[TomGarn]

I did a 'scientific' test of this
https://www.dropbox.com/s/qk4pnm5w2rghovj/PK3_0696.mp4

lister has a fixed Kopter here ... and a K-3 fixed on a tripod ... (in that video)

We usually should not use SR on tripod - they tell us - and of course that is right !
The SR of any kind will work to smoothen some kind of shaking - but there is none.
The body on the tripod doesn't shake - so the SR works on the image to calm it down.

In my case - with the anlog / mechanical SR of the K-7 - the SR now helps to keep the image
of the street still - but the car is moving and you can see it dancing up and down in that video.
Usually the camera must shake a little of course when handheld or fixed to the car directly. This
is the case with professional filming as well - mostly. And it has to be shaky - but that is natural
and OK and gives you a good feel for the action as well - The shoulder of the cameramann
compensates just a little then - but this is the same for our body - being a driver

Here I somehow managed a "flying eye" that glides over the street - and just the car is dancing.
This is the same effect like: "the station starting to move while the train is standing still" ... Right ?
Some interesting phanomenon !

If you do that car-video with a K-3 ... it may work out right just as well ? (or almost as well because it seems
to work a little shizophrenic - here and there - as if not clear what shakes next - the camera or the image.
So maybe the analog SR has some advantage here ... maybe more smooth and balanced ?

In my street-video: The car is wobbling a lot - but the sensor SR does the same "wrong" thinking
as the digital SR: It "assumes" the camera is shaking- but in both our examples it is not ...

1. We are not to put a camera on tripod with SR - for videos AND photos ... so the result has to be
thought about - because it is always opposite to what a SR is supposed to do - and planned for.

2. Panning is not possible with mechanical SR at all - you then ought to learn how to slow down at
the end of the pan and so somehow catch and neutralize the anti-movement, but that is very tricky.
I have tried it - sometimes you are almost lucky - most times not at all. You can do it but it just looks
very unnatural and stupid in most cases.

3. SR is just for stills - camera being handheld - not panning. Then it works OK (at least on my K-7)

---

Last edited by TomGarn; 12-31-2013 at 03:21 AM.

[kadajawi]

@Tom: If you shoot with a very long lens, even a tripod may not give you a stable video. Having SR do the rest helps.


The car sample is more pleasant with the SR balancing out the car movement. After all we are supposed to see the street, it doesn't matter if the car is sharp and without movement.


I prefer the camera to move a little... and at that point the SR comes in handy. I don't like the perfectly still "camera is on tripod" look, but those small fast camera shake movements are to be avoided too. They look bad.


Panning... well, yes and no. If you are going for a perfectly smooth pan, perfectly executed, then SR isn't for you. Use a tripod, switch of SR. But I think SR would do great in a more... cinematic sort of style. The camera is moving all the time, and the SR filters out the worst shake. The jidder, the stuff that leads to the jello effect. The panning problems may still be visible, but they go under in the constant movement. The sort of shots you would normally put the camera on your shoulder for. I think _that_ can be done handheld with mechanical SR, perhaps even better than that. (Though the look, the way of the movement may be too different...)

[TomGarn]

True. I didn't use K-7 with SR while doing documentaries yet.
But I know that I will see this problematic SR while moving the camera
up and down - and from left to right - I just know it - this is kind of a deep feeling
because I did these things often enough with all kinds of cameras.

True, many will not see that SR problem and understand what is happening there -
but they will truely feel it - while watching ... The audience just cannot say where the
problem is rooted - but I do films for my watchers - so I want to do it right - and not
make them vomit ... because of wobbling pictures dancing back and forth ~ ~ ~ ~ ~

One thing more: We all used to think how great it is to have a sensor-SR ... so we can
buy cheep glasses and have'm stabilized through our Pentax-body. For video this
advantage now has almost gone ... we did'nt talk about that much here. Why ?

Look at this video. It is done on a Canon 60D with a Tamron 17-50 / 2,8
It is almost the same Tamron I got for my K-7 (soon to be K-5)
but the one for for my Pentax is (of course) without IS (SR)

I use the opical IS of that Tamron always ON and because it is in the glass and not the body
and because it is not that strong as the Pentax sensor SR - it makes no trouble at all - but
here and there it is shaking much more, yes.

