the footprint significantly. A vertical layout also might make the marble tolerance less important since there will be more falling than rolling.
Hi Ivan, I think that you can eliminate the top and bottom lines because they will always be black. I don't know if that gives you any other benefit apart from saving marbles, but I needed to mention it.
I think anything that can move the black and white marble ratio to 50% will help a lot.
very soothing to sleep with this in the room going. and very portable and very good for an alarm clock
Are glow in the dark marbles a thing? They could be "charged" while in the elevator perhaps.
Yeah they are, that’s a really good idea too! A display visible even in the dark would be awesomeThe only issue is that most glow marbles are plastic, and I think he’s relying on the weight of glass marbles to keep them moving smoothly through the system. He could always use uranium glass marbles, which shine very brightly under UV light. They’re only a *little* radioactive :)
This is not an engineering project, this is ART! I love it - the pt 1, I was like "what a crap". Now I am fascinated. I looks amazing! Sorry for doubting you.
YES! After Pt.1 I thought, "Yeah, but you're going to have to tear out a huge section for the sorter." I thought about this "rejector" idea but didn't think it could be done so simply. This is quite brilliant!
That is a very impressive demonstration of the engineering process and quite relaxing to watch
Cool work! While it works eventually, it feels like a super bruteforce way to achieve this result. May be better to store marble in two separate reservoir based on their color, and dispatch the right marble to the right channel.
If you end up having two elevators in parallel, each with their marble color, you just need something at the top to push either a right or a while one. Or you could have two levels to "load" a while or a black marble into the elevator.
Having a second level for the white marbles seems like a really good idea! If it's above the black ones, the rejection mechanism can stay the same, and gaps created by it will be filled with white marbles automatically. Then the only problem to solve is how to separate the white marbles into their own reservoir.
Or forget about the black marbles altogether.....
Have you tried putting the 'staging' section on the top straight, instead of on the curve? I think the issue you're having is that the marbles stacked on the tight curves don't release as fast, but it might go fine if you stack them on the straight so they're moving when they hit the curves.
This is an absolute beast of engineering and invention! 👏
This is a very interesting build I’m looking forward to seeing the final result. It would be a very expensive alternative but you could try searching for silicon nitride and zirconium oxide ceramic bearings instead of glass marbles. they’re black and white, precisely round and would be less brittle than glass. Good luck 🤞
i had a similar thought, but with aluminium ball bearings, which are definately available, and i think could be anodised to create the tone difference
I know a guy named Martin that is watching this and finding every flaw in it.
3d printing combined with some basic electronics, make one of the coolest things ever, people are so incredibly creative with use of things at their disposal. can't wait to get my 3d printer
racism-powered clock, i approve
This is awesome. Need to make 2.0 where the marbles build the numbers from right to left, so you can create a way to only change the minute digits as the time changes.
Perhaps a top-down fill above each digit would be better. Then they can all be replaced independently.
How would you remove the minute marbles without clearing the entire board?
@@TheArachnoBot Since filling top-down would require almost entirely reworking the marble track anyway, a slightly smaller version of the release gate already being used to reset the display could be put under each digit. When it’s time to change that number, the gate(s) opens and dumps the marbles, quickly closing afterwards to allow the next one to build.
The ingenuity and technical expertise of some people is unbelievable.
Damnnn that's so sick man
Thanks!
So when part 3?
Love it, great example of how to solve problems by testing and improving.
Great content! One recommendation: i would rather hear how you plan to solve a problem and then why it didnt work instead of this was my solution and it didnt work. Even if you film afterwards please frame it as a possible success until its known to the watcher
Great comment! One recommendation: I would rather it have stopped at "Great content!"
@@notahotshot my comment wasn't a criticism, it was a recommendation of improvement to someone who may or may not listen to my comment. How is your perspective helping anyone?
Cool clock... the only problem is by the time it displays the time it's wrong. 😂
That's why he's setting it up to have the next number lined up beforehand.
Nope, you just have to determine how long it takes to build the numbers, and you can time the display appropriately.
@@vindrmasuta as long as it doesn't take 3 minutes to make either display. Don't forget that with his design there is no way to predict how long this will take because of how rejected balls are returned. Even putting them to the very back would be undesirable for some numbers.
How would you determine this? The 'rejected' pieces might be the desired pieces for the next number (and they are somewhat randomly distributed back into the mix) and with the return mechanism there is no way to guarantee it doesn't have to cycle through every waiting marble to get the correct time. Whose to say he doesn't have bad luck and one time it takes 5 minutes to build the display? or 10? @@notahotshot
Yup, cool gimmick unfortunately this will never work as a real clock with his current design. You'd need a faster way to pick marbles, or just speed up the lift A TON.
This is amazing!! I think listening to that well-engineered cacophony for an afternoon would make me crawl out of my skin, and it's brilliant.
Que lindo y que fácil es trabajar con presupuesto ilimitado. Se ahorra muchísimo tiempo en diseño. Porque es cuestión de probar y probar y probar hardware hasta que funciona como uno quiere. Sin tener que darle cuenta de los gastos a nadie.
This is one of the most impressive things I’ve seen on UA-cam
The amount of creativity, ingenuity and technical expertise that went into making this project is commendable. This video is a complete pack of thrill and is a delight to watch. The wise use of different electronic components and the genius way of putting it all together to put up a wholesome show is really worth appreciating.
Keep up the great work! Eagerly waiting to learn something new from the next part.
one of the best thing ive ever seen on UA-cam- Something different and cool
You have to be loved for the creations, edification, construction and portrayal in the video. Thank you very much for this type of technology and entertainment. You should exhibit the machine in Germany at the Mathematicum in Giessen. Thanks so much…
I cant wait to see pt. 3
Fantastic! Great to see life and creativity at work in an otherwise video wasteland.
Wow, I am completely speechless. I'd totally buy this if you started selling them!
Nice work Ivan!!! I need one just for the "noise" to relax
Very cool. Just one of those videos the UA-cam algorithm liked. I watched to the end.
ready for part 3!
We NEED part 3.
An awesome display of engineering skills.
looking forward to see how this turns out. will be super impressive to see it work in real time