[♪ Unreleased/Tight ♪] How tight music can the Marble Machine X actually play? In this episode we're going to find out! To make the Marble Machine X play tight music, we can calibrate it in many ways. And my goal with this video is for you to understand exactly how. Before when I made timing tests, I always had to crank the machine manually, but now I've installed this electric motor which will turn the machine by itself at a constant torque. This will give us the conditions we need for accurate timing tests! [♪ Wintergatan: Proof of Concept ♪] First I want to check if the Marble Machine X is playing tight in relationship to time itself. To do that I'm setting the hi-hat machine to play one tick for every revolution of the crankshaft. [metronome/hi-hat machine playing] I'm recording the hi-hat into the computer for a couple of minutes, so we can check if the timing is drifting over longer periods. [Martin records the metronome/hi-hat] This is a click from Logic and here comes the hi-hat from the Marble Machine X. Here we go! [Recorded hi-hat sounds] What? It keeps the tempo exactly like one... This is one and a half minutes later... Whaaat?! Let's listen one minute in here... Still playing tight! This is crazy good news, that it's so tight over such a long time period. This means, that the motor is keeping an even torque and strong enough to keep very very accurate tempo! Next I want to calibrate the timing between the hi-hat machine and the left channel of the kick drum. The kick drum is played by two independent marble releases and for this test we're only going to drop marbles from the left release. I'm programming a simple test pattern on channel 24, which corresponds to the left marble release of the kick drum. Pause... (Martin programming...) Hit... Pause... Pause... Pause.. Hit... So, now we should have: [Martin recording the kick drum] Oh, wow! I heard a programming error: The last kick-drum-beat was late! That was... My magnet was in the wrong hole. The marble machine does not play wrong! [Recorded kick drum] Here you can see, I programmed it too late. You can also see that the kick is a tiny bit after the hi-hat, but we can just check the consistency by aligning the first two strokes, and then... lets jump to the next hit! Not bad at all! Look at that! This little consistency test that I just did here is maybe the most important test for today. I'm very happy to see that the marble drops is tight with itself, so to speak: It uses the same amount of time to drop each marble. Now when we got that very hopeful result, we can move on to try to align the kick drum with the hi-hat. So I'm gonna turn this... 180 degrees... This one is pulled back and going to play later. [Martin recording the kick drum] The kick drum is still later than the hi-hat. Orange is the first take. The kickdrum has moved to the left Let's just keep on nudging the kick drum to the left. On these discs I can make a control-line Then I can open a bolt and I can shift the entire rhythm machine backwards. [Martin recording the kick drum] That was pretty tight! [Recorded kick drum] Hahahaha! You can see that the kick drum is moving closer to the left for each test. What a difference this motor has made for the entire Marble Machine X project! Maybe the Marble Machine X can become the tightest mechanical music instrument like ever made! Probably not true: The electromagnetic drum beater from Polyend probably beating me... [Electomagnetic drum beater from polyend] I noticed a pattern in all the tests: The first one was always a little bit more late than the others. Well, I'm guessing that the first one is so close to this edge, and let's say that I didn't really get the edge down properly, this would lean back and play later. So what I'm gonna do is I'm gonna shift the whole program over... here, so we get away from this edge. We can eliminate this issue when we're using the stud welding programming plates from HBS. That's for the world tour and for the producing of the record. I want this system to work when I'm composing. Shifting the program away from the edge did help: The four hits were now consistent and I nailed the timing for kick drum test E. [recorded the kick drum sound] Ah! The programming sections and their internal timing. I welcome you deeper down into this rabbit hole, let's go surgical! So, if you remember the dragon slayer video I did, I was fighting with not only advanced CNC machining tool-paths, but especially the internal timing between the four quarter sections. So I'm very happy to go back and revisit this topic, stab the corpse of the dragon a little bit with my new electric motor weapon. So far, we only tested the first quarter of the programming wheel. I'm gonna program the same kick drum pattern on all four quaters and I'm gonna put a snare drum so I know which beats comes from which quarter. I suspect, that the kick hits from the second quarter will be slightly shifted in some position and the computer will be able to tell us how we can adjust for that. Never mind the snare timing. That's for later! [Rythm/hi-hat machine playing] [Kick drum and Snare playing as well] [♪ Unreleased: My Dream Snare Sound ♪] The different number of snare drum hits makes it very obvious in the waveform which programming section is which. So I could divide the sections up, color code them and then measure all the latencies, and enter the values into my Dragonslayer chapter 2 spreadsheet. I tried to measure a lot of different hits and then I calculated the average of all hits to know exactly how much I should shift a programming section back and forth. To adjust the position of a programming section I can change the number of shim washers under this touching bolt. The touching bolt is resting against a touching plate on the neighboring programming section. The shim washers are 0.1 millimeter thin so in theory I should be able to slide the programming sections in 0.1 millimeter increments. I ended up doing five tests taking all the programming sections off and altering the shim washers, before I got the averages to very close to zero. From here the only improvement possible is to move to the stud welded programming plates from HBS, as they won't have this issue at all. HBS has a very nice video on their youtube channel showing them making stud welding tests for the Marble Machine X. This is going to be really really cool to see full size. So I've been sitting here, recording and entering the measurements in the spreadsheet almost the whole day. Sun is gone... I talked a lot about delayed gratification in this project and... To me it's becoming more and more the theme of this whole project. I think I'm making a masterpiece. Actually... Right now when I'm seeing the numbers in the spreadsheet, I think I am making a masterpiece. Actually... What's that? 13... 