Too much blur with TRV950 when making very small movements...
 

I just filmed a bit with a sony trv950e and this is the result (i just pressed record, no special settings)

http://users.skynet.be/m2/trv950e.avi

As you can see, without too much shaking the image is very very sharp, i notice almost no noise at all. BUT when i pan the camera a big strobing effect shows up and this is completely useless for even simple cameratracking. (the reason why i need a good camera)

Does this motionblur (well, i realy didn't move fast at all) be reduced with some settings on the camera ? (i didn't have the time to test everthing ofcourse) Or do i have to film with another method?

Will this blur be less with a vx2000e??

Help me out guys, i realy can't make a choice between these 2 ;-(

Clear frame:
http://users.skynet.be/m2/1.jpg
http://users.skynet.be/m2/3.jpg

Blur (destroys a good previous track)
http://users.skynet.be/m2/2.jpg
http://users.skynet.be/m2/4.jpg


am very satisfied with this imagequality, but that blur makes this camera (at the moment) useless to me...

PS1: anyone knows where i can find a movieclip of a traveling with a trv950? that would convince me to buy it, knowing it's possible to make clear shots)

PS2: Sound is NOT important at all for my camcorder! Imagequality must be very good, that's all)

Tom, you get pretty much natural results, that is what one could expect from a camera in full auto mode.
Unless illumination is very strong, camera tries to keep shutter speed at 1/50s for PAL or 1/60s for NTSC. At this relatively slow shutter the motion blur is very visible.
To avoid it you can chose either sport AE program or manually set shutter speed to something faster, assuming you have enough light to do this.

does a faster shutter speed needs a higher number or is it the inverse? (sorry am new to this)

higher speed is like 1/100 or 1/25? (am using PAL)




thx for your clear answer Vladimir, very much appreciated.

(i realy hope i can get rid of this blur, it's a big problem for me)

You should also try turning off the image stabiliazation (steadyshot). When the camera is stationary and something is slowly moving, or if the camera is slowly moving, the steadyshot may try to compensate and create problems.

<<<-- Originally posted by Tom de Boefer : does a faster shutter speed needs a higher number or is it the inverse? (sorry am new to this)

higher speed is like 1/100 or 1/25? (am using PAL)
-->>>

1/100 speed is higher. However, on TRV950 screen it's displayed as "100". So looking on the screen a greater number means faster shutter.
As a matter of fact, motion blur gives your video more natural look. If you chose very fast shatter, like 1/1000, you can get each frame much sharper, but the frame sequence would look stroboscopic.
Also, with such a fast shutter you need a lot of light for proper exposure.

You had a very light motion blur in that movie and personally it
added to the scene! Why are you so desperately trying to get
rid of it? For some scenes it might be better to use a high or
perhaps even a lower shutter speed.

Without knowing what your exact reasons are it is harder for
us to answer your question. But the answers given are good
ones to get "rid" of your motion blur

One piece of advise to you. Switch your camera over to manual
and try what each settings does (like shutter, iris, gain levels
etc.) with and without motion in the frame. Record it and write
down which settings you used for each run. Watch them on your
TV and make notes

i need this camera for cameratracking and cameratracking is not possible if there is lots of blur. (i need it only for PC video editing)

Then what you want to do is select a high shutter speed (something like 1/600th sec) and an aperture to give you correct exposure in the light you're filming in. Probably f4 outdoors.

This will give you very jerky pans and zooms if you watch the footage as a pure movie, but it sounds as if you're only after sharp frames. Why jerky? because at 1/50th sec you record everything that happens in front of your camera. At 1/100th sec you only record half of everything that happens, and the missing information makes the film look jerky, stuttery.

A small point re the OIS Boyd. You say "When the camera is stationary and something is slowly moving the steadyshot may try to compensate and create problems". Just to point out that I haven't found this to be so as Sony's OIS is purely camera motion sensitive and ignores the image totally. Which is why it works so well in low light situations off course - unlike EIS systems.

tom.

