Ahoy folks! I've gotten a lot of questions about rotating turrets recently, so I figured I'd post a proper thread about the topic. The objective for this topic is to discuss both the physical hardware and the user interfaces for various turret rotation mechanisms that I have used. I will cover both Big Gun and Fast Gun applications, and the benefits of each. I will be paying particular attention to the user interfaces, because this (in my experience) is the most important and underappreciated aspect of the entire topic. What's the point of having the perfect rotation mechanism if you get fumble-fingered in the middle of battle and point all your guns in the wrong direction? A proper user interface must be simple, intuitive, and easy to use. You generally have two options: position-control setups, and velocity-control setups. The defining feature of a position-control setup is that the operator specifies the position of the cannon, and the cannon figures out how to get there. This can be controlled with a dial for a wide range of positions, or with a toggle switch to select between two positions. It is precise, but can potentially have issues in the event of collisions or battle damage. In a velocity-control setup, the user specifies direction and speed of rotation, rather than position. The turrets will rotate while the user holds the controls, then stops the moment the user releases. This method tends to be simpler, but less precise. It can be difficult to aim the cannons if your ship is too far from shore, and you cannot see where the cannons are pointing. EXAMPLES: Lets start with a basic Big Gun application. The battleship Gascoigne is equipped with a very nice Indiana-style cannon with a continuous-rotation servo, and an elastic band to connect servo to turret. This is the most common rotation setup in Big Gun cannons. View: https://youtu.be/POMBl4basD0 Like most Big Gun setups, Gascoigne's cannons are mounted to the hull rather than the deck. Perhaps the most interesting thing to note is the user interface. The cannons are completely controlled with the right stick. Push the stick to the left, and the cannons rotate to port. Push the stick right, and the cannons rotate to starboard. Push the stick a little, and the cannons move slowly. Push the stick all the way, and the cannons move as fast as they can. The aft cannons have a servo reverser, so all of the cannons will converge to the same side for a full broadside. To fire the front gun, you push the stick forward, and pushing the stick back fires the stern gun. Most ships have physical endstops that stop the cannons when they reach the broadside position. The user interface is intuitive and simple, and given the relatively slow pace of Big Gun combat, it is very easy to learn and use. disclaimer: this particular ship has not been operated in about 10 years, and is in need of a refit. The turret sometimes gets hung up, and the blast shields, turret depression, and cable control are all missing, along with other needed work. Next up is a very different sort of Big Gun setup. This example is from SMS Prinz Eugen, a Viribus Unitis class battleship. It uses a very similar basic hardware setup, but the user interface is very different. Instead of a continuous-rotation servo, it has a 6-turn sail winch servo and uses a dial for precise position control of the turret. Take a close look at the ship's original radio. The dial on the top right has a collar with an extended setscrew in it. When I place my thumb on the setscrew, I can very easily feel the exact position of the dial. Using end-point-adjust, the travel of the turret has been precisely matched to the travel of the dial. So whichever direction the dial points, the turret will also aim. The right stick has no effect on rotation, only firing. Please note how quickly this setup can move the turret. In common velocity-control setups, such a high rotation speed would be too fast for precise control, unless you go to the endstops each time. Unlike the Gascoigne, this example gives me an excellent over-the-shoulder shot, but also only works because I'm only rotating the A turret. B turret is a dummy, and C and D turrets are fixed directly astern. The dial on the left is another example of how to improve tactile feedback for a dial, by replacing the short setscrew holding the dial in place with a longer allen-head screw. The result is the same: the dial is easier to use and provides good tactile feedback to my thumb. That one was used for the pump ESC. I rarely had a need to throttle my pump, but I had the option so I experimented with it. Unfortunately the original radio was cheap HobbyKing junk and died after several years, so I had to improvise a bit for these demonstrations. Before this ship sees action again, I will need to find another radio with dials in a suitable location. View: https://youtu.be/q8qY0BXGBwo "But Kotori87," you ask, "what happens if you get a little too close to your target, and the cannon gets bumped out of alignment?" The turret is only linked by an industrial rubber band, which slips quite easily. Fortunately, re-aligning the cannon is quite easy. Since this cannon is designed to use the maximum physical range of motion, I can swipe the turret full left and then full right. If it is mis-aligned, it will hit one of the end-stops and slip until the servo is properly aligned with the turret. This swiping motion only takes a couple of seconds, very easy to do if I have any doubts whatsoever about my cannon alignment. View: https://youtu.be/CAdYRFFUwXM and voila, good as new.
