Neills DIY Gimbal Design

[BIGTACO]

Here is another test video. This is an example of a "typical flight". Nice and slow. I flew around the park right next to our house (near downtown Houston TX). I haven't changed any settings yet on the gimbal or Hoverfly controller from my previous test flight(s), and same crappy camera. I took the raw footage and ran it through Sony vegas stabilizer @ 10% to remove the roll bounce I am getting.

This coming week I plan to begin working on my horizon (roll gain) setting in the HFC and dissassembling the gimbal to add additional members for rigidity. I ordered 4 vibration ball mounts to replace the rubber grommets I am using. http://www.barrycontrols.com/products/product.cfm?cid=7&fid=36 I don't think my grommets are really doing anything and I am sure vibrations are making it down to my gimbal which doesn't help my roll bounce issue.

https://dl.dropbox.com/u/91287103/HEX 4.wmv

 

[jcmonty]

Here is another test video. This is an example of a "typical flight". Nice and slow. I flew around the park right next to our house (near downtown Houston TX). I haven't changed any settings yet on the gimbal or Hoverfly controller from my previous test flight(s), and same crappy camera. I took the raw footage and ran it through Sony vegas stabilizer @ 10% to remove the roll bounce I am getting.

This coming week I plan to begin working on my horizon (roll gain) setting in the HFC and dissassembling the gimbal to add additional members for rigidity. I ordered 4 vibration ball mounts to replace the rubber grommets I am using. http://www.barrycontrols.com/products/product.cfm?cid=7&fid=36 I don't think my grommets are really doing anything and I am sure vibrations are making it down to my gimbal which doesn't help my roll bounce issue.

https://dl.dropbox.com/u/91287103/HEX 4.wmv

Wow, looks great (besides the "crappy" camera ). I think that you tilt is spot on and the roll is 95% there. I did notice the vibes a bit, but get a better camera on there before you chase all of your vibe issues. Especially if the camera you will be using is heavier. How are you vibe isolating at the moment? Are you using the grommets in tension or compression?

Barry mounts are nice, but the ones you linked to are for low frequency applications. If you are concerned about vibes transmitted by the props, that's a much higher frequency than those are designed for. Will they help? Probably, but maybe not as much as mounts designed for hi-frequency.

Just two cents from the peanut gallery lol. Anyways, the results you are showing are top notch. Get rid of the vibes and you will be at the same level as the CS gimbal for a smaller camera IMO

 

[BIGTACO]

Thanks! I am very happy with results so far.

Currently my grommets are in compression. I don't really think they are doing much to de-couple the bird from the gimbal.

I ordered P/N 275-2N ball mounts. For my weight the spec says they are for frequencies of 20 to 30 hz. What freqency range would I be producing from 6 avroto's? Also what type vibration isololation would you suggest? I could apply more to the camera tray design...

Wow, looks great (besides the "crappy" camera ). I think that you tilt is spot on and the roll is 95% there. I did notice the vibes a bit, but get a better camera on there before you chase all of your vibe issues. Especially if the camera you will be using is heavier. How are you vibe isolating at the moment? Are you using the grommets in tension or compression?

Barry mounts are nice, but the ones you linked to are for low frequency applications. If you are concerned about vibes transmitted by the props, that's a much higher frequency than those are designed for. Will they help? Probably, but maybe not as much as mounts designed for hi-frequency.

Just two cents from the peanut gallery lol. Anyways, the results you are showing are top notch. Get rid of the vibes and you will be at the same level as the CS gimbal for a smaller camera IMO

 

[jcmonty]

Thanks! I am very happy with results so far.

Currently my grommets are in compression. I don't really think they are doing much to de-couple the bird from the gimbal.

I ordered P/N 275-2N ball mounts. For my weight the spec says they will reduce frequency of 20 to 30 hz. What freqency range would I be producing from 6 avroto's? Also what type vibration isololation would you suggest? I could apply more to the camera tray design...