<<<-- Originally posted by Tom Hardwick :
Sony's OIS is purely camera motion sensitive and ignores the image totally. Which is why it works so well in low light situations off course - unlike EIS systems.

tom. -->>>

Tom, this is an interesting bit of info.
My understanding is that the main difference beween OIS and EIS is a way the motion is compensated, rather than a way how it's measured.
I thought they both use some sort of acceleration sensors to determine a camera shake. Once the shake is measured, OIS moves some optical element to compensate it, while EAS moves read-out window across a whole CCD area.
Is this right or there are more differences between the two?
Is there some web page detailing their operation?

When Sony's OIS system is on there are sensors permanently on standby in the camera body waiting for you to move the camera. If you *never* move the camera (on a tripod say) then every so often the OIS can give a tiny "wiggle" - sort of like taking a little breather. This is why you should turn OIS off if the camera is locked down. IF you have an imperfect fluid head (and many of them exist) then you would be well advised to leave SSSS on because it'll smooth out the irregularities of the tripod head friction.

Some EIS systems don't need to know if you're shaking the camera, they will still endevour to stabilise the image. This can be useful for film to video transfer work where the camera shake on the projected cine film can be stabilised by the EIS. This is very easy to demonstrate. Lock down the camcorder on a rock-solid tripod but have the EIS turned on. Point it at the cine screen and notice how the TV picture is steadier than the original film. But beware - you have to crop the original image quite severly for this to work. Another thing. EIS should *most certainly* be turned off for tripod work as it will assume subject movement (a fidgety interviewee, for instance) is shake that needs to be "corrected".

So yes, EIS is simply looking for the same info in successive frames and locking onto this. EIS can be made very powerful indeed, but the more powerful it's made the more sticky/jerky are the zooms and pans. EIS requires good light in which to work - and not only good light but decent contrast. OIS will smooth out your footage on a foggy day whereas EIS will give up completely.

For now, OIS is the winner, but its days are numbered. It's a bulky, noisy, power hungry, expensive solution to the problem, and the hybrid EIS systems are catching up fast.

tom.

Tom, thank you for the reply. I did not know that EIS system do not use acceleration sensors.
One strange thing remains though. TRV950 uses just a fraction of CCD area for video. One could assume that periphery of CCD is used for some form of image stabilization, even though it is supposed to be optical. Otherwise I can not explain why they waste this space. Can it be that TRV950 uses some combination of OIS and EIS?

That's interesting. When I first got my PDX-10 I had it locked down on a tripod with no movement, at full telephoto, shooting a full moon in the sky. When I played the tape back, at one point the the moon "jumped" a bit. Then I realized I'd forgotten to turn off steadyshot. After turning it off there was no problem.

TomH's advice re shutter speed is spot on in motion tracking work if you use Boujou whatever


the problem as he mentions is that at high shutter speed the motion is very stuttered since you can see two crisp images in each frame

the workaround for this (assuming you shot with tripod and pan) is to post some motion blur into the footage after you have done the tracking

after effects can handle this - you should use a value of motion blur in the direction of motion equal to the displacement between the images you see in each frame (fill the gaps with motion blur) to make it look as if it was shot at normal speeds 1/50 or 1/60 so that the fluid look is restored

that's good to know, when i have my 950 i'll think about it ;-)

The TRV950 and many modern camcorders have to have the words "mega pixel" emblazoned along the outer panels just to survive in the market place. This can be quite useful as Sony have found out. The PDX-10 makes use of this mega-pixel chip to give true wide-screen 16:9 and at the same time record much better stills than the 900 before it.

So the chip area isn't "wasted" Vladmir, it's used for stills, 16:9 (in the DVCAM version) and for the camera advertising.

OH, and OIS is never only optical, it's a combination of electronics, mechanics and optics. It's a hybrid system that uses a vibrating front element (on the VX2000's VAP) or vibrating internal elements (of the TRV950). These are pushed and pulled mechanically by information fed and processed electronically.

In contrast, EIS systems are purely electronic. Silent, cheap to manufacture and far more compact.

tom.

Tom, yes, I understand the area that "wasted" for video is used for stills and 16:9 mode (in PDX10).
I just thought that designwise for Sony it's more beneficial to use full CCD surface in 4:3 video mode. One could get a better low light and a wider wide-angle view. Both are not very good in TRV950. I assume one needs an interpolation engine to interpolate the pixels down from megapixel to the normal TV resolution. But they already have this engine because of digital zoom, at least in TRV950. So, essentialy, this interpolation comes for free.
This is kind of rhetoric question, as only Sony could answer this for sure. It would be a nice surprise for me if anybody else can explain this.

