Anti-Shake with IS - 'double' system?

First of all, this thread isn't about whether you like the idea of image stabilization or not, it's about making the next generation IS system more effective. Let's assume there is a market for this system.

Secondly, as I have written several times in this thread, it's not just a matter of attaching a Canon IS lens on a Konica/Minolta body. The two systems has to be INTEGRATED with one set of gyros and one processor so they won't 'over compensate'. Eg. the in-body system would work normally (say up to 2 degrees shake), while the in lens system would only step in when the in-body system can't handle the job (2 - 4 degrees shake).

I am very interested in WHY you think it will work or not. Please give a reason.

Thanx in advance

GregGory said:

I am very interested in WHY you think it will work or not. Please
give a reason.

Click to expand...

Why not just an improved single stabilizer - I don't really see what having two stabilizers in one system will accomplish - achieving perfect coordination would just be too complicated and expensive.

--
Misha

GregGory said:

I am very interested in WHY you think it will work or not. Please
give a reason.

Click to expand...

Can I give more than one? ;-)

First of all, the limitation on the effectiveness of IS is rarely the travel limits. So expanding the travel limits wouldn't help much.

Next, the travel limit on the lens is there because of the available AOV of the front section. In other words, the system would vignette if you exceed the in-lens limits anyway.

That part is not true if you have a full-frame IS lens on a crop-sensor camera. In this case, moving the sensor would increase available travel substantially. However, the real limit on IS effectiveness is bandwidth, not travel. There's no reason to believe the bandwidth's of the two systems aren't similar so the best they could do if working together is add a pole to the response which means the combined system could be at most 1 stop better (3 db - twice as good) and only at the corner frequency boundary and beyond.

For truely better IS, we need lighter components to move and/or more powerful actuators to move them so that bandwidth can be increased.

I frankely don't see why this is necessary as the current systems provide me with 3-6 stops on the best lenses which is A LOT of performance - I've gotten 2 seconds at 24mm on the 24-105L IS at 24mm and 1/4 second at 200mm on the 70-200 2.8L IS. What this means to me is that subject motion is basically always the limit because IS relaxes the handholding limitation so far. There are some exceptions (very long exposures) in which case a tripod does the trick.

--
Lee Jay
(see profile for equipment)

misha said:

Why not just an improved single stabilizer - I don't really see
what having two stabilizers in one system will accomplish -
achieving perfect coordination would just be too complicated and
expensive.

Click to expand...

There are two aspects of an effective IS system. Firstly, how well the system can counter 'mimic' the fine handshake movement. This can always be improved within the current systems, but it is quite good already, so the there's not much room for improvement and no 'additional stops' to gain.

The other aspect is simple geometry (how much the sensor can be moved, and how much you want to bend the ray) which sets the upper limit how much the camera can be 'off' target. This is where the new system would more than double the effectiveness in gained stops (theoretically at least).

Hi ljfinger!

Very interesting

Lee Jay said:

ljfinger wrote:
Can I give more than one? ;-)

First of all, the limitation on the effectiveness of IS is rarely
the travel limits. So expanding the travel limits wouldn't help
much.

Click to expand...

Hmm, I have the opposite impression.

Lee Jay said:

Next, the travel limit on the lens is there because of the
available AOV of the front section. In other words, the system
would vignette if you exceed the in-lens limits anyway.

Click to expand...

Yes, that's the point of the 'double' system. You don't have to extend the In-lens travel.

Lee Jay said:

That part is not true if you have a full-frame IS lens on a
crop-sensor camera. In this case, moving the sensor would increase
available travel substantially. However, the real limit on IS
effectiveness is bandwidth, not travel. There's no reason to
believe the bandwidth's of the two systems aren't similar so the
best they could do if working together is add a pole to the
response which means the combined system could be at most 1 stop
better (3 db - twice as good) and only at the corner frequency
boundary and beyond.

Click to expand...

That is true if you are right about the bandwidth being the limit, not the travel. May I ask where you have this info from?

Lee Jay said:

For truely better IS, we need lighter components to move and/or
more powerful actuators to move them so that bandwidth can be
increased.

I frankely don't see why this is necessary as the current systems
provide me with 3-6 stops on the best lenses which is A LOT of
performance - I've gotten 2 seconds at 24mm on the 24-105L IS at
24mm and 1/4 second at 200mm on the 70-200 2.8L IS. What this
means to me is that subject motion is basically always the limit
because IS relaxes the handholding limitation so far. There are
some exceptions (very long exposures) in which case a tripod does
the trick.

Click to expand...

You must have a cast-iron torso and HI-TEN steel arms! Most people only get 2-3 stops. I would love to gain 6 stops.

GregGory said:

GregGory said:

Next, the travel limit on the lens is there because of the
available AOV of the front section. In other words, the system
would vignette if you exceed the in-lens limits anyway.

Click to expand...

Yes, that's the point of the 'double' system. You don't have to
extend the In-lens travel.

Click to expand...

That won't help. You would need to expand the image circle in the lens to get the double system to work without vignetting unless, as I said, you were using a full-frame lens on a crop camera (essentially the image circle is already expanded).