Well if you are spinning a 2 bladed prop passing at 5000 RPM that's 5000/60*2 ~160 hz. Now, if you factor in that you have 6 props, it gets a bit more complicated since the pulses are not all at the same time. By pulse I am referring to the prop wash hitting the boom 2 times every rotation since you mentioned that you are pretty well balanced already.

Don't get me wrong, you may find that those barry mounts help. I know that the barry control brand is used successfully on much more serious and expensive payloads with success. It's really a tricky game trying to isolate vibration. Personally, I have found that the softer the dampener the better, but you have to match the load correctly. I currently use these in compression: http://www.shop.aglhobbiesllc.com/Landing-Gear-Camera-Mounts/Soft-4-40-threaded-damper.html. I use one per roughly 4 oz of load. The work well to eliminate jello, not so much for the bigger bumps. You may find that a stiffer barry mount is actually better if your case.

I am just working with a go-pro at the moment (vibration/jello machine). If you have a IS or OIS lens, that also helps a bit. So, you may find that once you up to a better camera, the problem may go away as well.

Again, take everything I say with a grain of salt, since I am really monday morning quarterbacking after you have already designed and built your own stellar gimbal.

 

[BIGTACO]

Thanks for the info. Interesting vibration mounts you found there. I'll log them in my favorite links folder for ref later if needed. I think what I will do is try the barry mounts and wait till I get a real camera. Then see where I stand.

No problem with monday morning quarterbacking... I love the exchange of ideas.

 

[Efliernz]

I have been following your design with interest - as my diy mount is very similar in design, except I have used 18mm square pine. No twisting problems and 5% of the cost! I too have the "roll bounce" jitter problem. I have tried tight toothed belts, 4-9Kg servos, rubber belts (to absorb the bounce) but I just cant get the last shake out of it. I know servos must have steps - and this is a problem for virtually every mount except Zemuse.

I have flown without the roll servo plug connected and it is perfect... a little tilting here and there is easier to look at than roll bouncing...

People seem happy to do the last vibration removal with software... There must be a way!!!

Pete

 

[jes1111]

The Barry mounts are good - jcmonty is a little off, though, in the suggestion that they are "designed for low frequencies". The term "Natural Frequency" (quoted as 10-20Hz for the Ball Mount Series) is actually the Resonant Frequency - the frequency at which, rather than attenuating vibration, resonance will cause it to magnify the vibration. Since we are playing with a quite a range of frequencies (all of which are really classified as "low frequency" in vibration isolation terms) then we actually want an isolator with the lowest possible Natural Frequency (and lowest Transmissibility at that frequency range). Which is precisely why "softer is better".

However, the downside to very soft isolators is the lack of impact protection (we can live with that) and a general lack of lateral control, i.e. the mount wants to swing like a pendulum, pivoting on the isolators themselves. There are a couple of ways around this - the first is to spread the mounting points as wide apart as possible and the second is to choose a type of isolator that exhibits higher lateral/radial stiffness while having low vertical/axial stiffness. One such example would be a silicone gel shape surrounded by a spring - the spring is happy to compress/extend but very resistant to shearing.

I'm not sure if the "bounce" would be due to servo backlash/slop. If it was then the pitch axis should exhibit equal amounts of the same problem. Timing belts have exactly the same backlash problem as involute gears and, of course, they cannot hide the backlash within the servo itself. Some specific timing belt profiles are specially designed for bi-directional positioning applications, but most are only intended for uni-directional power transmission where backlash is never an issue. However, these are hard to find and even harder to hook up to a hobby servo.

Of course, another possible source of roll "bouncing" is the controller itself. The degree of accuracy required to fly the MR is somewhat lower than that required for a gimbal. The gyros in this level of equipment are subject to noise, i.e. they are not that accurate. The gains you have set will affect their performance. Also, they are highly susceptible to vibration - so if you have any vibes reaching the controller then "all bets are off" in terms of accuracy.

 

[BIGTACO]

Thanks for the thorough set of info on vibration isolation.