I did not edit it - the cutter redid it a few times - when I told him what was wrong - then he quit
the job and uploaded it- he just gave up. But to some point it is usable. I do of course see my
own limitations here as well ... This is all done with displayviewfinder on - out of my bare hands:

You Tube

You Tube

[lister6520]

lister has a fixed Kopter here ... and a K-3 fixed on a tripod ... (in that video)

The reason for the camera being fixed is to fix one of the variables and figure out the anatomy of the SR algorithm as it were. A proper SR (using the gyroscopes) whether done mechanically or digitally in PP should not produce any movement of a fixed subject in such a situation. The test is not meant to simulate a real life situation but rather to analyse how the SR is designed.

We usually should not use SR on tripod - they tell us - and of course that is right !
The SR of any kind will work to smoothen some kind of shaking - but there is none.
The body on the tripod doesn't shake - so the SR works on the image to calm it down.

That is sometimes true and sometimes not. If you are using short lenses and/or a very heavy and steady tripod and/or there is no wind then it is true that SR cannot contribute anything. It shouldn't cause any harm either unless you are doing very long exposures (because of gyro drift) or panning rapidly for stills or video.
Anything but the heaviest tripods will vibrate in anything more than a breeze and with long lenses this becomes a problem. SR can and does fix that problem, if it is properly designed and working correctly. It does work fine on the K-r and I believe Pentax can and should put back in the K-3 what they already had and worked fine in the K-r.
As things are I occasionally still have to resort to the K-r instead of the K-3 for video, which for me is almost always 500mm on a tripod. It is a real pity I cannot use the K-3 for that as in almost all other aspects the K-3 is much better for video than the K-r but the crazy 'movie-SR' makes it almost useless for me.

If you do that car-video with a K-3 ... it may work out right just as well ? (or almost as well because it seems
to work a little shizophrenic - here and there - as if not clear what shakes next - the camera or the image.
So maybe the analog SR has some advantage here ... maybe more smooth and balanced ?

In my street-video: The car is wobbling a lot - but the sensor SR does the same "wrong" thinking
as the digital SR: It "assumes" the camera is shaking- but in both our examples it is not ...

The issue is not about whether it is anlog or digital or whether it is in lens or sensor shift.
The issues are three:
1. That the K-3 is not using the gyros to stabilise the image and instead relies on analysis of the video coming from the sensor. That is very odd as such a form of stabilisation is only used as a workaround when there is no gyro information available to feed the stabilisation algorithm.
2. That it is doing the stabilisation in post on the video coming from the sensor rather than doing it by shifting the sensor. Independently of the source of vibration information (gyro or image analysis) this method has one disadvantage that cannot be overcome. When using slow shutter speeds it can only correct for the apparent movement of the subject (due to camera movement) but cannot eliminate the blur that the movement caused. This results in stationary objects appearing to blur as the camera shakes, even if the stabilisation is perfect and keeps the objects looking stationary. The only way to eliminate such vibration induced blur is to use some form of machanical stabilisation. Whether by moving the sensor to track the image or moving a lens to keep the image steady on the sensor is immaterial. What is important is that during the time that the electronic shutter is open the image must remain motionless relative to the sensor.
3. The third issue is that the image stabilisation does not take into account rolling shutter. The stabilisation algorithm is applying an offset globally to the whole frame whereas it should be applying it on a line by line basis, applying a compensation relevant to the vibrational motion that was present at the time that line was being scenned. The ugly result of this is that jello effect now affects even stationary objects when the camera shakes. That jello effect appears due to panning or moving subjects is quite inevitable but there is no reason why we should have jello effect even on stationary objects when the camera shakes.

1. We are not to put a camera on tripod with SR - for videos AND photos ... so the result has to be
thought about - because it is always opposite to what a SR is supposed to do - and planned for.

As already mentioned above there are many situation where SR is useful, or indeed even necessary, when using a tripod.

2. Panning is not possible with mechanical SR at all - you then ought to learn how to slow down at
the end of the pan and so somehow catch and neutralize the anti-movement, but that is very tricky.
I have tried it - sometimes you are almost lucky - most times not at all. You can do it but it just looks
very unnatural and stupid in most cases.

This is not so. Panning is entirely possible with mechanical SR provided the coefficients are adjusted correctly. Having a choice of different coefficients to chose from is something I hope Pentax will introduce at some stage as no set is good for all situations. Shooting from a fixed tripod and handheld panning being two opposite extremes that could easily be accomodated by using different presets.