I like your argument but don't know enough about CCD technology to add anything other than conjecture. The TV screen is looking for a 720 x 576 rastor and generally bigger CCD pixels give greater low light performance so using double this amount (say) would reduce the individual pixel size to slightly less than half.

I do agree that the TRV950 has a very dissapointing wide-angle though - in fact I'd say it was contravening the Trade's Description Act to call it a wide-angle at all. It equates to a 49mm lens on a 35mm SLR and the camera cries out for a w/a converter even before a spare battery.

Not that this is so bad though as the TRV60 and lots more in the Sony current range have a 52mm focal length equivalent! Can you imagine even less w/a than the 950? Remember too that the TRV900 had a decent (almost) wide-angle of 41mm equivalent.

Of course these mega-pixel chipped camcorders do offer significantly more wide angle when used in the stills-to-Memorystick mode but this is more a by-product of the bigger chips than design intent.

One nice thing about shooting movies onto mega-pixel chips is that as the lens coverage is designed for a larger "gate", when you shoot movies you're only using the central part of the lens' image circle. This generally means that vignetting is less obvious at wide apertures and picture sharpness is improved into the corners of the frame. Soon as you shoot stills on the thing though all these advantages fly out the window.

tom.

Well, this thread has gone a bit away from the original question, but has opened some very interesting issues.

Just a small explanation about CCD: CCD is an acronym of "Charge Coupled Device" or how we called it 30 years ago: "Bucket Brigade". CCD is essentially an analogue shift register and mimicks the firemen bucket brigade that used the buckets to bring the water to the fire. The firemen have been standing in a row passing the buckets to each other. So the firemen have been standing on fixed point and the buckets travelled fast to the fire.

Same thing is happening in CCD. Each photo sensitive cell (called pixel - picture element) on the CCD chip represents a fireman. The cell accumulates a small electrical charge during the exposure time. After the exposure the shutter shuts down and the the electric charges are shifted through the shift register to the ouput of the CCD. It is important to know that the output of the CCD is analogue. This analogue signal is then converted into digital form by means of an analogue to digital (A/D) converter. The digital signal is afterwards processed in the DSP (Digital Signal Processor) and recorded to the tape or other storage device.

Knowing this we recognise that it is not essential that the pixel count of the CCD resembles the spatial matrix of the DV (720×576 in PAL). CCD can have lower pixel count and can still provide the analogue signal to the same A/D converter. Of course the quality of the picture will degrade with decreasing of the pixel count. Theoretically there would be enough to have 414,720 pixels to resemble the full quality DV picture. And most of todays better DV consumer/prosumer cameras include three CCDs each having 430,000 to 470,000 pixels (Sony VX2000, TRV900, Panasonic DVX-100E). Such a CCD can produce very good picture and still maintain good low light sensitivity by keeping the pixel size as big as possible.

But there is another effect caused by the sampling of the analogue signal. Sampling theory explains that it is possible to transfer only the frequencies up to half of the sampling frequency. And in the vicinity of the half sampling frequency the sin(x)/x function distorts the frequency response by lowering the high frequency level. That means less detail, worse colours etc. By increasing the pixel count this effect becomes less evident. Ideally the CCD pixel count would be about 1 MByte, but doubled only in horizontal direction (1440×576 pixels). That would double the sampling frequency and minimize aliasing (another nasty effect of sampling). Of course that would require much faster and complex (more expensive) electronics and nobody is doing that today. Besides, DV coding is not so simple. It utilizes 4:2:0 colour space (PAL) and spatially compresses the signal in order to fit it into 25 Mbps stream. So the coding algorhytm itself prevents the DV coded signal to be better. But still, higher pixel count above "theoretical level" of 720×576 definitely reduces sin(x)/x effect and produces better picture.

This is very evident in "megapixel" 3CCD cameras like TRV950, which produces astonishing sharp and colourful picture (when exposed correctly), using about 690,000 effective pixels per CCD. Due to the coding limitations of the DV signal there is no reason to increase the pixel count beyond that figure though.

