-
DV100 vs DNxHD vs CFHD
-
The lens is the first limiting factor the light meets as it enters the camera. It’s vitally important that it passes exactly the right level of detail too – not too little (or it’ll be blurry) or too much (aliassing artifacts). So the resolution of the chips, their type and if anything like PixelShift is being used are all factors in how much detail the lens needs to pass.
Graeme
– http://www.nattress.com – Film Effects for FCP
-
What I meant was – are you saying that the type of lens they are going to put onto the HVX wont make any difference anyway as the lens isnt up to it?
I have a digial SLR and while Im using the same detachable lenses – i can tell you that CCD pixel really does make a difference, granted we’re talking bigger leaps in resolution on digistill cams – but to say that capturing the full res would make no or little difference is only correct if the glass is soing to be whack – which I hope it isnt.
-
The lens has to act as an anti-alias filter, so that the image sampled by the CCD doesn’t alias and create artifacts. For pixelshift to work, you need to create some aliassing so that it can be anti-aliassed as more real detail is created by the pixelshift algorithm. I’m guessing that Panasonic will make the lens just good enough for the CCD to function properly, and that to keep sensitivity of the chips up, they won’t have any more pixels than they need, and so the camera won’t generate any detail beyond what the codec is capable of, and hence capturing the image uncompressed ahead of the codec won’t add in any extra detail save reducing any artifacts from the codec. If they were to put a “better” lens on the camera, and the chips were high enough resolution for full raster, there is a point whereby the lens would be too good for the chips and produce aliassing artifacts which would be objectionable. It is the camera designers’ job to balance these factors for an intended result.
Capturing uncompressed on F900 or Varicam makes sense as the chips are more detailed than the codec allows, hence the codec can deal with super-sampled information, and the lens choice would be made so that it works with the chips and resolution detail of the camera, not limited by the codec.
Graeme
– http://www.nattress.com – Film Effects for FCP
-
Thanks Graeme,
“there is a point whereby the lens would be too good for the chips and produce aliassing artifacts ”
Ok I believe you 🙂
as a mater of interest, if the lens is created in tandem with the CCD to stop aliasing aritfacts infront of a higher res CCD – then how does this work with a digistill cam? I can take the lens from my 300D and slap it on my 350D with 2Mpixel more resolution – no specially designed compromise here between ccd changes and lenses; and no aliassing artifacts either.
Im not saying your statement isnt true, just trying to understand how it happens only in video.
-
I think you’re missing the point Graeme.
I’ve used both DVCProHD from firewire (1200A deck), and footage ingest uncompressed to Blackmagic 10-bit.
There is a WORLD OF DIFFERENCE between the footage. You can argue all you want on the technicalities of this process, how you’re editing with that native codec, etc. but once you ingest via firwire the native DVCProHD codec signal, the damage IS done. Frankly, there’s something about ingesting a base-band signal that is much higher quality than the native DVCProHD codec. Maybe it’s the way that Apple decodes the codec compared to the hardware decoders in the Panasonic decks, but the footage is simply NOT THE SAME. This is the same concept behind the differences of working with firewire DV, and digitizing DVCAM off SDI via a high-end DSR deck to a 10-bit 4:2:2 format. There’s a big difference in quality with the SDI signal.
So, if you’re ingesting the HD-SDI stuff via blackmagic, etc. you’re now doing uncompressed 10-bit. Big files, and a lot of storage. That is where intemediate codecs come into play like Cineform or DNxHD. They allow you to work with the base-band HD-SDI signals with real-time effects, etc., not being constrained by the bandwidth and processing requirements that moving around multi-stream uncompressed HD would require. Furthermore these codecs are higher quality/more efficient than DVCProHD, they are full-raster, 10-bit precision, can survive multiple passes, etc.
Ingesting the native DVCProHD signal via firewire, adding effects, and then rendering out uncompressed does nothing for your quality in improving how the DVCProHD signal was decoded in the first place. When you do this, all you’re doing is preserving the artifacts of the original decode, whether that is banding, blocking, etc. I’ve tested and I’ve seen with my own two eyes that the software decode of these native DV codecs, whether it’s DV, DVCProHD, etc. is not as good visually as the hardware decoder required to output a base-band HD-SDI signal.
So, there is a big use for intermediate codecs, in that working with native DVCProHD is not that wonderful a solution for really high-quality work, especially post color-correction, etc. Digitizing off HD-SDI is, but there is a definite workflow bottleneck when you’re trying to move around uncompressed footage. Intermediate codecs allow you to re-definte the HD-SDI workflow by having the benefits of the HD-SDI stream decode process (and it’s quality improvements) and the flexibility that editing with the native codec allows, but now with visual quality that matches the uncompressed HD-SDI workflow.
