[Gimbal]

That's a nice test. Except that everything happens far too fast for the naked eye to see. It would suffice if the sensor is just a bit delayed to reach its starting position and it would still appear to sit locked in the middle to the naked eye.

Of course, but it would be a highly illogical behavior, whereas positioning in the middle as if SR is off would be a very logical (and simple) way of taking care of the situation. I can't really see a reason for it to be delayed either, either it is ready or it is not, a simple software switch. But you are right, without precise measurement one cant be 100% sure.

[dosdan]

SR works by electronic coriolis force sensors which measure angular velocities, not acceleration. This is why panning can be a problem

To permit SR during panning, wouldn't it be a matter of ignoring the horizontal/yaw component and use just the vertical/pitch & tilt/roll components. With panning, the angular momentum of swinging a long & heavy telephoto lens should "dampen" any muscular horizontal shaking force. Perhaps this could be done in a firmware upgrade.

Dan

[Gimbal]

At least SR works when panning in movie mode. Just pan and suddenly stop, and you will see that the panning stops softer then you did and then swings back a little, adapting to the new state of standing still.

Edit: Or wait a minute, that is not really the same thing....

[Lowell Goudge]

To permit SR during panning, wouldn't it be a matter of ignoring the horizontal/yaw component and use just the vertical/pitch & tilt/roll components. With panning, the angular momentum of swinging a long & heavy telephoto lens should "dampen" any muscular horizontal shaking force. Perhaps this could be done in a firmware upgrade.

Dan

Some makers do allow this. I forget who at the moment

[jamesm007]

That's a nice test. Except that everything happens far too fast for the naked eye to see. It would suffice if the sensor is just a bit delayed to reach its starting position and it would still appear to sit locked in the middle to the naked eye. So, the only test would be doing test shots on a vibrating bench, waiting for variable length amounts of time. AFAIK, nobody did this test.

The system can work while panning as I show and as Pentax says to use it. Lost shots are due to other factors. The SR system is not "locked". Its continually reading the gyros that can't detect motion and the hall sensors that can detect motion. Thats why Pentax SR needs not to be shut off for panning. The hall sensors know which direction you are moving the gyros read any shake. The calculations and movements are faster then we can imagine - (perhaps some can), try using live-view. The system is always correcting your movements during live-view. How is it doing that with no time to get a "lock".

Official Pentax Advertisement-
"The in-body Shake Reduction of the K20D helps to reduce the effects of camera shake. The K20D Shake Reduction (SR) feature allows you to shoot at lower shutter speeds without camera shake as the system is electromagnetically controlled to detect camera shake and moves a free floating image sensor to compensate. As SR is built into the camera body any lens, right back to screw thread, can be used to take advantage of this feature, even while panning."

Airshow challenge with the Pentax K20D

All of my Gull pics are me panning and are tack sharp. I lose no pics because I don't turn off SR nor does the owners manual tell you to turn off SR.

When you half press the shutter the system is only determining (but not moving anything) where to place the sensor platter using info from the gyros, lens, CPU. Then when you press the shutter button it can also read the hall sensors (before it lets the shutter fire), why does the Pentax system use hall sensors, and always has? Can hall sensors detect movement of the body versus sensor platter during a pan? During a high speed (shutter speed) pan shake is not important so there is no need for info from the gyros only hall sensors. The system is more than fast enough to hold the sensor still during high G movements and panning or better to say still long enough for the picture to be taken. Remember the sensor (photo) is just sitting on the bottom of the camera with SR off (or on) while you are panning. Then when you fully press the shutter button it jumps up (or pulled up by magnetic force) into proper position. If SR is on it may (as in some of my pics) try to stabilize the platter if it can't the CPU (or program) will just default it to a "off" position (centered with lens). Also no matter what speed the shutter is set at everything is under CPU control. So when you press the button it waits for the mirror to flip up and stop bouncing; enough time to pull up the platter and start reading the hall sensors before the CPU lets the Shutter fire. It all happens in a blink of the eye.

This can only be settled by official Pentax statement IMO.

Again I don't believe SR is making my pics sharper, but nor do I think its making my pics softer.

BTW handholding is what I speak of. Any panning or movements on a tripod requires the SR to be off as a tripod is the best SR.

[derekkite]

Some makers do allow this. I forget who at the moment

I had a canon super zoom that did that. Three modes; off, horizontal panning and full on.