In regards to the roll bounce, in my case I don't feel servo backlash is the main culprit. It may play a small part but as you mentioned my pitch should be showing the same issue. But, its not. Pitch is near pefect. With that said, to rule it out I do plan to add some felt washers between the roll servo output shaft and my pulley. I read this is what PH has done on some of their offerings. Seems like this will provide just enough friction against the two moving parts to help the issue. Currently it doesn't take much force at all to move the servo gearset from one "stop" to the other now that my servos/pulleys/belts are all broken in. This backlash is very very small, but it is there.

I am taking a process of elimination tactic at this gimbal. Where I see a deficiency I plan to resolve it and then move onto the next. This will not only flush out whats causing this bounce issue (hopefully), but will also work to strengthen the positive aspects on the system.

Another issue I see with my aerial video tool is the craft itself. I am flying this gimbal under a DJI F550. This craft can get the job done fine but its sort of like trying to haul a 25 foot powerboat up an incline with a supercharged Mini Cooper. Its not the best MR for the job. The span is only 550mm so its simply more suseptable to wind gusts, has low rigitiy in its air frame, is heavy for its size, and is generally sort of "twitchy" compared to larger crafts. If I took this same gimbal and stuck it under a 950mm MR the final video result would be improved.

I'm not sure if the "bounce" would be due to servo backlash/slop. If it was then the pitch axis should exhibit equal amounts of the same problem. Timing belts have exactly the same backlash problem as involute gears and, of course, they cannot hide the backlash within the servo itself. Some specific timing belt profiles are specially designed for bi-directional positioning applications, but most are only intended for uni-directional power transmission where backlash is never an issue. However, these are hard to find and even harder to hook up to a hobby servo.

Of course, another possible source of roll "bouncing" is the controller itself. The degree of accuracy required to fly the MR is somewhat lower than that required for a gimbal. The gyros in this level of equipment are subject to noise, i.e. they are not that accurate. The gains you have set will affect their performance. Also, they are highly susceptible to vibration - so if you have any vibes reaching the controller then "all bets are off" in terms of accuracy.

 

[jcmonty]

The Barry mounts are good - jcmonty is a little off, though, in the suggestion that they are "designed for low frequencies". The term "Natural Frequency" (quoted as 10-20Hz for the Ball Mount Series) is actually the Resonant Frequency - the frequency at which, rather than attenuating vibration, resonance will cause it to magnify the vibration. Since we are playing with a quite a range of frequencies (all of which are really classified as "low frequency" in vibration isolation terms) then we actually want an isolator with the lowest possible Natural Frequency (and lowest Transmissibility at that frequency range). Which is precisely why "softer is better".

However, the downside to very soft isolators is the lack of impact protection (we can live with that) and a general lack of lateral control, i.e. the mount wants to swing like a pendulum, pivoting on the isolators themselves. There are a couple of ways around this - the first is to spread the mounting points as wide apart as possible and the second is to choose a type of isolator that exhibits higher lateral/radial stiffness while having low vertical/axial stiffness. One such example would be a silicone gel shape surrounded by a spring - the spring is happy to compress/extend but very resistant to shearing.

I'm not sure if the "bounce" would be due to servo backlash/slop. If it was then the pitch axis should exhibit equal amounts of the same problem. Timing belts have exactly the same backlash problem as involute gears and, of course, they cannot hide the backlash within the servo itself. Some specific timing belt profiles are specially designed for bi-directional positioning applications, but most are only intended for uni-directional power transmission where backlash is never an issue. However, these are hard to find and even harder to hook up to a hobby servo.

Of course, another possible source of roll "bouncing" is the controller itself. The degree of accuracy required to fly the MR is somewhat lower than that required for a gimbal. The gyros in this level of equipment are subject to noise, i.e. they are not that accurate. The gains you have set will affect their performance. Also, they are highly susceptible to vibration - so if you have any vibes reaching the controller then "all bets are off" in terms of accuracy.

thanks for the clarification! That makes much more sense. At least my math with 160hz should be solid lol. Sorry if I steered anyone wrong. And it's only been three years since my vibrations class.. geez :black_eyed:

 

[R_Lefebvre]

I'm not sure if the "bounce" would be due to servo backlash/slop. If it was then the pitch axis should exhibit equal amounts of the same problem.