With all three Pentax SLRs that I have I have succesfully used SR while panning. There are some limitations but then again there are some photos that cannot be shot properly unless I pan with SR activated. One such example is shooting a (flying) plane at dusk with a 500mm lens at 1/100s. The trick is to track the target for a few seconds (with the shutter half pressed) so that the SR mechanism aligns itself to the constant movement. It will then effectively ignore that movement and compensate only for variations within that movement. I have tried it several times and it works and I have taken shots that I never could have without either SR or some elaborate motorised tracking mount.

You just have to know when to switch on SR and when not to.

3. SR is just for stills - camera being handheld - not panning. Then it works OK (at least on my K-7)

Maybe so, but I don't have a K-7 so cannot comment about that. What I can say is that on the K-r the SR comes in very useful when shooting video especially with long lenses on a tripod and I see no reason why they should not make it so also on the K-3.

[kadajawi]

Good one lister.

The only advantage that I see to analysing the video the way the K-3 does it is that it knows what movements follow, and thus can slow down to a stop from a pan. It basically knows what will happen in future. There are plenty of problems though of course, and apart from the loss in resolution this can easily be done on a computer, where you have more controls and better algorithms probably. Plus you can look further into the future.

It seems to me like this sort of processing is harder to implement than just having video stabilization through the sensor shift... Perhaps the new processor that Pentax uses since the K-30, that was supposed to be faster, was already equipped with this feature by Fujitsu, so they activated it rather than porting their own stabilizer from the older processor.

I believe I have heard that the K-3 does do stabilization in LiveView when you are in the stills mode.

@Tom: Your SR problems are a software issue. The lens probably has a more sophisticated algorithm to detect pans. Or perhaps more conservative because the SR can't compensate much. In any case these things can be fixed by Pentax. They should have several SR options... No SR, movie SR, movie SR with gyro, movie SR + mechanical SR (crop like movie SR, mechanical SR is active but additionally, based on gyro info and looking into the future the shake is further reduced, i.e. the remaining up and down etc. is reduced to get a steadicam look), mechanical SR (which has to be split into several modes, like one that recognizes pans and then won't or only slightly corrects the speed of the pan, so that when you stop it won't pan back).

[TomGarn]

yes, different to our real own bandwidth in experimenting with our SR we will come to better and more scientific results ;-)
You did a good work here lister ... quite deep descriptions - but at some points you didn't get me quite right ... but I don't
want to dig to deep again into this - it costs to much effort for my poor english ... but it is OK - we all know more or less
what is happening there with that mechanical SR ... I think K-7 -5 and -r will have the same SR ... ?

kadajawi allready made some additional points I think of too, but I still wonder why you, lister, can judge SR is good while
panning ... Even very slow pans over longer distance than the SR-radius - always need the sensor to return from its outpost,
so every time you have a counter-clockwise movement that is ugly ... and there is no way to neglect that when you watch
closely ... you would need a good anti-movement done by a good android / robot that could manage to compensate that all
the way ... for the whole momentum that needs to be adressed ... kadajawi saw that too ..

But it is true we need an outer switch for the SR in the future (like it was in the past with K10D) because putting SR on/off in
the menue is no good choice really ... Or bring back the status menue at least for that quick on/off - just as it is offered for
the photo-modus ... It's a shame the K-3 can do so many things well now, maybe even this too ? I don't know. But the missing
SR for me is a sad thing too ... but I will go for better work-arounds now - as well with the soon to come new K-5.

About the VC (vibration controle) of that Tamron lense for Ca.Ni.Cons: It may look better because it only does compensate for
half of the radius of our Pentax-sensor (as my feeling says) That may be the reason for better real endings of pans. Would be
nice if there was a software that could rethink and change this counter-movement - but that double work-out seems to vast for
this price-range ? ... Pentax will surely be good advised to ponder on that synthesis of both corruptes SR's in the future ...

I allready ordered my first NDx16 filter - another x4 and x64 will follow later maybe ... I decided against a variable ND filter
because I read everywhere they do not behave safe and well enough about full sharpness and true colors and much more.

What I still want is a better tripod ... and if possible a tablet / palm-top / smartphone with HDMI-input to watch the image
live in complicated situations and positions ... There seems to be none you can easily carry with you - just bigger monitors.