13 is a good number! Love it! Like in 2019, I wasn't able to focus like this on the grind. I was somewhere else, probably the project was too painful at the time. I couldn't really go deep, and now I'm just back in the midst of it, and it's absolutely wonderful! This will result in a tight drum beat by the end of this video, if you're having problem with the delayed gratification, at this very moment. Basically what I've achieved is that the blue and the yellow, brown and red have the same timing of the kick drum towards the hi-hat. Here's blue! [Recorded mix] Ignore the snare! For all the tests so far I had to manually set the tempo of the motor with the crank, but I got curious to see if we could actually program a defined bpm. For example 120 bpm. I called Alex and voila we can! He taught me how. We can now control the motor with the computer: The usb cable (is) going from the computer into this controller. So the motor is built in seven steps and the gear ratio from the motor to the pull is seven. So we have to take 120, which is my targeted bpm for the music, times 49, seven times seven, equals 5880... 5880... And hitting play... And off it goes! 120 bpm lining up the first beat. Computer click on... [Metronome from the computer] [Rythm/hi-hat machine playing] Wooooow! Okay, that is very impressive, but if I now go to one minute later in the recording, over here... This is crazy! The Marble Machine X is tied with the computer over like two minutes recording here. There are 38 marble drop channels on the Marble Machine X and now we can start to calibrate the timing in between them. I'm going to start by comparing the second kick drum channel with the one that we already fixed, so I'm adding some extra strokes on the programming wheel. [Martin programming] [Rythm/hi-hat machine, snare and kick drum playing]] The new kick drum is early, we can see it in the test results and we can see it in the waveforms and we can hear it. [Recorded mix] Here you can see the the discrepancy! Let's fix that! One way to adjust the timing is to move this up and down. I can shift it up like that... Moving the marble up had a bigger difference than I thought! Average is 7 compared to average 29. We can also see here, that the kick drum is closer to the hia-hats. [Recorded mix] But it's still early! With this measuring stick I can now see, that the marbles have exactly the same dropping height. So I don't want to put this marble higher, because it's good that they have the same dropping height. Another way to calibrate the marble timing is to shim these black and white registrators with 0.1 millimeter custom laser cut shims. We can just add some more shims underneath, move the registrator backwards and the kick drum will play later. Here's the results after the shims were closer to zero, but it's now too late... An educated guess will be to remove one of the shims we put in and this value should reach closer to zero. Let's try that! I'm pretty proud of this result. I expected to come close to zero by removing one shim, and we did. Here is the first channel that we had always... You can see they're aligned and I'm moving forward... to the second, and you can see how well aligned these are. Calibration of the second kick drum is done. So let's make a beat on the Marble Machine X, I programmed just a simple beat: [Beat from the computer] I programmed the kick drum pattern. Let's see if we got it right! [Marble machine playing the beat from the computer] It's tight! It's not doing too bad! So the marbles are running out. I don't have the funnels capturing the marbles, so I have to manually put the marbles up here every time. So remember to ignore the snare, we didn't calibrate that! [Recording of the marble machine playing the beat from the computer] It's so good! It's so good. This is crazy! Oh, nooo! I pulled my... oh no!!! I pulled the main.. electricity out for the computer. Ahhh! Perhaps I should do something about this electric installation :)... Oh, I'm hearing it sparkling... This is bad... okay... It's perfect! Hearing the kick being this tight is one of my happier moments in the MMX process. Because this means, we can achieve this tightness for all the channels of the Marble Machine X. This is fantastic news! It's computer tight, almost. It's very very very good... I did the same calibrations for the snare drum and a little bit for the marble hi-hat, I burned most of my energy on the programming sections, so it's a little bit rough at this stage, but this is where we are at the moment. Hidden in the midst of the Marble Machine X we have the beautiful time keeping device, the indexable clutch built by machine thinking. This device will also shift the timing between the kick drum and the hi-hat, but this is meant to be used for when changing tempos. So if we play a song in 120 bpm and then go to 80 bpm, we're going to use the indexable clutch to compensate for the latency issues. But to keep things a little bit simpler in this video, I will revisit the indexable clutch later. If you want to know how the indexable clutch is working, you can check out episode 77 where that is explained in detail. The motor for calibration is amazing, the Marble Machine is another beast now! I'm still maybe only halfway there or not even, I can still improve on this, but you can hear that it's coming to life. And when using my programming pins I've seen the measurements, that the same pin have the same... errors. So, now I believe even more in the stud welding system from HBS that Sebastiaan and Marius are working on. So for now I'm going to leave it like this, I'm going to move over to the picard of the bass, and the drum drops and the bass drops and the funnels so I don't have to manually put the marbles in, but that's for coming episodes. Thank you so much for following the build! This is really really promising!