Well, I've been thinking that maybe the problem with the roll axis is simply that it has much higher rotational inertia. Therefore any servo slop will have a greater effect.

I would be interested in trying to use an RC car spring/damper to preload the roll axis. Just a light spring. It will be constantly pushing against the servo in one direction, but it should eliminate/lessen the amount of backlash in the gears since it will be pushing the gears on one side constantly. And it will also help dampen any movements. You could probably throw one on really quick. Just need the damper, a couple holes, and ball-mounts.

 

[R_Lefebvre]

Here is another test video. This is an example of a "typical flight". Nice and slow. I flew around the park right next to our house (near downtown Houston TX). I haven't changed any settings yet on the gimbal or Hoverfly controller from my previous test flight(s), and same crappy camera. I took the raw footage and ran it through Sony vegas stabilizer @ 10% to remove the roll bounce I am getting.

This coming week I plan to begin working on my horizon (roll gain) setting in the HFC and dissassembling the gimbal to add additional members for rigidity. I ordered 4 vibration ball mounts to replace the rubber grommets I am using. http://www.barrycontrols.com/products/product.cfm?cid=7&fid=36 I don't think my grommets are really doing anything and I am sure vibrations are making it down to my gimbal which doesn't help my roll bounce issue.

https://dl.dropbox.com/u/91287103/HEX 4.wmv

I think this looks great. At least equivalent to the any of the pro-hobby gimbals like PH, etc. I do see a tiny bit of vibration coming through every now and then. I also see a some slight shaking when you are descending. I assume the copter is unstable when descending, this is somewhat normal, though I had hoped that a Hex/Octo would be better than a quad. A better camera might clean this up with OIS.

So you have a few issues to work through, but a great first effort.

 

[BIGTACO]

This is a valid train of thought. I plan on trying to lessen this slop by using a different technique. I plan on placing a felt washer between the servo output shaft and the pulley. I'll select a washer with just enough thickness to create the right amount of friction to resist that axis from easily moving from one side of the backlash to the other under external load, while still being free enough to allow the servo to do its job W/O binding. If this doesn't help I will look into your idea.

Out of curiousity what type spring damper setup are you speaking of. Can you post a pic?

Well, I've been thinking that maybe the problem with the roll axis is simply that it has much higher rotational inertia. Therefore any servo slop will have a greater effect.

I would be interested in trying to use an RC car spring/damper to preload the roll axis. Just a light spring. It will be constantly pushing against the servo in one direction, but it should eliminate/lessen the amount of backlash in the gears since it will be pushing the gears on one side constantly. And it will also help dampen any movements. You could probably throw one on really quick. Just need the damper, a couple holes, and ball-mounts.

 

[R_Lefebvre]

This is a valid train of thought. I plan on trying to lessen this slop by using a different technique. I plan on placing a felt washer between the servo output shaft and the pulley. I'll select a washer with just enough thickness to create the right amount of friction to resist that axis from easily moving from one side of the backlash to the other under external load, while still being free enough to allow the servo to do its job W/O binding. If this doesn't help I will look into your idea.

Out of curiousity what type spring damper setup are you speaking of. Can you post a pic?

I don't like the felt washer thing because it's dry friction, which means stiction. What you need is fluid damping. And alternative approach to the felt washer idea, is using two washers with a super thick grease between them. Sort of like the friction plates in a viscous limited slip differential. The grease I have in mind, I'm not sure the trade name or generic name, but is usually is used for greasing Polyurethane suspension bushings on cars. It's thick, and sticky, like honey. You'd get a form-factor kind of like the felt washer, but no stiction.

Dampers I'm talking about are simply these:

View attachment 6076

 

[BIGTACO]

Hmm very interesting. I am going to digest these as possible options to try out and investigate further.

Thanks!