Next up lets take a look at a Fast Gun example, with the SMS Schlesien. This little predreadnought has a traversing stern cannon that covers a 30-degree arc. This is perhaps my best example of intuitive user interface. Consider, for a moment, what actions must be taken to bring a stern gun on target. You turn away, and fire as she bears. With that in mind, this setup uses the rudder input for rotation as well. When I turn away from my target, the turret traverses 15 degrees towards my target, bringing the guns to bear just that little bit sooner. I know people are very interested in the specific hardware setup, but I think the ultra-simple user interface is the more significant development. There are no extra switches or dials to fumble, no extra brain-power required to operate. I had to do a lot of design work to accomplish this within the tight confines of a predreadnought, but a larger ship like a Deutschland or other proper CA can duplicate the effect with a separate servo quite easily. With that said, lets take a look at how it works. View: https://youtu.be/9SzNQqzratU OK, now for the actual mechanism: These photos are from an early prototype, but you can clearly see the basic idea. The two cannons (black ABS) are mounted in a turret adapter ring(also black ABS) that converts the basically square profile to an round D shape. You can also clearly see the Millenium Falcon plate, which I will discuss again in a bit. Now we get a much better view of the assembly as installed. The turret assembly sits inside the barbette, carefully sized for that perfect snug-but-not-so-tight-it-cannot-rotate fit. The Millenium Falcon plate serves two roles. The first is to act as a retaining collar, holding the turret assembly down in the barbette while still allowing it to rotate. The second role is to interface with the rotation pin sticking up from the rudder servo. When the deck is screwed down, the rotation pin fits securely between the two blades of the millenium falcon plate. As the servo rotates the rudder one way, the pin moves in the opposite direction, dragging the millenium falcon plate (and the stern cannons) with it. As you can guess, this makes the rudder servo a potential single-point failure, so I use a high-torque metal gear waterproof servo, and thanks to the fantastic design of the Deutschland kit I can swap the servo in just a couple of minutes if something goes wrong. And yes, I did exactly that in the middle of the campaign game at Nats 2019. It is also worth pointing out the other minor adaptations I made to support rotation. The gas lines connect to swivel fittings, so there is no danger of rotation unscrewing a fitting and causing a leak. The area around the gas lines is kept as clear as possible, and the solenoids are given a small amount of room to slide forward and back as the cannons pivot. The overall travel is only about 1/4" or 3/8", but this flexibility does help reduce stress on the rudder servo. As for collisions and such, the interface pin has a small amount of flex, allowing it to absorb impacts due to collisions, bb impacts, or other stress. Lastly, we have a top view of the setup. You can see how the cannons fit in the turret adapter ring. Also noteworthy is the barrel fittings: male threads on the barrels and female threads on the cannons. This is a much more rigid setup than the traditional arrangement, allowing these cannons to operate without a deck-mounted barrel brace. My original cannons (seen in the first pair of photos) had the traditional setup, but they were unable to maintain a consistent range setting. Ultimately I don't think a little predread like the Schlesien really needs a pivoting stern gun. I built this because I could, not because it was necessary. But other ships could definitely benefit. Imagine how much more effective a triple-stern cruiser could be, or a triple-stern battleship like Scharnhorst or North Carolina. What about proper rotating sidemounts? Well ultimately I haven't built one yet. I am sure you can figure out how to fit a sidemount into a similar turret adapter ring, and have it spin freely inside a barbette. You've got plenty of options for rotating it, too. Gears, belt drive, servo arm, etc. The most important thing is to consider is not the mechanism itself, it's the user interface. All of the examples listed here have succeeded because of their ease of use. The mechanisms may be complex, but the user interface must be simplicity itself to use. The same thing applies for any controls, be it rotation, depression, or dual throttles. I cannot count the number of times I've seen people struggling because they have buttons or dials in awkward places, or a stick they push up/down to move something left/right.
Perfect timing!! I started to draw the turret rotating joint this morning! The #1 thing that I was trying to solve NEXT was how to get a servo to rotate 180 degrees in a positional setup. I have been unsuccessfully searching for "6-turn sail winch" for days now!! THANK YOU!!!
Pololu - FEETECH FS90R Micro Continuous Rotation Servo ^Also available on Amazon... 4 for $23 and I could have them tomorrow...
