Stabilising in a confined space.

[Neirin Jones]

 

I'm not sure I understand that. My idea was a board hard mounted to the camera base with a foam cushion between that and the plywood "bobsled".

 

 

Ok from the 'whiteboard' the plywood and foam bobsled bit is clear, but from the diagram (and I may be getting this all wrong) I have understood M1, M2, M3 and M4 (m4 latter removed) to all be physical points of contact, so supporting arms to the camera base. I was wondering if they should have a foam contact to the the conveyor belt?

 

 

Perhaps I misunderstood.

 

But I am following that the general rule seems to be that multiple layers of foam will deal with varying frequencies of vibration.

 

Thanks to all for helping out, especially the nerds. :)

[Chris Millar]

By pivot point I meant the two edges along the side at board/foam/conveyor interface . Looked at sideways the CoM is above that, to get it in line with it (or below it for poops and giggles) you'd have to swing your masses below the conveyor level ;)

 

But yeah, Neirin he actually means masses, balanced in air:

 

 

Go for it if you like, but I think a simpler approach will be the eventual winner - pretty much what he has drawn minus the weights ;)

[Gregg MacPherson]

@Neirin

Hey Neirin,

 

I didn't think you would missread those drawings. Obviously, bad drawings. I'll just explain the masses and rods. The normal 19mm rod base for a Red is given very long rods. These rods support heavy masses m1, m2. Rig something similar for the masses m3, m4. I didn't intend that the rods flex like the wire walkers pole. maybe the rods will not be stiff enough.

 

The safest thing if one doesn't have an engineer there is to try the simple plank and foam. The only advantage I see from the curved plank (bobsled) is that the bobsled shape can conform to the conveyor with a layer of foam to maximise the contact area. This is a default approach given that I don't know the shape of the conveyor and where the little disturbances are strongest. You could make a sketch showing the rollers under the conveyor, or take a photo. I had assumed that the conveyor was cambered with rollers set at an angle, hence curved section of bobslet, but maybe it's flat.

 

Are there places on the conveyor that run smooth? What is the sectional shape of the conveyor?

 

Brians idea of using a succession of foam densities to damp different frequencies, I hadn't thought of that before. Sofar I hadn't thought much about the foam damping. I was thinking of heavy inertial damping for small angular disturbances, because I think angular displacements are the most noticable on screen.

 

If you do play with the mass rig the tube or solid rod needs to be very stiff. Stainless from the industrial tube supply shop may be ok. Or 4130 steel from the aircraft supply shop. The foam between the camera and the bobsled should be glued (super 77) in place. You could tether the bobsled to the conveyor with a rope or wire to the rear.

 

You could test this with no risk to the camera if you choose. Or use a cheap cam with the same sensor size.

 

What is the coonveyor speed? Do you have any test footage from trial rigs sofar?

[Gregg MacPherson]

 

By pivot point I meant the two edges along the side at board/foam/conveyor interface . Looked at sideways the CoM is above that.....

Let's assume the CoM is the pivot point. The CoM being 50 or 100mm above the bobsled should be fine.

[Chris Millar]

[Brian Drysdale]

You may wish to use a mass combined with the foam damping, as used in earthquake damping systems in buildings. A steel plate may serve if the camera isn't heavy enough on it's own, you could mount the camera on top of it.

[Chris Millar]

Yip, 'mass spring damper' ;)

 

 

M is your camera system - c and K are your foam - your task is reduce y_dot to zero given the context of input disturbances y_0

[Neirin Jones]

@Gregg

 

 

So, here is the test footage we shot on the recce. Shot with a Nikon D800 DSLR with 24-70mm Zoom @ 24 mm. absolutely no stabilisation was used for this shot, the camera was placed on an empty cardboard box (so adding very little weight) and send down the conveyor.