[kadajawi]

I think fixing the algorithm/adding another option should be trivial. I'm not really a programmer, but I can think of a way how to do it. Basically you'd keep the normal SR system in the K-5 which seems to always move into the direction you have shaked the camera to (with a slight time delay and a rubber band). It is easy to get the camera to start the pan faster, you just begin panning with a small jerk and then continue slowly. Either the manual could say that such a movement indicates to the camera you are panning... think of gestures you can do on your smartphone... or the camera will notice that when the movement is relatively constant into a certain direction, say over 1 second, that direction will be locked on more or less (it can be adjusted over time). The camera will then slowly move the sensor back to the center along the axis of the panning direction (even if it is diagonal), and mostly compensate for any abrupter direction changes up and down. For the stopping part, the camera could see an abrupt change of speed as a stop, and slightly fade out. The sensor will be off center, which means the videographer will be advised to pan a bit to center it. You could also leave away the fade out. There are different ways of doing it, so there should be different options, with an info text explaining what it does.


I'd also like to see an overlay, just 2 squares, which tell you where the sensor is, and what it's maximum range is. That way you know how much further you can go, and you can adjust better to what the camera is doing. If we want to go further, there are a lot of buttons un-used when shooting video. Turn them into SR controls. There could be a button to tell the camera you're now panning, please switch to panning mode. Another to deactivate it. You could have one button that is used with one of the dials to control the strength of the SR. You can have a hold position by all means button, like AE-L, the camera will then remember the current position and try to hold it until it is impossible, for a perfectly stable, tripod like shot. Here the overlay mentioned earlier would come really handy, as the videographer will then know if they have to move the camera a bit to get the right position. It would also work great for shots on tripods where the tripod isn't 100% stable. They could implement a steadicam mode which combines mechanical and electronic SR to get really smooth flights, to get more latitude (the electronic SR would balance out the slow up and down movements that do not introduce jello and weird motion blur, while the mechanical SR is filtering out the fast bumps that will look ugly).


Will this have a steep learning curve? Yes. Basically you'd need to put it in the manual, and maybe an on-screen help system (perhaps when you activate the enhanced video SR which unlocks all these nerdy features). Most of these suggestions shouldn't be very hard to implement. Maybe they'll take a day, a week for a programmer... then some more time for testing. That's it. Some of your issues could even be fixed by changing parameters!

[kadajawi]

Found a great video shot with an Olympus. You want proper stabilization? Get an Olympus. Seriously.


The range of motion that the Olympus can correct is ridiculous. There are no distortions at the corner that my K-5 would do. There is no rotation (the K-5 can only correct this to about 2°, which isn't nearly enough). I recon with some practicing you can get that camera to look like you've used some proper rig... even a steadicam perhaps.

Also very impressive:
Olympus OM-D 5-Axis Stabilization Tested by a Man Suffering From Hand Tremors

Olympus is actually advertising and stressing that their IS is now used for video, and that it is better than the electronic IS they used previously. Sony too is getting the 5 axis system from Olympus, and I have no doubts that they will make use of it for video.

[kadajawi]

Review of the OM-D E-M1 by EOSHD:

It’s less noisy in low light and it has an extremely effective stabiliser which allows you to shoot video entirely handheld without a tripod. It is that effective.

If you then add Warp Stabiliser in post you can get a completely locked down shot from the palm of your hand. This is a MAJOR advantage over all other cameras, be it the GH3 right up to the Alexa. All those cameras need a rig of some kind. The E-M1 can get away without one. Also – there’s less of a crop in Warp Stabiliser because the shot is much steadier to begin with. In fact often there’s virtually no movement at all, not even a gentle sway. Jitter is effectively gone.

First look - Olympus E-M1 vs Panasonic GH3 » EOSHD.com

Pros don't use image stabilization, eh? The whole article is basically about how awesome the stabilization is. Pentax used to have it too (it wasn't as good as the Olympus, but it's pretty helpful already). Oh well.

[TomGarn]

I will keep an eye on Olympus ! This sounds really great !
Maybe it really is even better than GH3 - which is great news.
Btw.:
In the seventies I owed an Olympus OM1 - with two glasses: 35 and 100mm.
It was the smallest of all SLRs at that time - and I loved it !