I figure I should post an update to this. I have swapped the Gascoigne's continuous rotation servos for GoBilda 5-turn servos from Servo City, but was unimpressed by the rotation speed. While that rotation speed is great for velocity control, it is inadequate for position control. This is entirely due to the huge diameter of the magazines, though. The GoBilda servos are actually quite fast, it just acts slow due to the ratio between magazine diameter and rotation drum diameter. I'll also drop a warning about the cheap non-brand-name sail winch servo from Hobby King. It was absolutely terrible. All they did was replace the standard potentiometer in a VERY cheap servo with a 6-turn pot, and the result was appallingly bad. It could not point the turret at the same position twice, simply due to bad position tracking. It also took forever to reach a specified position due to its large deadband, so it would get kinda close and then slow to a crawl, spending perhaps 10 seconds to reach its final position. The result was entirely unusable, even for non-combat sailing applications. They sometimes stock other sail winch servos, which I have not tested yet. Are you looking for the cannons themselves, or just the rotation servos?
The question came up on my Mikasa build thread of what an ideal TX setup would look like. So here's a brief discussion of user-interface design. This is based on my experience with a variety of Big Gun and Fast Gun control schemes, along with my experience in combat flight simulators. I've already discussed position control vs velocity control above, so this will assume a position control setup. First thing's first. How can you interact with the TX? Most of your digits are required to hold the TX. So effectively you've got two thumbs and two index fingers for user input. Each of these can do separate tasks simultaneously, ie left thumb steering while right thumb fires. The first user-interface challenge is to identify the major tasks, figure out which ones must be performed simultaneously, and then assign them to the four different fingers. First come the basic ship tasks. Throttle, rudder, and pump(s). Throttle and rudder are on the same stick, so those are continually operated by the left thumb. That wholly consumes your left thumb, now we are down to three inputs. Pump(s) is a simple on/off switch. I like it on the left side, assigned to the index finger. Operating the pump requires only a small time commitment from the index finger, so it is still available for other tasks. Now we get to the important stuff. Fire control. Each set of turrets requires rotation, depression, and firing. By far the most common solution is to directly control the forward turrets, and simply have the aft turrets mirror the forward turrets. However, a computer Director system can be used to take bearing and range information and calculate aiming for each turret group. Either way, you need to provide some form of direction, depression, and firing. Direction is the most important function, so lets give that to the right thumb. Depression needs to be controlled simultaneously with direction, since you at least have to un-depress the cannons when crossing one broadside to the other. I find it convenient on the left index finger, but it could easily go on the right index finger instead. That just leaves firing, which is traditionally on the right side. Thumb or index finger, either works. My ideal setup looks like this: Left thumb, throttle and rudder. Left index, pumps and depression. Right thumb, direction. Right index, firing. Now we need the actual user interface. What controls on the TX, and where they go. The left stick is pretty well figured out, no need to mess with that. You also need some switches and, ideally, a slider for depression. The Taranis does this very well. Its "slider" is very conveniently placed for the left index finger, and all of the switches can be easily accessed and distinguished. A linear potentiometer could be a suitable replacement for a dial on radios that don't have a slider in the right location. The real challenge is the right side of the radio. A simple velocity-control setup can use the right stick as-is, but pretty much any position-control setup replaces the right stick with a 270-degree dial and a pair of buttons. Ideally the dial would have detents every 45 degrees, and a nice, long protrusion so you can easily feel the intended aim point. Both of these are important, as they provide tactile feedback about your aim without taking your eyes off the battle. The detents are especially useful if you've got a computer Director, since you can feel when the aft turret is able to engage. You can then make a conscious choice whether to stop using the aft turret, or turn to keep it in arc. I've seen people try rotary encoders for true 360-degree input, but it never truly satisfied me. Maybe because they never had good tactile feedback, maybe because of the mental gymnastics required to either 1) aim from an imaginary point in the center of the ship, or 2) switch aim points from front turrets to aft once they rotate far enough. Probably the inadequate tactile feedback. The other advantage to a 270-degree dial is that the cheap FlySky radio has two of them already, all you need to do is relocate them. Last is firing. Most firing setups either use a stick (which we've removed), or buttons. Buttons are nice, and can be located just about anywhere. Buttons on the front face are very easy, but would use right thumb for firing. That is less than optimal. Buttons on the back face are better from a user-interface perspective but raise other concerns. Will the buttons get pressed if you set the radio on the ground? Can you feel the difference between front and back buttons? I think two big triangle-shaped buttons should be used on the back of the radio, with a trigger guard to prevent accidental firing. I am very curious, however, about a momentary 3-position switch. A nice big switch that spring-returns to center would provide all the benefits of buttons, bit it would be a single component that eliminates tactile confusion between different buttons. I'd also like to take a moment to discuss Computer Director fire control. This is a possible upgrade to position control that uses a computer algorithm to aim each turret group based on bearing and range to target. Note that I have only seen one, I never got to use it outside of bench-side tests. View: https://youtu.be/094RDEzKGJw As you can see, a computer director allows you to bring more firepower to bear on targets not directly to your 90* broadside. This is pretty significant, if done correctly. It is still vitally important, however, to have a good user interface. The only example I saw used a rotary encoder, which is good for 360 degree rotation but provided poor tactile feedback for aiming. It also required a custom controller with a zigbee bluetooth wireless link for control. I'm not an electronics engineer and have little experience with that level of computer design, so I don't feel comfortable making such setups myself. That particular example had different levels of convergence and depression for different ranges. Setting the different ranges and calibrating range-based depression for each turret is a HUGE amount of work, even before you add wave action, battle damage, and listing. The director mostly provides benefit for long-range fire at ranges of 6ft+, where you're basically at zero depression. Once you get close enough for depression to matter, you're not shooting full broadsides and you're concerned more with angle-of-impact to maximize your underwater penetration. If I were to develop my own computer director, I would only have it bring all turrets parallel when in arc, and wouldn't bother with the complexity of calculating convergence or depression. Question. Is it possible to write my own programs for the FrSky Taranis or, preferably, the FlySky radios? I would love to code a director system directly into the TX.
