First shoot with HVX?
- Thread starter Jim Arthurs
- Start date
khmuse
Tech Guru
I have been following this somewhat painful thread and feel that poor Juan (pertierr) may have reached his end.
For Stevet,
The act of quantifying (sampling) an analog signal is to make it into a series of numeric values, hence, digital. The subsequent delivery format (DV25, DVCPro, HDV, etc) is the result of a signal processing stage (usually either a general purpose DSP or an application specific ASIC).
The analog output (composite, Y/C or color difference) of a digital camera is from a device named a video encoder (often a 656 format encoder). This encoder is in essence a digital to analog converter (plus a few filters and a DC restoration stage). Therefore, the analog output of a digital camera lacks the digital signal processing stage used to create the format output, but is still no more accurate (and in fact less so due to the additional conversion steps) than what is possible from the video encoder, and since these are not intended to be primary signal paths, they can be lacking in quality due to their execution.
Expecting that the analog output requantized to a higher data rate digital one being higher quality than the primary digital output is not terribly realistic as the quality of the encoder precludes this possibility. Camera manufactures simply don’t put their fiscal resources into aspects of the camera that would provide little benefit to the average user.
For Stevet,
The act of quantifying (sampling) an analog signal is to make it into a series of numeric values, hence, digital. The subsequent delivery format (DV25, DVCPro, HDV, etc) is the result of a signal processing stage (usually either a general purpose DSP or an application specific ASIC).
The analog output (composite, Y/C or color difference) of a digital camera is from a device named a video encoder (often a 656 format encoder). This encoder is in essence a digital to analog converter (plus a few filters and a DC restoration stage). Therefore, the analog output of a digital camera lacks the digital signal processing stage used to create the format output, but is still no more accurate (and in fact less so due to the additional conversion steps) than what is possible from the video encoder, and since these are not intended to be primary signal paths, they can be lacking in quality due to their execution.
Expecting that the analog output requantized to a higher data rate digital one being higher quality than the primary digital output is not terribly realistic as the quality of the encoder precludes this possibility. Camera manufactures simply don’t put their fiscal resources into aspects of the camera that would provide little benefit to the average user.
Mr. Blonde
RED Advocate
Well, the shot is quite bad but still, you can't pass judgement on the HVX from one frame grab, end of argument.
khmuse
Tech Guru
Stevet said:
Hey Steve,
I didn't want to imply that you were advocating that analog outputs were higher quality, but I suspect that some reading this thread might jump to that conclusion. I agree that it is possible (and may be probable) that DCT hasn't been applied to the data that is used to generate the analog video output, but without knowing the structure of the firmware, I can't say with certainty.
A concept that may have not be discussed is that, depending upon the point where the analog video encoder stream is derived, it may still be subjected to a fair amount of signal processing and would be limited by the data path width and thus would exhibit artifacts such as quantization noise and offset errors. Things like PAR compensation and gamma curves are applied to the analog outputs (since they are intended to be used as a monitor) which are derived from the in camera signal processing stages. DCT may be absent, but the analog outputs are by no means a pure analog signal path.
Zig_Zigman
Well-known member
Well, it IS a video camera...you were thinking it should look like super16mm?
Anhar_Miah
Veteran
Today’s lesson: Analogue and Digital
Analogue
A continuous signal over a given time period.
Digital
A signal composed of only two states namely "on" or 1 or "off" or 0, this convention is known as binary or Base 2 (mathematical notion).
"Today’s lesson was brought to you by the number 1 and letter D"
# doodadoododo, Can u tell me hooow?, how to get to Sesame streeeet!#
Chill guys
Cheers
Anhar
Analogue
A continuous signal over a given time period.
Digital
A signal composed of only two states namely "on" or 1 or "off" or 0, this convention is known as binary or Base 2 (mathematical notion).
"Today’s lesson was brought to you by the number 1 and letter D"
# doodadoododo, Can u tell me hooow?, how to get to Sesame streeeet!#
Chill guysCheers
Anhar
khmuse
Tech Guru
Stevet said:
It can not be after as this is a digital process that can not be applied to an analog signal.
Since the HVX has yet to be released, we can only speculate as to its topology. However, if we look at the structure of the DVX as an example, we can make an educated guess as to how the HVX might be configured.