GregGory said:

That is true if you are right about the bandwidth being the limit,
not the travel. May I ask where you have this info from?

Click to expand...

My own experience with 5 IS lenses and I've seen an analysis on it somewhere. Maybe the Canon Bebit site, I'm not quite sure.

GregGory said:

You must have a cast-iron torso and HI-TEN steel arms! Most people
only get 2-3 stops. I would love to gain 6 stops.

Click to expand...

Actually, being good or bad doesn't matter that much. The difference is where your shakes happen to occur. I'm very good at low-frequency pointing but have a lot of high-frequency shakes. The IS system is great at eliminating the low-amplititude higher-frequency shakes but intentionally lousy at removing the lower-frequency pointing motions. Therefore, IS suits my problem well.

--
Lee Jay
(see profile for equipment)

ljfinger said:

GregGory wrote:

That won't help. You would need to expand the image circle in the
lens to get the double system to work without vignetting unless, as
I said, you were using a full-frame lens on a crop camera
(essentially the image circle is already expanded).

Click to expand...

So 'oversized' (compared to the sensor) IS lenses are need (just like the KM lenses), an easy fix.

Lee Jay said:

My own experience with 5 IS lenses and I've seen an analysis on it
somewhere. Maybe the Canon Bebit site, I'm not quite sure.

Click to expand...

If it comes to your mind, please post it here!

Lee Jay said:

Actually, being good or bad doesn't matter that much. The
difference is where your shakes happen to occur. I'm very good at
low-frequency pointing but have a lot of high-frequency shakes.
The IS system is great at eliminating the low-amplititude
higher-frequency shakes but intentionally lousy at removing the
lower-frequency pointing motions. Therefore, IS suits my problem
well.

Click to expand...

Yes this is my main concern, the high amplitude low freq. motion, but I suppose it's not impossible to go slower with the IS prism? The gyro 'monitoring systems' might be a bottle neck, but probably could be improved over time.

Maybe this system should be tested on fixed lens small sensor cameras then gradually move to larger sensors (larger IS prisms).

GregGory said:

The gyro 'monitoring systems' might be a bottle neck, but probably
could be improved over time.

Click to expand...

Most types of solid-state accelerometers are notoriously lousy at low-frequency. This is why they are not used in inertial navigation systems.

--
Lee Jay
(see profile for equipment)

ljfinger said:

Most types of solid-state accelerometers are notoriously lousy at
low-frequency. This is why they are not used in inertial
navigation systems.

Click to expand...

Is this because the accelerometer technology isn't matrue yet, or is it a 'dead end' where no significant improvements can be expected?

GregGory said:

ljfinger said:

Most types of solid-state accelerometers are notoriously lousy at
low-frequency. This is why they are not used in inertial
navigation systems.

Click to expand...

Is this because the accelerometer technology isn't matrue yet, or
is it a 'dead end' where no significant improvements can be
expected?

Click to expand...

I'd say the basic technologies are not going to get any better in that area. There are more accurate technologies but they are expensive and large. I think a cheap, accurate at low-frequency accelerometer has yet to be invented.

--
Lee Jay
(see profile for equipment)

GregGory said:

ljfinger said:

My own experience with 5 IS lenses and I've seen an analysis on it
somewhere. Maybe the Canon Bebit site, I'm not quite sure.

Click to expand...

If it comes to your mind, please post it here!

Click to expand...

On the KM in body method, the system is barely ever at the limits. It can moves something like nearly a 1/4 inch. But it typically barely moves. There is video on the net of the system running on a stand so you can see it. Its amazing to see how far it can travel.

Greg,

Yes, if they were integrated together you could use each end to best effect.

The in-camera AS of Konica-Minolta works by moving the sensor in an X-Y plane. That is it does not tilt the sensor but slides it up/down and right/left.

The in-lens IS and VR of Nikon and Canon work by tilting an element inside the lens.

The KM system works at a broader range of freqency than the IS/VR systems, but KM does worst with tilt errors.

So, with a very clever, and complicated, stabilization algorithm, one could design a better system using both methods. It would work on x-y errors and tilt errors.

GregGory said:

The Konica-Minolta Anti Shake sensor mount and the in lens image
stabilisation systems (Canon, Panasonic etc.) have proven their
worth for some time now. Both systems are good for 2 - 3 stops give
or take. I'm wondering why nobody (as far as I know) has combined
the two systems to gain 4-6 stops (probably even more)? As far as I
can see there is no reason why it shouldn't be possible to combine
the two sytems. Of course the IS in the lens must be 'hot wired'
with the AS in the body otherwise they would 'over overcompensate',
but theroetically it should't be impossible.

Looking forward to an explanation

Click to expand...

--
Rick

hardsuit wrote:
I foregot to tell you something. don't cross the streams.
why?
it would be bad?
bad , what do you mean bad.
Imagine every molocule in your body accelerating
at the speed of light and then exploding.

yah, I guess that would be bad.