It seems like at worst the SR doesn't ruin anything, at best it gives you a couple of stops. Pretty good implementation I would suggest.

Before my K5 I had an Olympus E420 with no stabilization. I used a 300mm zoom and got reasonable shots with short enough exposures.At 400 iso it looked like what I get at 1600+, so I can keep exposures shorter on my K5. I got a shot the other day, 420mm from a distance of 250 ft a bobcat in a tree. Handheld 1/1000 sec 640 iso, f4. The cropped result is shown here: Bobcat on dog walk | Exploring Kootenay Lake

I would have been much slower on my oly and probably would have missed the shot. The SR no doubt helps, but the higher iso seems to make much more difference. I suppose that in even lower light conditions the SR would give me the couple stops to make the shot possible.

[Gimbal]

Then when you press the shutter button it can also read the hall sensors (before it lets the shutter fire), why does the Pentax system use hall sensors, and always has? Can hall sensors detect movement of the body versus sensor platter during a pan?

The hall sensors are used to detected the position of the sensor platter relative the body. In order for the SR system to move the sensor platter to any given position it has to know where the sensor platter is. With other words, they are feedback sensors for the servo-system positioning the sensor platter.

With "lock" mentioned earlier means that the SR system has built up enough information about the shake frequency, amplitude and phase and "sort of" a built up a general direction in which the camera are pointing. Kind of you are wobbling back an forth but the mean direction is in the middle, so it takes a while to work out a mean direction. (Although it's not a direction in absolute numbers, it's all relative.)

[falconeye]

Kind of you are wobbling back an forth but the mean direction is in the middle, so it takes a while to work out a mean direction. (Although it's not a direction in absolute numbers, it's all relative.)

To be frank, we have no idea what the SR does while initializing.

Hand shake "noise" is a type of 1/f noise but with a "shoulder" around 4 - 10 Hz. I documented this in some of my papers. So, any initial SR measurement shouldn't take longer than the selected exposure time plus 1/4s or less. Because best you can do is measure amplitude and phase by integration over the intended exposure time.

You're right about the Hall sensors. They're part of the shift motor feedback loop and shouldn't be even mentioned in a discussion of proper SR operation. It only confuses the topic.

[audiobomber]

Some makers do allow this. I forget who at the moment

This is common with in-lens stabilization, I've not heard of it for in-body.

I had a canon super zoom that did that. Three modes; off, horizontal panning and full on.

Canon only uses in-lens stabilization.

[Gimbal]

To be frank, we have no idea what the SR does while initializing.

Right, we can only guess based on the cameras behavior and what is needed to make a camera behave like that. A stab at reverse engineering.

I usually cover my ass with "IMO" or "it is my understanding", but I missed it this time.

[jamesm007]

To be frank, we have no idea what the SR does while initializing.

Hand shake "noise" is a type of 1/f noise but with a "shoulder" around 4 - 10 Hz. I documented this in some of my papers. So, any initial SR measurement shouldn't take longer than the selected exposure time plus 1/4s or less. Because best you can do is measure amplitude and phase by integration over the intended exposure time.

You're right about the Hall sensors. They're part of the shift motor feedback loop and shouldn't be even mentioned in a discussion of proper SR operation. It only confuses the topic.

When you have SR off or on it uses data from the Hall Sensors to keep the platter square. These same Hall Sensors will read when the body of the camera applies any amount of force that may lag the sensor platter. This is what makes it work while panning with SR off. I know you hold to the feedback loop without the CPU during shutter activation. However if you see pictures of a panning movements taken at 1/2000 and the system says "Stabilized" its not lying. I will show Pentax stating such latter.


Remember the sensor platter is rolling on ball bearings. Put a baseball on the floor of your car and then accelerate and stop and go backwards. What happens to the baseball? The is what the Hall Sensors read. This is how it keeps the sensor platter still during panning with SR off. It does in fact work with while panning with SR off, can be no question of that. Now just add SR on, the system can still compensate is my belief and what I am reading.

Now your point of not having enough time to develop a compensation signal from the gyros before the feedback loop, I understand what you say fully. But again reading the patents, looking at schematics and meta data from pictures tells me its always forming a compensation signal and there is no feedback loop for SR only the Hall Sensors. This is where me and you differ.

