Need engineer for hire

[John Fields]

I'd be willing to pay what it costs

---
But you must have some ceiling, no?

I'm only asking because I'm interested, and if your ceiling is high
enough, that's what I'd use to determine whether the investment of
resources, on my part, to do a paper design of the system in order
to generate a responsible bid would be worthwhile.

 

[Jeff Liebermann]

I'm looking for someone who can take on a project for my company, to
design a very accurate PTZ mount for a small camera. The mount will
need to have a resolution of around 100,000 steps (0.0036 degrees per
step), and high accuracy positioning (repeatably it should go to the
same exact spot when it's told to via serial/network communication
interface). The camera weight will be around 1lb, so it doesn't have
to be too heavy duty. It also needs to maintain a minimum speed of 60
deg/s for movement (ideal is 100-120 deg/s).

Sanity check. Methinks the number of steps is a bit excessive.
Assuming a 30 degree viewing angle (my guess), 0.0036 degrees per step
equals a:
30 / 0.0036 = 8,333 pixel
wide display. Assuming a square image area, that's:
8K * 8K = 70 MegaPixels
I don't think that's very reasonable and well beyond the state of the
art in imaging resolution.

Working backwards, and again assuming a 30 degree viewing angle, a
good B&W 8 MegaPixel imager will be about 3000 pixels across, which
works out to:
30 / 3000 = 0.01 degrees/pixel
which methinks is more reasonable.

Of course, a narrow viewing angle, as in a telescope will work, but I
don't know of any 1 lb telescopes.

Continuing to brainstorming with insufficient information...

Perhaps do the positioning in two stages. Stepper motors,
motor-encoders, or selsyn:
<http://en.wikipedia.org/wiki/Selsyn>
drives to do the coarse positioning. The fine positioning is done
electronically using one of the various image stabilization methods
found in todays digital cameras and telescopes:
<http://en.wikipedia.org/wiki/Image_stabilization>

I recently saw a vehicle mounted camera that could generate perfectly
stabilized images while literally bouncing down a dirt road. (It
could also detect motion and center the image on the motion). Sorry,
but I can't seem to find the manufactory.

Anyone interested in a project like this?

Not me. I'm buried, lazy, obsolete, burned out, etc.

Anyways, thanks guys - looking forward to hearing from some talented
hackers

Methinks this be somewhat more complexicated for the average hacker.

 

[Tim Wescott]

Wow lots of responses.

By low cost I'm talking between $1000 and$2000 per head. With a small
camera I really don't see the need to go heavy duty mounts - that
sagebrush unit is huge and heavy, and probably 10x times what I'm
looking at for cost. I'm wondering if it's using stepper's (big &
heavy).

There is a zoom lens on the cam, so this is just a Pan-Tilt head but
with communications hardware added for the zoom (so it's all on one
interface). There are several cameras we are looking at, all at less
than 2 lbs and fairly small dimensions.

BB - Yes for the most part the answer is in the software but the
higher accuracy the mount is, the less processing needs to take place
on the imaging.

$1000 to$2000 per unit for 13 arc-seconds of accuracy? Or when you said
"resolution" did you really mean resolution, but the accuracy could be
less?

I assume that when you specified repeatability as being "exactly" where it
was before, you really meant "within the accuracy of the unit".

You'll find if you really need that kind of accuracy that you'll learn all
sorts of things about how apparently rigid materials are really quite
flexible in the presence of such things as light breezes and sunlight.
You'll also learn a lot about backlash, friction, possibly anti-backlash
gears, and other things you may never have wanted to know.

Expect that this will probably take at least a small team, and that you
won't get the price down to that level until you're building hundreds of
units per month.

Send me mail off list if I haven't discouraged you yet -- tim at
wescott design dot com (take out spaces as appropriate, replace phonetic
punctuation, etc.).

--
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

 

[john jardine]

[...]

two words: inverse kinematics

[...]
Add: Have no general solutions

Idea seems reasonable though. Being bloody minded I've just opened up a JVC
PTZ domecam, thinking there might be a bit of room to fineagle in some
mechanical contrivance such as a linear mass compensator. Not a sausage of
space. 'Tis engineered to perfection.