--
' lets see what's out there.....engage'

GregGory said:

GregGory said:

I frankely don't see why this is necessary as the current systems
provide me with 3-6 stops on the best lenses which is A LOT of
performance - I've gotten 2 seconds at 24mm on the 24-105L IS at
24mm and 1/4 second at 200mm on the 70-200 2.8L IS. What this
means to me is that subject motion is basically always the limit
because IS relaxes the handholding limitation so far. There are
some exceptions (very long exposures) in which case a tripod does
the trick.

Click to expand...

You must have a cast-iron torso and HI-TEN steel arms! Most people
only get 2-3 stops. I would love to gain 6 stops.

Click to expand...

Well, I've managed to get some (I think) reasonable images at 35mm equivalent for 1/4 of a second with no stabilisation. I guess that the 2 second pictures you took at 24mm, that would be about 36mm with FOV multiplication factor? Assuming we are both similarly capable of holding a camera still it sounds like roughly a 3 stop advantage... maybe I'm being less fussy, too...

Here's a 1/4 sec handheld image, anyway. The jumper looks pretty sharp to me, to the point that other limitations of that camera are a greater factor.



p.s. Any Archers listeners out there? I think that might be the last picture taken of Mary Wimbush before she died...

ptodd said:

Well, I've managed to get some (I think) reasonable images at 35mm
equivalent for 1/4 of a second with no stabilisation. I guess that
the 2 second pictures you took at 24mm, that would be about 36mm
with FOV multiplication factor?

Click to expand...

No, that was on a 5D.

Here's a 1/4 second test shot at 200mm on a 20D:
http://photos.imageevent.com/sipphoto/samplepictures/20D_6765.JPG

--
Lee Jay
(see profile for equipment)

ljfinger said:

ptodd said:

with FOV multiplication factor?

Click to expand...

No, that was on a 5D.

Click to expand...

Oh, ok. Still, that essentially means that there is less of a gain than I thought...

Lee Jay said:

Here's a 1/4 second test shot at 200mm on a 20D:
http://photos.imageevent.com/sipphoto/samplepictures/20D_6765.JPG

Click to expand...

That certainly does seem to be pretty steady. I'd be quite interested to see some sort of comparison, or if you could just comment how long an exposure you think you can get away with without the IS.

Cheers,
Peter

Mojo151 said:

Greg,

Yes, if they were integrated together you could use each end to
best effect.

The in-camera AS of Konica-Minolta works by moving the sensor in an
X-Y plane. That is it does not tilt the sensor but slides it
up/down and right/left.

The in-lens IS and VR of Nikon and Canon work by tilting an element
inside the lens.

The KM system works at a broader range of freqency than the IS/VR
systems, but KM does worst with tilt errors.

So, with a very clever, and complicated, stabilization algorithm,
one could design a better system using both methods. It would work
on x-y errors and tilt errors.

Click to expand...

All current camera stabilisation (lens or sensor) AFAIK corrects tilt errors
(rotations around two (of three) axes). You can't measure
translations easily (only accelerations), and translations only have
significant effect for close up anyway.

Just my two öre
Erik from Sweden, F Z 5 owner

ptodd said:

ljfinger said:

ptodd said:

with FOV multiplication factor?

Click to expand...

No, that was on a 5D.

Click to expand...

Oh, ok. Still, that essentially means that there is less of a gain
than I thought...

Click to expand...

1/24th would be the 1/f rule, 2 seconds is about 5.5 stops.

ptodd said:

ptodd said:

Here's a 1/4 second test shot at 200mm on a 20D:
http://photos.imageevent.com/sipphoto/samplepictures/20D_6765.JPG

Click to expand...

That certainly does seem to be pretty steady. I'd be quite
interested to see some sort of comparison, or if you could just
comment how long an exposure you think you can get away with
without the IS.

Click to expand...

I need about 1/400th for that shot without IS. That's about 6.5 stops.

--
Lee Jay
(see profile for equipment)

ljfinger said:

Lee Jay said:

Lee Jay said:

Here's a 1/4 second test shot at 200mm on a 20D:
http://photos.imageevent.com/sipphoto/samplepictures/20D_6765.JPG

Click to expand...

That certainly does seem to be pretty steady. I'd be quite
interested to see some sort of comparison, or if you could just
comment how long an exposure you think you can get away with
without the IS.

Click to expand...

I need about 1/400th for that shot without IS. That's about 6.5
stops.

Click to expand...

OK. I was thinking that perhaps a big part of it was that you might be able to work somewhat beyond the 1/f rule. I feel I sometimes can get vaguely acceptable results up to, for example 1/4 @ 35mm, which is about three stops longer... A lot of this hinges on our standards though, I suspect mine are relatively low. I've done very little telephoto stuff, so I can't really comment there.

Still, I see 1/f as a very rough rule of thumb, for something which is rather individual in practice. This is why I was interested to hear how felt you could do without IS, and from what you say, it seems like that lens is doing a pretty incredible job for you. I wonder how much difference there is in the various different lens / sensor based approaches, and how much they will develop in future.