 

I think the video gives you most of your answers, this stretch of conveyor is typical to the whole machine, with the different surfaces from rollers to belt. Except this section is much more easily accessible, as you turn the corner it will enter a more confined and enclosed space, as originally mentioned 2'x 2'.

 

Hope this makes everything more clear, thanks to everyone who continues to contribute on this.

 

 

Neirin.

[Chris Millar]

Light foam on the bottom but a good thick whack of it, longish board on top then even more foam or a cine saddle on top of that - stabilise the rest in post.

 

If it doesn't work try something else - possibly simpler - it isn't a toughie :)

[Gregg MacPherson]

@Chris

For Neirins benefit.

y=position.

y_dot=velocity.

y_dot dot=acceleration.

y0=any position on the undulating surface, shown here at maximum amplitude.

All those are in the vertical dimension.

 

c=damping coeficient such that change in damping force = c*change in position.

K=spring coeficient such that change in spring force = K*change in position

 

We can take the spring mass damper concept as given, though I think it's unfair to assume that of Neirin.

 

@Anyone

Here are the ideas that I think are key.

 

There can be many rollers under the rig, so if the roller geometry is constant the reaction loads will average out. Uneven roller geometry will produce disturbances thet will need damping. The foam should give some damping. The mass of the camera rig should give inertial damping, absorbing a large number of small disturbances from the conveyor.

 

I'm fairly convinced that small angular disturbances to the camera are the most visible on screen. As the camera rotates about its horizontal axis the FOV moves up and down.

 

A large (long) base board may react the angular distrbances to the camera. For both angular and linear disturbances to the camera one should add mass. I would add that as far from the camera as possible. A heavy system will absorb the small disturbances. Small high frequency loads should not be visible.

 

A plank or big long base board needs to be reasonably stiff to help even out loads onto the foam.

 

If the simple plank idea works the camera can be hard mounted to that. If not then one could try a rig like in my drawing with extended masses, but with a flat bobsled.

 

 

@Neirin.

Can you get in there again to experiment? If you use the cheap camera you should try to recreate the mass distribution that the Red kit will have.

 

Try the plank with foam on the straight conveyor. I would start with quite light open cell foam about 50mm thick and have masses on hand to get very heavy. The foam can get compressed a bit ok. Take photos, include a pic of the foam compression on the rollers.

 

If that works go to a smaller plank that will navigate round the corner OK and work on that. You may need cuts, different rigs for different shots. I brought up before the idea of an operator handling a tail pole on the rig. You may need something like that on the corners, but who knows.

[Chris Millar]

c=damping coeficient such that change in damping force = c*change in position.

 

damping force = c* (y_ dot - y_0_dot) ;)

Edited July 5, 2014 by Chris Millar

[Gregg MacPherson]

 

damping force = c* (y_ dot - y_0_dot) ;)

OK mofo, got me there. Damping, by common sense, has to be a fn of velocity. While previously writing I eyeballed the picture and saw what looked like spring force equals K times displacement and unthinkingly kept speedwriting re c. I didn't see the dots (true) untill writing right now.

 

This does not constitute an admission that pictures of toppled scooters or inane kittens are genuinely usefull.

[Chris Millar]

Neirin, if it interests you, please go ahead and try to figure out what we're debating, but if not, worry not as we're just having a technical pissing competition - it won't affect your project :)

 

Toppled scooters:

 

The axis of rotation isn't always about the CoM, in a constrained system it can be about the point of contact with the constraining surface >> think dominos

 

Inane kitten:

 

Same idea, it isn't all about the moments and products of inertia (the inertia tensor). A (partially or fully) constrained system - such as we are dealing with - has more considerations than your solution implies.

 

...yet in the same breath (minus the whitespace), those considerations are too complex to be adequately addressed in the context of a forum discussion, as such, I think it's better 'we' just go with step 1 and see what happens. It's very likely that foam and board will work.