 

[Nico Coesel]

Wow lots of responses.

By low cost I'm talking between $1000 and$2000 per head. With a small
camera I really don't see the need to go heavy duty mounts - that
sagebrush unit is huge and heavy, and probably 10x times what I'm
looking at for cost. I'm wondering if it's using stepper's (big &
heavy).

Nope, the sagebrush unit uses servos. From your other replies it is
clear you are still underestimating the difficulty of what you are
trying to achieve. Sagebrush does custom designs. It won't hurt to ask
if they can build a unit in large quantities for a price you feel
comfortable with.

A fish-eye lense and software to un-warp the image may be more cost
effective and you'll be a apple to pan & tilt at infinite speed
without any mechanical wear. There are very nice high resolution
'machine vision' cameras on the market that can be used for such
purposes.

 

[John Larkin]

Wow lots of responses.

By low cost I'm talking between $1000 and$2000 per head. With a small
camera I really don't see the need to go heavy duty mounts - that
sagebrush unit is huge and heavy, and probably 10x times what I'm
looking at for cost. I'm wondering if it's using stepper's (big &
heavy).

There is a zoom lens on the cam, so this is just a Pan-Tilt head but
with communications hardware added for the zoom (so it's all on one
interface). There are several cameras we are looking at, all at less
than 2 lbs and fairly small dimensions.

BB - Yes for the most part the answer is in the software but the
higher accuracy the mount is, the less processing needs to take place
on the imaging.

Just curious: why is absolute pointing accuracy important? Why not
just slew it to within, say, 1 degree of the ideal direction, clamp
the drive, and take pictures? Does the exact center of some object
have to hit the exact center pixel of the sensor?

You could still use a good encoder to tell you the actual az/el if
needed. But the drive could be something soft and sloppy, toothed
belts maybe.

John

 

[Michael Brown]

Sanity check. Methinks the number of steps is a bit excessive.
Assuming a 30 degree viewing angle (my guess), 0.0036 degrees per step
equals a:
30 / 0.0036 = 8,333 pixel
wide display. Assuming a square image area, that's:
8K * 8K = 70 MegaPixels
I don't think that's very reasonable and well beyond the state of the
art in imaging resolution.

Working backwards, and again assuming a 30 degree viewing angle, a
good B&W 8 MegaPixel imager will be about 3000 pixels across, which
works out to:
30 / 3000 = 0.01 degrees/pixel
which methinks is more reasonable.

Of course, a narrow viewing angle, as in a telescope will work, but I
don't know of any 1 lb telescopes.

Continuing to brainstorming with insufficient information...

Perhaps do the positioning in two stages. Stepper motors,
motor-encoders, or selsyn:
<http://en.wikipedia.org/wiki/Selsyn>
drives to do the coarse positioning. The fine positioning is done
electronically using one of the various image stabilization methods
found in todays digital cameras and telescopes:
<http://en.wikipedia.org/wiki/Image_stabilization>

I recently saw a vehicle mounted camera that could generate perfectly
stabilized images while literally bouncing down a dirt road. (It
could also detect motion and center the image on the motion). Sorry,
but I can't seem to find the manufactory.


Not me. I'm buried, lazy, obsolete, burned out, etc.


Methinks this be somewhat more complexicated for the average hacker.

--
Jeff Liebermann je...@cruzio.com
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558

Jeff,

The reason for the extended accuracy (and again, we're talking roughly
13-18 arcseconds) is because we often work at the far extents of the
zoom lens, we're not imaging the whole wide shot at once in high res
as that wouldn't call for such accuracy.

 

[Michael Brown]

Nope, the sagebrush unit uses servos. From your other replies it is
clear you are still underestimating the difficulty of what you are
trying to achieve. Sagebrush does custom designs. It won't hurt to ask
if they can build a unit in large quantities for a price you feel
comfortable with.

A fish-eye lense and software to un-warp the image may be more cost
effective and you'll be a apple to pan & tilt at infinite speed
without any mechanical wear. There are very nice high resolution
'machine vision' cameras on the market that can be used for such
purposes.