For P2 this process is totally irrelevent, since there is no HD-SDI decode process. You already have the native codec on-disk ready to work with. But if you ever want to work with HD-SDI uncompressed for the highest quality possible, intermediate codecs are a huge boon.
BTW, when Qrez is released by AJA, that should make for a very nice intermediate codec in FCP compareable to CFHD or DNxHD.
Jason Rodriguez
Virginia Beach, VA -
I’m starting to get out of my depth here with my knowledge of such things. For one, Digital Cameras use a single chip with a Bayer pattern, wo things are not as simple as they are with 3CCD video cameras, and I have no idea on the relative resolving powers of the lenses in question. If anyone here has more insight into the optical block and lens of a video camera in relation to filtering to stop aliassing on the CCD, I’d certainly appreciate some insight.
Graeme
– http://www.nattress.com – Film Effects for FCP
-
Jason, I’ve had discussions about this over DV over SDI and DV over Firewire, so I set up an experiment. I got a friend with some pristine 35mm footage, a digibeta deck and a DVCAM deck with SDI and firewire to help. We used the 35mm footage, telecined to Digital Betacam as our reference, dubbed it over SDI to DVCAM, and then brought it into FCP via uncompressed over SDI and natively over Firewire.
Then I compared them. The luma on the SDI and Firewire (after decoding with the Apple DV codec) was identical in every way down to a fraction of a percent. Comparing back to the uncompressed capture over SDI from the Digital Betacam, it’s impossible to say which of the lumas from DVCAM was better, just that they were ever-so-slightly different, by about 1 LSB in places.
Next I checked the chroma. As you know the Apple DV codec does not smooth chroma to improve it’s generational performance for compression back to DV. An SDI output will smooth the chroma to 4:2:2, just as the analogue outputs smooth the chroma. By applying a chroma smoothing filter to the firewire footage in an uncompressed timeline, we could compare them both as 4:2:2, and again, the difference was negligible.
The only way that DVCproHD will look any different over firewire as it does over SDI is if the software DVCproHD codec is inferior in it’s implementation compared to the hardware one. Are you saying this is the case??
Ah, Qrez, the hardware codec. Pretty useless if you want to move the footage to another machine that doesn’t have a Kona2 card in it??
Graeme
– http://www.nattress.com – Film Effects for FCP
-
The overlay is not full res.
-
Something’s different, because the files definitely do look different.
Try doing some green-screening sometime with DVCProHD over HD-SDI and then over the same footage on Firewire.
You will notice a difference, especially on challenging footage.
Seems as though the software DVCProHD codec can suffer horribly from macro-blocking in the shadows and saturated color gradients. I hate color-correcting DVCProHD footage for these reasons, there’s not much you can do before you see some pretty terrible-looking stuff, if you’ve under-exposed the camera a little too much. I’ve talked to Pana about the problem, and they acknowledge that much of it has to do with the 8-bit precision and compression on the codec not being able to really take advantage of the full dyanamic range that the FILM REC mode on the Varicam can deliver.
BTW, for a quick visual comparison between DVCProHD and CFHD, go over to one-river media and look at Pixlet 100% vs. DVCProHD. Ignore the white-count, because CFHD has a very, very low white-count (the gradients and other areas that Pixlet does bad in are perfectly preserved with CFHD). But they look about the same visually, with the edge going to Cineform. And again, their white-counts are completely different (the white count of CFHD is close or very comparable to Apple’s 10-bit 4:2:2 Uncompressed, so again, it’s a very high-precision codec, and way beyond the other 8-bit codecs that have very high white-counts).
Jason Rodriguez
Virginia Beach, VA -
Yes, you must have Prospect HD and Premiere Pro 1.5.
You can now purchase Prospect HD stand-alone for $1,500, but it needs a 3.4Ghz P4, and that is without the ingest/output card (Aja XENA HD).
With HD-SDI card, you’re talking around $6,000 with the Adobe Video Bundle and Prospect HD. Also for ingest/export you’ll need a much beefier machine, basically a dual Operton 252. Since you don’t need a drive array, once you’re all done, figure under 10K for the whole package.
Go over to http://www.cineform.com and click on Prospect HD for the different choices you can make.
Jason Rodriguez
Virginia Beach, VA
Reply to this Discussion! Login or Sign Up