P.S.
I will go on with my K-5 and forget about heating up hot-pixels - if possible ...

Last time had lots of trouble with aperture - ISO - shutter-speed when I did a small clip with it -
but which also showed my it can be used better once I get used to it ... So now I first press green
button - for getting the right f-stop ... then pressing AE-L to hold sensitivity and s.speed - and than
manually choose the right aperture - which is troubling me sometimes because the livewiew doesn't
show fading and rising of the changing brightness salways - in order to always give out a bright image
for controle mostly ...
It's a hard way to remember that work-around for me now
but if I learn that it maybe possible to controle the issue ... ?
What did you find out about it ?

[kadajawi]

I will keep an eye on Olympus ! This sounds really great !
Maybe it really is even better than GH3 - which is great news.
Btw.:
In the seventies I owed an Olympus OM1 - with two glasses: 35 and 100mm.
It was the smallest of all SLRs at that time - and I loved it !

P.S.
I will go on with my K-5 and forget about heating up hot-pixels - if possible ...

Last time had lots of trouble with aperture - ISO - shutter-speed when I did a small clip with it -
but which also showed my it can be used better once I get used to it ... So now I first press green
button - for getting the right f-stop ... then pressing AE-L to hold sensitivity and s.speed - and than
manually choose the right aperture - which is troubling me sometimes because the livewiew doesn't
show fading and rising of the changing brightness salways - in order to always give out a bright image
for controle mostly ...
It's a hard way to remember that work-around for me now
but if I learn that it maybe possible to controle the issue ... ?
What did you find out about it ?

Well... the Oly only does 30p, there is nothing else. No more, no less. Which is quite limiting. And unlike the 70 Mbit or so the Pana can do, you're stuck with a maximum of 24 Mbit. Basically all that speaks for it over the GH3 is stabilization, but that might just be enough.


Haven't tried what you are writting there. I usually just choose the aperture I want, let the camera do the rest and when I'm happy with the exposure I push AE-L. I wish there was a WB-L button too... that would be really useful. Setting WB by hand often results in results I do not want. Anyway, haven't tried letting the camera decide the aperture... could make sense. Hmm...

[TomGarn]

Mbit
In this article they said, that those 24 MBit are the same quality as GH3 's 70 in edit suite - I also
wondererd a little about that ... (I don't edit too much on image parameters anyway ... (So GH3
is closer to that broadcast minimum of 50Mbits ... good)

30 fps / 25fps
Well 30 fps is not our western European standard -so I don't understand that - same with K-7 btw.
Now I use the typical 25 fps in my K-5 - so a little more stuttering is there of course ... and we are
not used with that because we had 50i on tv for ages ... but well ... if the shutter speed doesn't add
to it, one may get used to that ... and if you have some footage you want to sell for broadcastusage
25 is better I think - or you would have to render it down from 30 - which is stupid and not very
practical as well ...

Exposure:
It is very important to decide for the right one of those 3 metering surounding distances. I will
try full radius now ... because the choice is to bright easily when you don't include the skyline while
looking for the right aperture. Mostly the camera-dispaly also is much to bright - and you cannot
easily turn it down visibly - it resists to show up and pull down the brightness ... or aperture. I am
sort of perplexed about this effect.
I usually want to know first what the camera thinks is right - and then decide for myself. And this
first decision is critical, because you cannot manyge to see the true light that comes in - on display -
and there may be that faster speed-shutter included allready - or ISO allerday steps up.
Later on you may see in the colour noise on dark parts, that the camera has even added automatic
ISO .... so I can only hope the green button - with my right grey-filter on ... (at least 2 f-stops less -
and in sun: 4) ... does the right exposure metering ... if you can understand what I mean) ...

WB
I took some time to manage this manual WB-measuring ...it's a little complicated and crazy, but
it is great they have 3 memory-positions for WB now on the K-5 - that is professional !
But mostly I use ordinary: cloud / shadow / sun .... because they really give you the true special
impression later on ... It gets difficult only on very blue skies, with warm sun in the same scene ...
and with different low artificial lights - but mostly fixed positions are best.

In Califonia I always had 7300 Klevin whenever I matched WB - that was funny so I still remember.
> Pro cameras show you the number in the display so you can know by that if your WB is done rightly.
That manual change of colours is great - but I would only use it with a real reference studio monitor.