Discussion: Issues with 6DOF

Once the self-locking breaks loose from friction the platform falls instead of being lowered.

Some worm gears are better than others where backlash is concerned. I've seen some where the wormshaft itself moves considerably back and forth causing backlash before even counting the backlash in the gearing itself.

My opinion is the best ones are the ones that have a screw in the worm gear housing that butts up to the end of the wormshaft that allows for adjustment so you can get it as tight as possible against its bearings before binding starts occurring.

Lubrication will be somewhat quickly squished out from between the gears so if the problem happens or gets worse after break-in then that could be the issue.

Your description about it breaking loose and falling is very interesting. That sounds like a real possibility.

As for lubrication, my gearboxes are filled with synthetic gear lube. Not sure if that is of interest or not.

I agree about the adjustable worm drives. The problem is the reducers are basically the most expensive part of the whole motion sim. I bought my gearboxes on the basis that others had used them and reported being satisfied, and that they were the least expensive tried and true option for my setup. If I knew this was an issue I'd have looked harder for an alternative. At this point replacing them just is not financially possible.

As I said though, I think that problem is eliminated now on my rig. At the cost of travel, but with VR I think it's fine. In the future when I add a G-seat even more so. So the only real issue at this point is that the softwares don't allow for my odd crank angles. Hexpod took care of that for me in his software, now I'm only waiting for him to fix the "fake kinematics" option and I'll be ready to fly SimTools!

I'm also super, super hyped about upcoming FlyPT Mover. It has all the options and features I ever dreamed of and more!

Deleted post because I'm not sure my conclusions were correct. I'll report back after I do more testing.

Ok confirmed, it wasn't the weight and it does happen in both directions.

Anyway, point is that it seems tuning is still going to be a battle against jerky actuator direction changes.

I only had unsealed gearboxes in mind meaning grease filled. If you have a sealed gearbox lubrication shouldn't ever be an issue as long as its positioned so the gears are always bathed in oil.

WG-262-060-D Those are the gearboxes that are the source of my problems. I have them inverted but I filled them to spec for that type installation.

I didnt read every thing, but a wormgear can be backdriven when there ratio is lower than i50
with higher ratios it cant.
Its got all to do with the tooth angle in side the gearbox, with a small ratio you get a big tooth angle
but from i 50 and above you get a tooth angle below 3 degree and makes it self locking

That is interesting, thanks! I knew there was such a thing as a backdrive worm drive but I always just assumed they had some kind of black magic special design. =)

Mine are 50:1, no way you are going to backdrive them. I can turn the input shaft with my fingers, but I can stand on a 200mm servo arm on the output and it doesnt' move a bit.

Alright friends, I've got an update for you.

First, I tried heXpod software with a proper profile made by Hexpod himself for me custom tailored to my sim dimensions. It worked great for normal flight, but as soon as I did anything fun (aerobatic or dogfight maneuvers) the limits would kick in... Sim stops moving completely for a couple seconds, then starts moving again very suddenly. Doing a loop had this effect 100% of the time, and it happened other times as well so until "fake kinematics" is fixed I'm stuck with using BFF. =(

So the rest of my testing was done with BFF. I would post this on the 'writing a motion software from scratch' thread, but since to my knowledge you guys aren't very familiar with BFF I didn't want to clutter up your thread with my observations that might not be helpful. I'll post here instead.

With the dampers my sim is capable of much smoother motion than before, but I still face some challenges in tuning. In particular, there is still that "bump" when the direction of an actuator changes. It's mostly irritating when multiple actuators change direction at the same time.

Anyway, the results of my tests tonight were surprising. BFF has two filter sliders for each axis except pitch and roll. Why he didn't give a filter on those two I have no clue. Turning up the filters really slows down the apparent latency of the motions, but didn't really help much at all with the bumps.

I haven't had enough time to draw conclusive results, but so far the way it's looking is that I actually get the best results if I don't use the filters at hardly at all! If I don't use the filters (at least not more than a tiny bit) there is no lag caused by them in the cues. So if I keep the filters very very low or off and instead I turn down the P (as in PID) setting it seems to do a better job of smoothing it out with less delay than is caused by the filters.

I'll have to do quite a few hours more testing and tuning before I can say for sure that this is how it worked out, but that's what gave me the best results in the first couple hour session of tuning experiments.

I will post as things progress to let you know what works best for me. I'm hoping I can switch to SimTools soon.

Are you using simtools 100% washout for heave, surge, sway and yaw?
If you stick with 5deg on rotations and 50mm on translations and still going out of your workspace (cuts in motion) it can be that depending on the aircraft (high performance vs low performance) you have to tune the forces with quick acces sliders or directly in Tuning Center.

Nah, just need the fake kinematics mode fixed.

You can’t fix something which was never broken man!

I spend hours trying to explain how the things works.
You may read again my emails

Alright guys, today I did my first tests with FlyPT and WOW. Really really great stuff, I can't wait for the new version! The filters look like they are definitely going to fix what is left of my smoothness issues.

On a different topic I have an issue you guys might have opinions on.

When the platform is near it's target position (small movements needed) each actuator kinda starts and stop repeatedly (and not all at the same time) until it reaches it. Each movement is like a little bump in the road, but I'm usually flying! They are kinda moving like zombie servos, all jerky and awkward. Very noticeable in straight and level flight in calm weather.

I have two limitations possibly causing this, one of which I plan to solve as soon as possible.

That is, my AMC1280 is set up for only 8 bit speed signal. I don't know if it will fix this, but all that is needed is to cut a few traces and solder on a few header pins and I can change it to 12bit speed output. I can't handle that tiny soldering, but my brother might and he is coming to visit very soon. My old AMC was running 12 bit, but I was focused on bigger problems so I can't say for sure whether or not it had fixed this.

