DVL-Digest 755 - Postings: Index FW: Resolution - Engineering vs. Marketing - (2) FW: Resolution - Engineering vs. Marketing - "Perry" Nice start Jan! Let's get one thing out of the way, Kevin is confusing lines and line pairs. DV can resolve 720 lines which is the way some folks measure video. Line pairs are used by printers and photographers. However, I have a few instant comments: 1)Jan mentions a sampling frequency of 18MHz. There are two completely separate sample frequencies, that (effectively) of the CCD and that of the subsequent DSP chain, I assume she means the DSP frequency. To give information content input to 18MHz sampling requires a minimum of 960 samples per line but as Jan alludes to in her post, this ignores the effect of the optical anti-alias filters. If you really wanted to see 960 lines AND get some effective ant-aliasing then you would need a lot more samples. 960 CCD pixels probably get you about 800 lines GROSS(with good anti-aliasing). 2)Jan glosses over the most important part of all in terms of accepted broadcast measurement, which is the translation between the optical chart and the sampling. As I had understood it, the 4:3 charts are calibrated for equivalence to vertical resolution. This means that the 800 line grating actually has 800 lines placed in 75% of the picture width, or has 1067 lines in the total width. For a camera to state that it has 800 lines resolution, allowing for anti-aliasing, would need around 1200 pixels! The situation for 16:9 is even worse and the numbers would be increased by 1.3. 3)The manufacturers like to strip off all the normal processing for specification measurements because it gives the best (artificial) signal/noise numbers. Most factors like gamma and detail add noise, and the coring functions used to hide noise will also hide the extreme tail end of resolution measurements. If they want to play these games they should state clearly on the specifications: "made in a completely non operational condition, and not available on any external connector". 4)Jan makes a fundamental mistake in her analysis: accepting that the detail processing is removed for testing gives a better signal/noise measurement, restoring it doesn't give a better resolution! This is because they are measuring EXTINCTION FREQUENCY. Detail enhancement can only boost what is already there and anyway is normally done at just below or just above video cut-off (or both) of about 5 MHz for NTSC or about 6MHz for PAL. It can make no improvement to resolution as measured here, and may well make it worse. 5)Using CCD pixel offset is indeed a way of getting more resolution, but it is only applicable to white/grey objects and has some fundamental problems with anti-aliasing. Jan doesn't say that her cameras use it I notice! 6)I think Jan has proven what those of us in the business have always known, and what she herself admits: there is a big gap between what a real engineer measures and what Marketing pretend they measure! Perry Mitchell Video Facilities http://www.perrybits.co.uk/ FW: Resolution - Engineering vs. Marketing - "Perry" Kevin posted: >Defining things in terms of frequencies makes sense when the entire signal chain is analog; these days most of the signal chain is digital, from the CCD to the tape format to the SDI distribution, through MPEG-2 satellite uplinks . If the final product is on DVD or digital cable, the only time it becomes analog is when it hits the viewers TV set, and not even then if they are using a computer or LCD screen to view the DVD.< But the two most important parts are still analog Kevin, the CCD sensor and the display device. Even an LCD display uses an analog modulation, and I don't know of any current device that uses digital brightness modulation. Perry Mitchell Video Facilities http://www.perrybits.co.uk/ (diese posts stammen von der DV-L Mailingliste - THX to Adam Wilt and Perry Mitchell :-) [up] |
 slashCAM  News HD Camcorder database One-on-One Cam comparison About our tests About us |
|
|