One-chip cameras shoul be limited for the same reason to 1,5 megapixels size (red and blue colours are only sampled by half the sampling frequency).

Soooo, putting a "Megapixel label" on the camera is only partly a marketing trick, it really enhances the picture quality. Of course it also increases the still picture quality, but that is not why serious videographers are buying the camcorders for. More important would be to keep the CCD size equal or bigger than 1/4" for a consumer and 1/3" for a prosumer camera in order to increase the low light sensitivity.

Higher pixel count is usefull only when using a generic 16:9 format, like Sony PDX10. It is a pity that TRV950 doesn't use the same technology.

Nikolaj

Nikolaj,

This is my first post to this forum. I have carefully read and re-read your post regarding oversampling and your reference to the fourier transform and frequency domain. Excellent post.

I had read a less detailed post on over-sampling and noise reduction wrt to the TRV-950 and that led, ultimately, to my purchasing the unit.

As long time signal processing guy I really appreciate your detailed post and your specifying the max useful resolution for 2X oversampling (and noting that nobody will do it because of the electronics issues).

I assume you work in the video image processing industry?

And to the reason that this is my first topic to post on:

I had a similar blur problem making a moving video of my son on his new roller blades Christmas morning with the TRV-950. The video came out blurry even when he momentarily stopped and looked at the camera. I have owned the TRV-950 for 5 weeks now so it is still new to me.

I had sharpening turned all the way down (having discovered how much noise the SONY sharpening function adds to the image). Also, I was not able to replicate the problem later with or without sharpening and the blur problem was haunting me until I read this thread.

In future I will TURN OFF THE STEADY SHOT. I just tested panning with it on and off and the steady shot, when panning, does add blur.

Great forum.

I'm most surprised to read that the OIS of your TRV950 adds blur to the image Mike, and I'm tempted to say that you may well have a fault with your unit. I've just tested the PDX10 and agree with Sony that for all intents and purposes the Super Steadyshot can be left on at all times, and only turned off if you want to conserve battery power. Unlike other manufacturers, Sony only give viewfinder warnings when the OIS is turned off, it's that transparent a technology.

Maybe you could explain in more detail what you mean by, "...does add blur". You mean the image takes on a soft focus look that isn't there when the SSSS is turned off? In theory, if a steady pan speed is maintained the floating elements will move to their bump-stops where they went when the pan was first started. When the pan decelerates to a stop the elements will again centralise in the optical path awaiting the next camera movement. As far as my tests show, having these elements out of alignment doesn't effect the resolution, but then again,in any interlaced TV system, subject or camera movement halves the vertical resolution at a stroke.

tom.

I started a thread a few days ago about tracers, or, ghosting, caused by overexposure, with my PDX10. This wouldn't happen to be the same thing as your "blurring", would it? Try adjusting not only exposure, but also color levels as well and see if this affects the phenomenon.

Great post Nikolaj. Yes, the CCD is an analogue device. However somewhere in the camera, probably just after the AD converter, the images coming from the CCD array must be moved in real time into some kind of DSP that can do the downsampling. Presumably the video cannot be moved fast enough through this bus if the full pixel count is used. This would explain why the PDX10 can use the full size of the CCD for still pictures but not for 60i video... but it does seem to be able to use the full width, at least in my NTSC version.

This would also explain why this kind of camera, even having a high pixel count CCD, will need a lot of changes to it's circuitry and not just a realtime MPEG2 encoder in order to record in the 'almost HD' resolution of HDV.

Unfortunately the PDX10's circuits cannot be upgraded easily so this camera will never do HDV in it's current form. I wonder sometime why we put up with what they sell us... can you imagine paying $2k for a computer that will not be able to write to next year's Office? More on this in another thread...

Tom,

Not sure what happened during the "blur" that occurred with my son's skating back and forth. Disturbed me for sure. Am testing my autofocus and will start a separate thread for a question on that.....could be I have a problem with my auto-focus system.

Shawn,
I definitely do get "blooming" or ghosts around white objects that are zebra striped. Have to take the exposure down a bit and they go away. Little surprised by this (and other features I am learning about).

In a separate post I will ask folks who have a TRV 950 to run a simple test and let me know what their result is.......