hyalinejim

Ghost free! Big thanks for that @tomsemiterrific. Looks clean as a whistle on my end, no motion ghosting. If I did what you did with my cam, it'd look like we were all on mushrooms. .. @mercer or any other XC10 owner, can you check if your results are clean @20000 ISO like tom, or sick like kidzrevil and me?

It's barely noticeable at low ISO. Everywhere in between is more or less noticeable, depending on what ISO, how much motion, and in what tonal range. It's not that I ever intend to shoot at ISO 20000, it's just that it's incredibly noticeable and is the best test of "Does my XC10 have ghosting artifacts?" There are 2 possibilities: 1. Only some XC10s have ghosting and kidzrevil and I should return our units. 2. All XC10s have ghosting and we should ask Canon for a fix. All it takes is one sample clip to falsify no.2

@tugela This might be the case, but the ghost comes from the actual previous frame, not an intermediate frame. You can see that in the frame by frame video I posted a while back. I'm pretty sure that it's temporal noise reduction as it increases with ISO.

AFAIR the XC10 does have some interlaced modes but with interlacing there would be odd and even fields - you know, the horizontal serrated effect. I don't think that footage is interlaced but I shudder to think what interlacing + ghosting would look like Hopefully other XC10 owners can post ISO 20,000 motion stress tests so kidzrevil and I can decide whether we should send our units back to Canon or ask them for a firmware update. All it takes is one sample video that DOESN'T show ghosting in a situation where my cam will produce it and I'll pop my XC10 in the post tomorrow. But I suspect it's all units.

No idea, it's not my vid. Ghosting is independent of shutter speed though - SS only changes the shape of the ghost due to motion blur. I agree that in this case it's not the end of the world - a test video / tourist vid. But there are times when this sort of motion artifacting is flat out unacceptable. I would hate to be in a situation where a client points and says "What's that? Can you get rid of it please?" In short, no, because if you play back frame by frame 25p footage the traces are from the previous frame, not a 50p interpolation

What would you say the difference between them is in terms of DR - they must be quite close right? I mentioned the superwhites because when you switch to WideDR from CLog the zebra area increases if there is slight clipping. However, I think this is just because of the way CLog spreads out the highlights and the actual clipping point looks to me to be more or less the same from my very unscientific tests.

Yes at 1000 there is a slight smearing of the image in HD, but 4K shouldn't be affected as much. I didn't notice a difference between the two when it came to noise - both were at ISO500 so not that much noise anyway except for shadow areas. Yes, that video has very clear ghosting in some shots - like double vision! 43s, back of the man's head - or the lady. Or when the white truck goes past, pause it and you'll see 2 sets of tail lights!

I also find it hard to believe that other people have not spotted the image ghosting and that makes me think it might just be some models. Only way to know for sure is to see footage from other units. It looks like @tomsemiterrific and @mercer are on the case and will post something in the other thread In other news, more tests today confirm for me that the image is already losing detail at ISO1000 in HD. Also, WideDR is better than CLog in HD because there's less banding and the colours are nicer. CLOG: WIDE DR: It also makes it easier to see the screen. I don't think there's any difference in dynamic range. I also learned that there are superwhites and 100 zebras are not necessarily clipped - I can get that info back in 32bit workspace in After Effects.

@tomsemiterrific +1 I would love to see that flowerpot shot. If it's too much hassle to upload to vimeo maybe you could dropbox it or something. Regular Canon support want me to send my unit in as well.

I agree. I went into a store in town today and checked out their display model. It had the exact same behaviour as mine. I've sent the files from both cameras to my Canon CPS rep, because I haven't heard from the regular support guys in a few days. In the meantime I'm trying to keep ISO at 500 whenever possible. Interiors are manageable at 24mm f2.8. If the ISO goes up the image goes to mush and the ghosts start coming out. I'd like to see Canon release a firmware update on October 31st. If you want to chance sending in your unit, ask Canon for a loan replacement. You might be entitled to it under the CPS programme. If kidzrevil and I are correct in our hunch, this affects all units and not just ours. In that case the only solution is for XC10/15 users to contact Canon and ask them to offer an option to disable noise reduction. Can you check this @HugoS316 and upload a quick example at ISO 20000, waving the camera around in front of something sharp and contrasty with underexposed areas? If your unit doesn't have ghosting, this will encourage kidzrevil and me to send our units back to Canon. Thanks in advance.

No such luck @kidzrevil! This one was in HD:

Which one are you saying is the XC10 @hijodeibn? Top or bottom? It's slightly less sharp and less detailed and has more compression artifacts (HD on top, resized 4K below, look for sharpness on book cover text and look for compression mush under the table): Is it worth 10x the file sizes for HD delivery? Probably not at ISO500 for well exposed footage, but maybe at higher ISOs or if crazy post is required. I can see the difference if I toggle back and forth between them but that's serious pixel peeping. Now, for HD shooting it looks like if you shoot Wide DR instead of CLog there is less banding and less compression artifacts (top = WideDR, bottom = CLog, contrast expanded) This is obviously an extreme push here to show that there's less banding in WideDR. However, it does make a noticeable difference in a real world grade, particularly if sharpness is applied to the CLog footage. Dynamic range seems to be about the same with regard to highlight clipping point. If you switch to WideDR the zebras increase but the info is still there in the superwhites - just not as gradual a roll off into white as with CLog. There's also a big difference in colour rendition between WideDR and CLog which I'm going to investigate more fully in daylight. It looks like Wide DR might be nicer.