This is great. I have a couple questions, but I wanted to answer you first… OpenTx radios support limited programming in Lua.
Yes, you can. I had a full converging fire mod to the source code back when the code was still mostly fixed point math on 8 bit processors. Lost the code when Google code closed, sadly. Learned a ton about fixed point math that I am not likely to use with more modern systems.... The first place to look would be the source and particularly their custom math libraries and headers.... Or do it in lua. The nice thing with companion is that if you compiled both, you could simulate the function in companion for testing....
I’ve had the chance to re-read everything; again great info here. On my Deutschland, I’ve written what you are calling a computer director. It’s written in Lua and runs on my OpenTx Taranis. So far I only have the front cannon printed, so the whole thing is not quite obvious yet… but it will be in the coming weeks. I’ll make a video in that thread as I’ve promised. So before I get into my questions for ya’ll; let me write a bit about what I have done right now… and what I learned during yesterdays dog fight. Like you pointed out, the biggest concern I had was my mental load while fighting. I wanted one position control and a single trigger, and for OpenTX to figure out how to point and fire all available cannons. I took a different path than the rotary encoder. I thought the right stick could give me both position and distance. Which you can start to see in my position control video: https://youtube.com/shorts/lawA_oLJYJQ Think of the stick is telling the controller “what is the distance and bearing to the target”. The center of this stick is my ship, stick up is forward, stick right is 90-deg broadside. Once the system knows where the target is, it can figure out which cannons can see the target. I liked the idea of the stick because, as you say, I have tactile feedback about where it is pointed. Lastly, once the controller knows which turrets can see the target, it also know which cannons to fire. Mix that with a “big gun reload” timer; and I don’t need to think about anything but holding down the trigger & pointing the cannon. —- OK so, on to my question. Cannon rotate was always going to be awkward for me to learn. When I was dog fighting with @Huru MorDae, I realized I kept getting mixed up which side he was on. My initial thought is this is a “me” problem & I just need more practice. However, I’ve not tried other designs before. So I’m curious if you think the XY-stick is going to hold me back? Also, curious if you think the rotary encoder (or anything) is superior in some way?