Just to add a little more info:

Sony EVI-D70: 0.075 degrees resolution (270 arcseconds)
Sony BRC-300: 0.0048 degrees resolution (17 arcseconds)

They both have an impressive resolution but relatively the same
accuracy (much less than the resolution, probably 360 arcseconds or
thereabouts). It seems doable in small form factor but I'm not sure
the encoding methods they use, notably both cameras have the same
accuracy.

I'm definately interested in pursuing ideas regarding a software based
system but the biggest factor is image resolution. I don't know
whether or not I can achieve similar quality using a high rez machine
vision camera using software panning versus a normal res camera
zooming and moving and such. Certainly it's benefits are directly
financial, but quality is of big importance as well. If there are any
programmers out there that have some experience doing this and want to
give it a shot, hit me up.

 

[Vladimir Vassilevsky]

I'm looking for someone who can take on a project for my company, to
design a very accurate PTZ mount for a small camera. The mount will
need to have a resolution of around 100,000 steps (0.0036 degrees per
step), and high accuracy positioning (repeatably it should go to the
same exact spot when it's told to via serial/network communication
interface). The camera weight will be around 1lb, so it doesn't have
to be too heavy duty. It also needs to maintain a minimum speed of 60
deg/s for movement (ideal is 100-120 deg/s).

As noted by many, it would be difficult to satisfy those requirements
because of the mechanical constraints.

It could be possible to use several fixed cameras pointed at the
different angles and synthesise the rotated image by means of the
digital video processing.

Anyways, thanks guys - looking forward to hearing from some talented
hackers

Hackery doen't seem to be an option in this case.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

 

[JosephKK]

Michael Brown mikeb@360replays.com posted to sci.electronics.design:

Wow lots of responses.

By low cost I'm talking between $1000 and$2000 per head. With a
small camera I really don't see the need to go heavy duty mounts -
that sagebrush unit is huge and heavy, and probably 10x times what
I'm looking at for cost. I'm wondering if it's using stepper's (big
& heavy).

There is a zoom lens on the cam, so this is just a Pan-Tilt head but
with communications hardware added for the zoom (so it's all on one
interface). There are several cameras we are looking at, all at less
than 2 lbs and fairly small dimensions.

BB - Yes for the most part the answer is in the software but the
higher accuracy the mount is, the less processing needs to take
place on the imaging.

You can barely get 1 degree accuracy at that price point. (us$1000).
I am actually surprised that you can even get that accuracy in a
standard product at all.

 

[JosephKK]

Michael Brown mikeb@360replays.com posted to sci.electronics.design:

I'd be willing to pay what it costs Right now I'm trying to get
an estimation on who has the capabilities of developing the mount,
and a rough estimation on the engineering costs. We would be looking
at proceeding with this in January of 2008, as far as a timeline
goes. It may even lead to a full-time position if it is desired.

Actually i expect that you will have more problems dealing with
inconsistent zoom of the zoom lens that the camera positioning error.
Small errors in the zoom lens are much harder to deal with than mere
lateral / vertical image shifts.

 

[JosephKK]

john jardine john.jardine@idnet.co.uk posted to
sci.electronics.design:

message [...]

two words: inverse kinematics

[...]
Add: Have no general solutions

Idea seems reasonable though. Being bloody minded I've just opened
up a JVC PTZ domecam, thinking there might be a bit of room to
fineagle in some
mechanical contrivance such as a linear mass compensator. Not a
sausage of space. 'Tis engineered to perfection.

Even if you could readily stuff high resolution position sensors into
that enclosure, it may not be enough. The expected zoom lens kills
everything.

 

[JosephKK]

Michael Brown mikeb@360replays.com posted to sci.electronics.design:

Jeff,

The reason for the extended accuracy (and again, we're talking
roughly 13-18 arcseconds) is because we often work at the far
extents of the zoom lens, we're not imaging the whole wide shot at
once in high res as that wouldn't call for such accuracy.

You should try thinking about different solutions when you mix in that
zoom lens. Zoom lenses are not all that repeatable. That part of
the solution needs to be pressed into the software. Not that is easy
there either.