Time for a game of guess the camera! Which is XC10 and which is Magic Lantern RAW? (I've downsized these so resolution advantages aren't apparent.) If you need a hint take a closer look at the camera strap. This is good news for me as it means I can match the two closely enough. The 5D3 was shooting in crop mode so DOF is more or less the same. Interestingly there is a difference in exposing the two cameras. I have to close down the 5D3 by about one stop to reach the same clipping point as the XC10 when they are set to ISO 400 and 500 respectively. In other words, if I used the 5D3 to set exposure and ETTR, and used the same settings on the XC10, the XC10 would be one stop blown out in the highlights. I'm not sure what the implications of this are.

@UHDjohn Good to know that zebras are on green! Thanks for that tip. I haven't seen much banding in blue skies. Will look again. If you'd like more customisable buttons and control over NR it might be a good idea to contact Canon support. They have an option for feedback / criticism. They've already shown they're responsive to improving the XC10 through firmware updates.

If true @HugoS316 that would be a massive relief because we can hopefully get our cameras repaired / replaced. Would you be so good as to record a few seconds of high ISO movement like I did above? @tomsemiterrific do you see ghosting on your XC15? @kidzrevil Did you register your camera on CPN? If so, they say there's a 5 day turnaround for returns.

I'm going to guess no. You could try plugging it into your TV to see what you get. Would an external recorder see the same?

Yes, and if you hold it at waist height with the screen flipped up it looks like you're just figuring out how to use it, or fiddling with some settings. Looking forward to the results @kidzrevil.

Some landscape shots from a country walk yesterday. 4K CLog ETTR at ISO 500 to 1000. FilmConvert Kodak 5213 Vis3. Very pleased with sky v foliage v skintone colour separation using FilmConvert on XC10 footage.

No, this was HD so no RS reduction. It's definitely not post because I can see it on the LCD, and in the file SOOC.

Very good examples kidzrevil. I'm also in contact with Canon support about this. @mercer I see it most at high ISO. If I whack up the ISO to max and wave the camera around a bit I'll see it very clearly on the LCD screen. Nothing else matters, not IS, picture style, shutter speed... nothing. As far as I'm aware if there's movement within the scene in a high contrast area there will be ghosting in the shadows. I'll try to upload a vid of what to look for. EDIT: Here it is. This is max ISO just to show the problem. IS is off. Shot in CLog and contrast expanded in post. Try to keep your lunch down!

Here's a 4k 305mbps frame grab of a scene with deep DOF and a lot of fine detail at only ISO 1000: Here's a crop of the top right hand corner at 100% Look at those rocks and shrubs. Pure mush. Here it is again resized to 1080: It's still fucking mush!!! There's no ghosting here as the camera wasn't moving. It's spatial noise reduction (within the frame) as opposed to temporal noise reduction (between frames). Both are perfidious and noise reduction should really be user selectable. Now, I'm sure the aggressive NR is helping the compression algorithm BUT... if I could switch it off I would. I'd also love to hear from other XC owners about whether they see aggressive noise reduction and motion ghosting in their units. Because if it's just my unit and kidzrevil's I'm sending mine back to Canon pronto!

Check out http://www.nextodi.com I use one of their older units (ND2730) so I can swap 256GB CF cards while shooting MLRAW. It works flawlessly except that it really needs to be plugged in to AC power after the first 256GB transfer (which takes around 40min). Newer models may have faster transfer speeds and longer battery life. Anyway, if you need a unit that accepts a variety of cards it might be worth a look.

@Tim Sewell Forgive me for mentioning it if you already know this: FilmConvert will blur your footage at anything less than the 35mm sizes to simulate the softness of smaller film stocks. If you like the size of 16mm grain but want to maintain sharpness then pick the next size up. I'm very happy with XC10 + FilmConvert. Will upload some nice landscape stuff later. Also I've found a great example of spatial noise reduction killing detail at low ISO. Hold on to your hats

@kidzrevil Thanks for making this thread. Canon don't know yet that this is a big problem. On page 76 of the manual they write: "When using Dynamic IS, the edges of the picture may be adversely affected (ghosting, artifacts and/or dark areas may appear) when compensating for a high degree of camcorder shake." BUT the ghosting appears whether or not IS selected or not, and at all ISOs. Now, my video shows faint ghosting artifacts at ISO 500, and kidzrevil's high ISO frame grabs show severe ghosting. So it gets worse with ISO. So it's probably due to temporal noise reduction. Forget frame grabs for a second: it's visible on the LCD at moderate ISOs as soon as anything moves. This makes the footage next to useless. I don't know how this slipped past Canon and the BBC guy. If you only shoot static scenes everything's hunky dory. But look what happened when DVInfo panned across a chart: Adam Wilt knows it , I know it, kidzrevil knows it, but... Canon doesn't know about the ghosting yet So if you own an XC10, contact Canon support and tell them your camera does mad shit against your will. They've shown their willingness to improve image quality through firmware updates. Ask them to fix the ghost that is haunting your camera. Because if your scene demands you record at anything above ISO 500, you might as well shoot it on a potato.