um... where to start? I am not sure there's anything inherently wrong with using the stick as input. I haven't tried it, so I'm not sure how intuitive or effective it is. Do you still have the centering springs on the gimbal? I bet it would be easier with friction sliders instead of centering springs, so you can set it and take your hands off. Especially if that is your input for both direction and range. That said, I personally would prefer to have direction and range on completely separate inputs. That way I can make quick adjustments to range or direction without affecting the other input. If they're both on the gimbal, it would be extremely difficult to adjust one without affecting the other. OK, I took some pictures of different firing scenarios. First up, long range: At long range (6ft+) Director-controlled rotation is useful, but convergence and elevation are not significant. A lot of gunfire takes place at these ranges, since turrets can rotate much faster than ships can move. However, all cannons are pretty much at zero depression. From this range and beyond, your shots are just ventilating the upper hull anyway. There is also little benefit to converging turrets except on truly giant ships like Montana and H39. Ships like Gascoigne, SoDak, KGV, etc. are compact enough that you're getting quite a good density of fire even without convergence. Your shots may skip once or twice on the way over, and the normal rocking of the ship is larger than any adjustments your elevation servos would be doing. Next up is medium range. Ranges of 3-5ft see the most benefit from Director-controlled convergence and elevation. The differing angles between turrets is small enough to have minimal effect on impact angles/penetration, and you're close enough that elevation servo adjustments are more significant than the normal rocking of the ship. You still can't put holes deep in the red, but a well-placed volley will damage the waterline or just below. Concentrate six, eight, or nine-plus rounds in a small area on the waterline, and you can easily cause balsa failure. This is my preferred engagement range for broadside engagements, but few other captains are willing to stay on a parallel course with me at these ranges. They usually either close to short range, or turn away. Lastly we have short range. Ranges of 2ft and less are usually where the truly critical damage occurs. You're close enough to penetrate armor deep below the waterline, and you're getting enough down-angle to actually get your rounds down there without skipping. However, at this distance several issues show up. The first is impact angle with convergence. You're starting to get some pretty bad impact angles if you converge too much. In the first photo, Mikasa's guns would likely bounce off Gascoigne's sides. In the second photo, Gascoigne's guns would definitely bounce off Mikasa's sides. In the third photo, both ships can only effectively shoot with their stern guns due to the excessive convergence. Photos 4 and 5 show alternatives. Photo 4 just has Mikasa's cannons parallel, instead of converged. In this position, I can fire the front gun with maximum depression, then quickly adjust elevation and shoot with my stern gun. This is easy to do with a properly-placed dial or slider for depression control. The 5th photo shows what you'd get without Director control. Not great, not terrible. The other issue that shows up at these ranges is that different turrets will require different amounts of depression. Whether you converge your turrets or just bring them parallel, you still need to set different elevations. I should also point out that while this photo shows two ships in a sustained broadside engagement, most shots at this distance are from passing engagements in opposite directions. You have to set your cannons, and then you only have a second or two to make adjustments as you shoot. Once I learn to code (IF I learn to code), I would prefer a director that only brings all turrets parallel. I wouldn't worry about convergence or elevation control as part of the calculation. You'll probably have great difficulty using a gimbal for input that is accurate with range as well as direction. If it only tells direction, that input can be greatly simplified since you can just jam it in the direction you want without trying to feel for just the right range. Also, I strongly advise against having a single fire button and hard-coded timer. Since we have such a slow rate of fire, it's important that you make each shot count. Remember what I said earlier about shooting the front cannon, then adjusting elevation on the rear cannon before shooting? Can't do that if they all fire at once. Also, details like timing your shot with the rock, impacts between ships, etc. are all considerations at close range. It's a really bad idea to just hold down the fire button and wait for the timer to expire.
This is interesting. At least with the 1-1 dogfighting, I was constantly moving the stick to point at 'where he is now'. At least from that day, I can't see taking my hands off the stick. That said, we'll see what happens after I get a few fights under my belt. This is a good idea, I will try out the system using a three-position switch for depression. - Note: My code calculates depression data, but as you see from my build thread, that's day 2 for Deutschland! I think you've captured the value of a gun director pretty well right here. In the systems with a servo-reverser, my back cannons are worthless until I turn. A gun-director will let front and rear cannons come to bear. This makes a lot of sense to me. From the dog-fight, he was in this range, then he was gone. Worth a shot or two only. Again I'd expect the gun director to show it's value here, because you can send more steel in that moment. Really good points here; I'll have to update my math. Which is a really good point. I'd love to simply add a set of secondaries and be done with it. Oh, if only my hull had more internal space. This surprises me? I understand you to mean that convergence isn't really worth much. Just get holes in the target and be done with it. (I do like that it sounds less complex...) Yeah, now that you mention it, I could feel this during the dog-fight. With depression, it would be a lot more fidgety. OK, I think you've sold me. I'll try your three-position switch; once I start designing for depression. Interesting. I did code it differently that I described it above. - On a single trigger pull, it sends a single round. - If I hold the trigger down, it goes full auto until all cannons have fired. - It does respect the reload timer, so it won't fire a cannon which is still 'reloading'. - It also prioritizes firing the barrel which has the most amount of ammo remaining. I went with this algorithm because I didn't want to install a bunch of accumulators. I figured I could time-slice the barrels in rapid succession, giving the pressure system a chance to recover, and get nearly the same amount of steel in the air. I honestly had expected this design to be inferior to an accumulator based one; so I was hoping my cannons were efficient enough not to need accums. However, based on what you are saying, maybe this is better than I expected. With a 6-second reload time; I am looking forward to being able to fire individual barrels. I hated the idea of feeling helpless while I wait for my damn timer to elapse! Point taken. This won't work in my algorithm either, since I'm not in control of which barrel will fire. I'll take this into account once I add depression.