 

I threw up all the math and hoohah I guess as I'm aware of the theoretical model vs real world issue attendant with these sorts of discussions. It's fun - sometimes - to go all out with theory with my (our?) smarty pants on, but I don't know how many reports I've written over the years with the disclaimer of 'experimental error' stamped as a permanent footnote on each page. Perhaps I should have been clearer about that.

 

Anyway, since we're talking about derivatives, lets try something interesting:

 

OK, so we have position, say that position is static at 0m until time 0s and from that point on it increases at 1m/s.

You have a plot that is colinear with the abscissa until 0 from which is deviates and drives up with a slope of 1 from then on.

 

Differentiate it.

 

You now have a plot of velocity wrt time.

It is 0m/s until 0s, from which it jumps to 1m/s from then on (a Heaviside step function)

 

Differentiate it.

 

You now have a plot of accel wrt time.

It is 0m/s² until 0s, from which it jumps to infinity m/s² for an infinitesimal duration about t=0s (a Dirac delta)

 

Differentiate it.

 

You now have a plot of jerk wrt time.

It is 0m/s³ until 0s, where it ...

 

Where it ... >> what?

 

Keep in mind the system, didn't fail at the 'level of jerk' (nor jounce, snap, crackle, pop or mofo) - it failed at every level as a model of the real world - kinda like a forum discussion can do.

 

You did note all those conduits, racks and what not from the video?

[Gregg MacPherson]

 

......The axis of rotation isn't always about the CoM, in a constrained system it can be about the point of contact with the constraining surface >> think dominos......

As the plank gets longer and the foam gets harder yes the "pivot point" tends toward the foam surface. In what I have suggested it probably would not get that far. But I think that is unimportant. In the drawing I made the camera "pivot point" is close to the CoM. Do you agree? How would my "skooters fall over" so to speak. What will provide this force? Where do you think it will be applied?

Edited July 5, 2014 by Gregg MacPherson

[Chris Millar]

Think about a domino.

 

It pivots about its lower edge - because it is constrained (for the intents and purposes of the discussion) by the ground - it does not pivot about it CoM.

 

Maybe I should have called it it's pivot "line" (ha ha - now I'm putting in qoutes!) , but that doesn't preclude the reality of a pivot point, just not in this case ...

 

Your ideas are perhaps more relevant in space.

 

(what a beach is to a dog, space is to free body diagrams)

 

:blink:

 

 

I'm off for a bike ride :)

[Gregg MacPherson]

 

Think about a domino......

I got the domino idea the first time round. Now what do you think will make my "skooters fall oever" so to to speak. What offers this force and where in the system (the camera rig) do you think it will be applied?

 

Perhaps just explicitly translate the scooter metaphore. Otherwise I'm left guessing. Does the conveyor apply unexpected loads in the X axis. Does the PA bump the camera?...?

Edited July 5, 2014 by Gregg MacPherson

[Gregg MacPherson]

@Chris

I confess I am a little over reactive to some of the message impulses that seem a bit, obscurist rather than revelatory. What about we confine ourselves to the most simple direct advice that is useful to people such as Neirin. Which commonly requires more info about the physicality of the problem, and more info about who's working the problem on site.

 

Good bike ride? I'm just off for a bikeride through Orphan Black, which a friend recommended, but which I find a bit "constrained" in concept development (to recycle a recently used concept here), fitting to expected templates, and a bit given to, WTF, have to call it characiture rather than character. But I suppose I am a bit given to be disparaging once I see some signs that displease me.

 

- Most interesting on westernized internet TV recently....Rectify, a well observed introspective narrative.

[Neirin Jones]

**(obscenity removed)** hell, i think i'll just stick a bit of foam under the camera. Enjoy your pissing contest.

[Chris Millar]

BIke ride, yes very good...