The other limit is I'm using the 10 bit analog sensors. Do you think that would contribute to this behavior? I don't want to replace them until they wear out unless they are causing a problem.


Trip

If you're still only using P start adding some I to your PID. That will help an actuator reach the setpoint before stopping.

By using “fake kinematics” feature, while one actuator lacks stroke to accomplish the demanded pose, instead of blocking the last possible pose, you permit the others to move.

Doing so, you are forcing your platform to reach the poses which are impossible to accomplish.

This is inducing false cueing.

It’s important to understand that “fake kinematics” is based ACCORDINGLY to the definition of your geometry.

Once you decide to limit the course of your actuator (due to mechanical or driving issues you’re facing) you’re modifying the definition of your geometry.

By limiting the arms revolution you are limiting your actuator stroke and the variations of its length.

Following the min and max angular values you gave me, basically your actuator overall stroke is about 110mm. That’s not much.

I’ve hard coded a both directions angular limiter on your demand first as an “out of reach” limiter than, on the bit output level which is supposed to work in a similar way as “fake kinematics”.

As mentioned above, considering that you’re changing the definition of your platform you may not expect that the 3D model motion correspond to your hardware motion.

I am aware of the false cues problem, but Dirty agrees with me, as does FlyPT to a lesser extent.

My platform as it sits right now moves beautifully with Ian's BFF 6DOF. I want to get the same results with SimTools but with your software it is impossible. FlyPT looks to be my solution.

I don’t believe you tried the last version with output limitation man.

But I am happy for you, you agree with yourself

If you have too much slack between your motor and gearbox which has the feedback pot attached to it - then that is going to cause instability in the feedback loop. You have masked some of this with the gas shock which keeps a load against the gearbox, but likely it is still there in some instances. Its a hardware issue - No amount of software filters will fix this. I have seen this exact issue on a larger AC power rig in the past which he fixed 'mostly', as the rig became much heavier through the build. It is yet another reason why I went to ball screws, but if you had a encoder attached to the back of the motor shaft any small slack in the gearbox would be irrelevant.

@SeatTime I know what you mean and I've got a little of that but not enough to matter I think. I'm not 100% sure but I think what I'm asking about here is a different issue.

It's not back and forth... it's that the actuator has just a very little way to go to reach the setpoint and instead of slowly but steadily moving the last little way it's like the second hand of a clock. Start, stop, start, stop.... but always the same direction and each time it moves it moves at the same speed following that pattern until it reaches the setpoint. This happens when the setpoint is just a couple degrees from the current arm angle.

@BlazinH I tried adding I to PID and it seemed like it did nothing, but I didn't try it specifically looking at this particular issue so I will do so thanks for the tip.

Update time! I'm making excellent progress and things are looking very promising!

Problem 1: [SOLVED] I think. Experimental .LUA not yet tested.

Problem 2: [SOLVED] I think. Experimental LUA not yet tested.

Problem 3: [SOLVED] by extensive filter options in FlyPT software

Problem 4: [GOOD ENOUGH] by using SimTools tuning center, plus limiting pitch axis travel to the lower value (forward vs. back) so that it is symmetrical, plus having my sim neutral position pitched up a bit.

Problem 5: [PENDING] It seems the latency was being caused mostly by all the filters I was using in BFF. I don't know yet what the latency cost of the FlyPT filters will be but I'm optimistic.

Problem A1: [SOLVED] by hydraulic damper/gas springs on my servo arms plus FlyPT filters to get a nice ramp on direction changes.
----
Other recent developments/current blockers to my getting back in the (virtual) air:

"Notchy" feeling for small motion sim movements looks to be dramatically improved by switching the Thanos AMC1280USB to 12bit ADC from 8bit. This is a prime example of something you would almost certainly never notice in a racing sim but in flight it drove me nuts. =) I am also looking at possible ways to decrease the slowest speed my sim will move via AMC and VFD settings etc. but I think it's acceptable now.

AMC1280USB sensor feedback (10 bit analog ADC partnered with 12 bit analog ADC speed output) is wonky at the low end of the sensor feedback range. Approaching zero it seems to go into an exponential curve. This prevented me from managing a temporary workaround for the motion sim software bug that cropped up tonight. I have messaged Thanos about it, though it shouldn't affect me any longer after FlyPT publishes a bug fix. =)

FlyPT is squashing a bug that (when combined with the Thanos AMC1280USB sensor weirdness) is preventing me from flying just yet, but he has been very responsive to messages and quick with the fixes. Also his new 2.0 software (FlyPT Mover) should have a first Beta available hopefully tomorrow. =D Very exciting stuff.

FlyPT .LUA for DCS needs major updating and a no-Euler-angles revamp as well as "on ground" detection or landing gear telemetry channels. For now I will use SimTools for DCS telemetry with a new experimental .LUA I'm excited to test as soon as possible, but it's always nice to eliminate a software layer when possible so I'm hoping one of our community geniuses will get a new .LUA made for FlyPT DCS World direct mode. =)
----

Other news: @pmvcda @Dirty and @hexpod are still hard at work making incredible improvements to the software side of our hobby. The new version of SimTools was released today (Thanks @yobuddy and others involved!) and in addition to the work from Yobuddy and other SimTools devs it includes some of the fruit of their debates and discussions. Tons of work is being done with filters that are a real game changer, but also they are discussing other major improvements and ideas. FlyPT also has an experiment in the works relative to VR motion compensation as well. No certainty that it will lead to anything useful but only time will tell.

That last one...
Don't expect to much, but I have some ideas.