This past November 18 2011 I saw far in the distance a set of large birds coming my way. I ran and got my K20D and DA55-300mm. I started to take pics hand holding. The first few pics I had to manual focus as they were too small for AF to work. Then as they got closer I could switch over to full AF-C 11 point mode. I had to pan the camera vertically. They came at me from afar and went over my head. In all I have 24 pics. I could take that many because of the couple second delay during the MF time and the switch to AF. All of them from the first to last say stabilized in the meta data. The same thing happens even if I pan across horizontal even if I pan 360 degrees up/down all around me. There can be no doubt by my pics the system at the very least needs not be shut off. Only if you are shooting at a low shutter speed.

The SR system does not lie. If a picture says "Stabilized" in the meta-data the CPU is not confused. If its not ready it will say so, if its not stabilized it will say so. Again its best not to state things that are not said in Pentax literature and moreover owners manual. The owners manual only says to turn off SR and use a tripod while panning if you don't have enough shutter speed. In no place does it say to turn off SR if your going to pan.

Falk your going off of your thoughts of the patents and your experiance, but no facts, but so am I. But if we both stick to what the Pentax owner manual says we would not touch the SR switch unless putting the camera on a tripod. However the Pentax Ad I showed you says to use it panning? Also the speed of the system is probably much greater now then with the original patents. The shakey Hand is to generate a superior SR alogorithm like in low light shake. However it can generate one while panning if you half hold the shutter button during the action.

Here is the first and last (you can read meta data) of 24 pics taken a couple weeks ago while hand holding panning and going over my head. The dark area in the last pic is my house. I had to stop tracking because they flew over my house (high in the air) and when I tried to get them the were heading away from me (not the best view of a bird). I thought at first they were Hawks but in fact they are vultures. Of course I have much better examples than this. Just these are easy to get to as I have to go to work

[jamesm007]

To be frank, we have no idea what the SR does while initializing.

Hand shake "noise" is a type of 1/f noise but with a "shoulder" around 4 - 10 Hz. I documented this in some of my papers. So, any initial SR measurement shouldn't take longer than the selected exposure time plus 1/4s or less. Because best you can do is measure amplitude and phase by integration over the intended exposure time.

You're right about the Hall sensors. They're part of the shift motor feedback loop and shouldn't be even mentioned in a discussion of proper SR operation. It only confuses the topic.

How about in a discussion of the need to turn off SR while panning even though Pentax literature and manual do not say shut it off while panning? Only if you have motion blur then shut off SR and use it on a tripod. Really I think the system is much more capable then you or others believe. This is why even though the owners manual does not say shut it off panning people think they should. That's IMO until Pentax says other-wise. And I mean Japan engineer.

[Gimbal]

I have no doubts about SR working while panning, as long as you pan in a steady pace.

But if you are not panning in a steady pace, then things might get ugly. Say that the SR is ready and you suddenly alter speed or angle of the pan (the birds suddenly turned) and fire, the first milliseconds of the altered speed might be considered an involuntary shake and the SR will do its best to compensate for it. But that is in error because you have actually altered your pace to follow your target in their altered track, and the result will be less then perfect.

So my advice is to use SR if your doing long smooth pans, turn it off if your tracking fast and quick action (that is, fast and quick camera movements).

[Lowell Goudge]

in looking at this whole discussion, I think we need to step back a second and think about this.

I believe shake reduction is based upon acceleration, not displacement.

As such, when panning, if the angular velocity is zero then SR will not or should not cause any blurr in the image, but unfortunately that is only true if you are at right angles to the driection of travel of the subject, any other viewing angle will have acceleration (assumed laterally) as you track and as a result there could be image blurr in the direction of motion due to the panning.

BUT, is that blurr better or worse, than an unassisted panning shot at the same shutter speed, I doubt it, because all the other camera motions including horizontal jitter as you pan will be corrected.

just my $0.02

[Gimbal]

I believe shake reduction is based upon acceleration, not displacement.

As such, when panning, if the angular velocity is zero then SR will not or should not cause any blurr in the image, but unfortunately that is only true if you are at right angles to the driection of travel of the subject, any other viewing angle will have acceleration (assumed laterally) as you track and as a result there could be image blurr in the direction of motion due to the panning.

But a steady angular velocity (as in a steady pan) COULD be filtered out, and then SR would work wile panning as well. Pan with live view on and you will see (I hope, as I don't have the camera with me to try it myself) that the sensor plate isn't bouncing into the wall again and again.