I don't like the felt washer thing because it's dry friction, which means stiction. What you need is fluid damping. And alternative approach to the felt washer idea, is using two washers with a super thick grease between them. Sort of like the friction plates in a viscous limited slip differential. The grease I have in mind, I'm not sure the trade name or generic name, but is usually is used for greasing Polyurethane suspension bushings on cars. It's thick, and sticky, like honey. You'd get a form-factor kind of like the felt washer, but no stiction.

Dampers I'm talking about are simply these:

View attachment 8048

 

[R_Lefebvre]

Hmm very interesting. I am going to digest these as possible options to try out and investigate further.

Thanks!

Now if you used 2 on the axis, opposing eachother, that might be better. Less force on the servo gears in steady-state, but might still eliminate slop. There's a lot of damping now. Damping rates are of course adjustable on these things, via oil weight and piston hole size.

 

[jes1111]

Axial damping won't eliminate the delay or the thump when reversing direction, but it could stop the camera rotating freely between the limits. Various DennyR designs had RC Car dampers on them - but he removed the spring and used (I believe) a VERY light oil. The alternative mechanism is rotary dampers: http://www.google.co.uk/search?q=ro...0NaW10QW1q4GQAw&ved=0CEgQsAQ&biw=1595&bih=944

 

[BIGTACO]

Here is another test video...

In this video I flew during lunch near my office. Mid video I went up maybe 250 feet or so for some higher altitude testing. I plan to take a process of elimination tactic and apply one change to the gimbal at a time and then video test. Compare to previous video(s), then move on to the next change if I am not satisfied. My Barry Controls ball mounts should be arriving today or tomorrow. Once they arrive I'll install them and do another test video just like this to see any change in raw video quality. I'll repost this video W/O 10% software stabilization + the new video with ball mounts installed in raw format as well for comparison. If the ball mounts don't help enough I'll move onto adding rigidity to my gimbal structure...

http://www.youtube.com/watch?v=PKho12lud5M

 

[Stacky]

The place you are flying looks remarkably similar to where Hoverfly are based.

 

[jes1111]

Looking good

I hadn't considered your mounting to the frame before. Looking back now at the images of the finished unit, I see that your mounting plate is rectangular, i.e. it is deeper front to back than it is side to side. This could have something to do with the observed behaviour. The narrower side to side spacing of the mounting points allows more pendulum-type movement (corresponding to the roll axis) than front to back (corresponding to the pitch axis). I reckon it will be important to have the mounting points equally spaced (i.e. square), as widely spaced as possible (so the suspended mass has minimum mechanical advantage over its mountings) and, of course, centred exactly over the CG point. Note the design of the Zenmuse in this respect - very widely spaced.

How do you intend to apply the Barry mounts? In compression or in tension?

 

[BIGTACO]

Good eye Jes!

My mount pattern is driven off the bottom plate premachined slots of the DJI F550 frame. I chose the widest possible slots on the plate. This resulted in roughly 80mm length x 70mm width mount pattern. Ideally a wider mount pattern would help the issue I agree. In order to pull this off though I would have to devise some type of intermediary plate between the bottom plate of the hex frame and the top mount plate of the gimbal. I opted against this in the beginning for weight / complexity reduction.

After I install the new ball mounts if I still have issues I may look into redesign of the gimbal frame mount situation. I feel a gimbal of this level would do much better under a larger more rigit bird.

I had planned on mounting the ball mounts in tension. See red circle.

View attachment 6136

Looking good

I hadn't considered your mounting to the frame before. Looking back now at the images of the finished unit, I see that your mounting plate is rectangular, i.e. it is deeper front to back than it is side to side. This could have something to do with the observed behaviour. The narrower side to side spacing of the mounting points allows more pendulum-type movement (corresponding to the roll axis) than front to back (corresponding to the pitch axis). I reckon it will be important to have the mounting points equally spaced (i.e. square), as widely spaced as possible (so the suspended mass has minimum mechanical advantage over its mountings) and, of course, centred exactly over the CG point. Note the design of the Zenmuse in this respect - very widely spaced.

How do you intend to apply the Barry mounts? In compression or in tension?