In the DVX, the 1394 interface is physically near the analog input/outputs. I will look at a schematic for the DVX tomorrow, but from memory I can tell you that the analog interface is derived from the same data stream as the 1394. If you are further interested, I can provide you with a link to the Si used for this aspect of the I/O of the DVX. In the case of the DVX, the analog outputs are downstream of the DCT compression stage. There is little incentive not to use the same data stream as it would add costs and since this is only intended as a monitor output, would provide little added value to most users. The same is true when the DVXs analog inputs are used, they encode to the same format as the 1394 interface. In fact, most of the Si on the market is bi-directional A->D and D->A when converting from/to analog video to/form 1394.
Perhaps the HVX will differ, but until I look at a schematic, I am not going to make a definitive statement one way or the other.
Dmitry Kichenko
Well-known member
Getting back to the original topic of the discussion *sigh*, I must say I didn't like the screen grab at all.
If the camera is moving and that's just a single frame out of the whole clip that's fine. Otherwise if it's a still shot, that's a rather strange choice of the camera angle.
Then, the key doesn't look too good to me either. Going for the "fake reality look of Sin City" sounds like a cheap excuse to me, personally. Heads of the characters seem to have a blurred edge to them, which looks like a not so succesful key. Indeed, a capture out of the raw footage would be much more preferable.
Once again, this is criticism towards the shot, not the camera.
If the camera is moving and that's just a single frame out of the whole clip that's fine. Otherwise if it's a still shot, that's a rather strange choice of the camera angle.
Then, the key doesn't look too good to me either. Going for the "fake reality look of Sin City" sounds like a cheap excuse to me, personally. Heads of the characters seem to have a blurred edge to them, which looks like a not so succesful key. Indeed, a capture out of the raw footage would be much more preferable.
Once again, this is criticism towards the shot, not the camera.
I agree, although the problems I see look more like codec problems than actual problems on how it was all shot.
The sky has terrible banding and artifacts. You can see problems with keying all around the edge of the car. For example, look at the edge of the windshield on the left of the picture, one can see stepped gray pixels surrounding the edge. Same thing with the woman's shoulder, etc.
Even though the woman's hair is supposed to be one of those standard tests for keying, it looks terribly soft. I agree with Dmitry to the point that perhaps it was corrected in post. You can also barely make out the two last letters of "BUICK" on the car. it looks sharp(ened?), but i'm not sure if it actually captured that many details.
Just my opinion. I for one know that Mike Caparele is extremely skilled at what he does, so this makes me think that the problems I see are codec issues. Furthermore, this is only one frame, and a very particular one. No conclusions can be reached just from this one image.
The sky has terrible banding and artifacts. You can see problems with keying all around the edge of the car. For example, look at the edge of the windshield on the left of the picture, one can see stepped gray pixels surrounding the edge. Same thing with the woman's shoulder, etc.
Even though the woman's hair is supposed to be one of those standard tests for keying, it looks terribly soft. I agree with Dmitry to the point that perhaps it was corrected in post. You can also barely make out the two last letters of "BUICK" on the car. it looks sharp(ened?), but i'm not sure if it actually captured that many details.
Just my opinion. I for one know that Mike Caparele is extremely skilled at what he does, so this makes me think that the problems I see are codec issues. Furthermore, this is only one frame, and a very particular one. No conclusions can be reached just from this one image.
Dmitry Kichenko
Well-known member
It appears that the road is perpendicular the horizontal line, but the car is at an angle. Is it skidding? The singer must be drunk. That explains the woman's scared face .
.
stephenlnoe
Well-known member
Way back 10 pages ago this: http://toshpit.blogs.com/images/Car_headlights.tif was posted. Take a look at the files size of the frame grab. A measly 1.95 MB for an frame grab? All this conversation is about a screencap that isn't raw off the camera.
For example, here are two screencaps from a 720p timeline:
Uncompressed raw 720p
Uncompressed composite 720p
Either way the image file size size is 2.7 MB uncompressed per frame. The screencaps are sourced from a JVC HD-100. It makes me wonder what the source of the provided tiff image came from? It would be nice to have more info about how the editor captured the frame.
BTW: I invite you to zoom in on the screen caps down to the nitty gritty. They are VERY clean. A lot of people hack on HDV around here but you can't argue with results. JVC pulled a rabbit out of their hat with the ProHD codec.
For example, here are two screencaps from a 720p timeline:
Uncompressed raw 720p
Uncompressed composite 720p
Either way the image file size size is 2.7 MB uncompressed per frame. The screencaps are sourced from a JVC HD-100. It makes me wonder what the source of the provided tiff image came from? It would be nice to have more info about how the editor captured the frame.
BTW: I invite you to zoom in on the screen caps down to the nitty gritty. They are VERY clean. A lot of people hack on HDV around here but you can't argue with results. JVC pulled a rabbit out of their hat with the ProHD codec.