 

Windy place is Williamstown:

 

 

 

Pushed about 60km on legs that haven't done much for a while so I cramped up on the way home - had to sort of push the bike from under me and hobble over and lean on a car while a balcony full of beer drinking 'cool people' looked on feeling a bit awkward in a shall we help him or not kinda way, I politely hobbled two more houses down to put them out of their misery. It's amazing how cramp can just go in the space of a few seconds, making me look like I was faking it after all.

 

And yes, contest is alright to come home to too!

 

Foam good.

[Neirin Jones]

Foam good. :)

[Gregg MacPherson]

Hard mount the camera to a board, sit the board on foam, if you still see some vibration on screen, add masses. Masses near the camera will be effective with simple linear (vertical) vibration. Masses to the extremities will be effective is the disturbance is angular.

 

With a few items on hand (board, foam, masses, large monitor, longest lens you will use) the experiments for this could be done in 1/2 hour or an hour.

[Stuart Brereton]

Mount the camera on the longest plank of wood you can find, that still stays out of shot.

[Chris Millar]

What offers this force and where in the system (the camera rig) do you think it will be applied?

 

Does the conveyor apply unexpected loads in the X axis.

 

Well, first we have to define axes - Y-up or Z-up in your world ? Right handed or left handed coords ? - X or 'NOT X & NOT UP' 'along' the conveyor ? What is 'along' the conveyor and so on :wacko: :) (lets not bother)

 

Anyhoo, there are two main forces applied to the rig (without foam), simplified/modelled here as a rigid rectangular prism. It's own mass driven by gravity 'down' - Mg - yes, this acts for all intents and purposes 'though' the CoM. The other force isn't so easy to define in terms of it's point of action, as it is constantly changing. For the most part is integrated over the areas of contact defined by the top of the rollers that provide:

 

- normal pressure to react against the weight and any other resultant accelerations and/or impacts (coefficient of restitution anyone :( )

- friction both fore and aft (to drive the unit forward via the torque from a/the driving rollers) - also friction lefty/righty (stopping it sliding off the side)

 

You'll no doubt know that the friction forces are a function of the normal force and the contact area (again, a pressure).

 

Keep in mind this is a simplified model. We know that the camera jumps about, how could a camera jump about with such a ideal system? Because the real world is all fudged up, bendy, out of line, non-linear etc. There are three main (simplified) cases to look consider (see orange in 2nd image).

 

These points (actually more like planes, but even that is again, simplified) are:

 

- where the forces will be applied.

- where the system will pivot about.

- what I call the pivot points.

- are not at the CoM.

 

I guess I left out the case where the camera flies up and has no point of contact.

 

Yes (some of) these forces will create a rotational moment about the CoM, but they are coupled with the CoM making moments about them. They are the constrained features here (via friction).

 

 

 

 

Any further steps toward reality get much more mathy... Then throw in some foam then it's time for computational mechanics.

 

In the meantime, have you had a crack at differentiating a Dirac delta?

 

You might note that a Dirac delta doesn't make sense in the first place - and if that doesn't sense then neither does having something at rest start moving.

[Gregg MacPherson]

 

Well, first we have to define axes - Y-up or Z-up in your world ? Right handed or left handed coords ? - X or 'NOT X & NOT UP' 'along' the conveyor ? What is 'along' the conveyor and so on .....

Do we have to go through this all point by point?

 

The axes in the pic you uploaded sufficed sofar, or did you change your mind?

 

Somehow the key ideas that any good problem solving engineer would imediately think about are being ignored or obscured in all this. I think this was wrongly identified as a pissing contest.

 

Do you agree or dissagree that what I called "inertial damping" will mask small high frequency vertical disturbances from the comveyor?

 

Do you agree or disagree that angular disturbances (about the horizontal axis orthogonal to x, paralel to the roller spindles) are the primary concern re vertical instabilty on screen. I'm reasonably convinced of this, but all the solutions I've thought of will also work if we have simple linear (non angular) disturbances of the camera.

[Chris Millar]

Do we have to go through this all point by point?

 

"(lets not bother)"