Now that you have seen the footage...
 

OK, I asked this before, but it wasn't the time.
Now that you have already seen the FX1 footage, I'll bring the question back.

Can we compare a downconverted HDV>SD footage from FX1 against a SD DIGITAL BETACAM or DVCPRO 50? Or perhaps, a BETACAM SP/SX?

OK, forget about the pro controls of these cameras, just overall image quality.

No, and that's not why you should be getting it.

"Can we compare a downconverted HDV>SD footage from FX1 against a SD DIGITAL BETACAM or DVCPRO 50? Or perhaps, a BETACAM SP/SX?"

all the above will look BETTER then HDV downconverted to SD ..

IMO shooting with the camera in SD will look a little better then shooting 1080i and downconverting to SD ... but even in standard DV mode it is not going to beat out digibeta , DVCpro 50 etc ..

I'm not trying to be cantankerous or start a fight or anything...that just honestly doesn't make sense to me...unless the concern is with the MPEG2 compression. Why wouldn't starting at a higher resolution and then downconverting look better than starting at the target resolution? It seems like if you record more information initially, you'd have more, even down converted... Please show me the error of my ways... (as this was my initial reason for pushing to get this cam as everything I do is SD right now...

Kevin, your theory is completely correct if we ignore factors like the lens, compression and CCD sentivity. True a progressive source would be better for downscaling, but generally HD sources scaled to SD will output perform SD native sources. Note: many high-end SD sources actually use higher resolution (than SD) CCDs to gain this scaling benefit. Basically you aren't losing anything by shooting in HD, because over-sampling is good.

I never did understand how that works. Mathematically it shouldn't as far as I know....

<<<-- Originally posted by Aaron Shaw : I never did understand how that works. Mathematically it shouldn't as far as I know.... -->>>


Perhaps because matematics laws do not apply here, because we are not dealing with a linear system.

A compression system is random ruled by definition, and this seems to be what is happening in HDV for now.


Carlos

Compression isn't random. It is written with math in any case. Math is also very much not limited to linear systems. All sorts of things can be done.

I honestly don't see how that would work and I am of the opinion it is a myth until someone proves otherwise (please do!). You only have so many pixels... you have the same number of pixels with a downconverted source and a source that is native. Can anyone explain the thinking behind this?

<<<-- Originally posted by Aaron Shaw : Compression isn't random. It is written with math in any case. Math is also very much not limited to linear systems. All sorts of things can be done.

I honestly don't see how that would work and I am of the opinion it is a myth until someone proves otherwise (please do!). You only have so many pixels... you have the same number of pixels with a downconverted source and a source that is native. Can anyone explain the thinking behind this? -->>>

If it's not random, or at least a great part of it, then why don't you get the same result every time you do a scene?

Modern compression systems are lossy, that is they throw away info they find to be redundant to be able to fit the media you use.

So until someone shows me otherwise, video compression systems are not linear.


Carlos

There are so many variables in this argument that it could go on for a long time. FIrst we have to look at the chain in the camera from lens to tape. From an electronics point of view we can start with the CCD's. These are analogue and produce a charge proportional to the light falling on them(Charge Coupled Device) that has to be measured and turned into a number for the rest of the system to process. The higher the resolution of this process the better. The number of bits used to measure this number. These sets of numbers are then used for all the camera function/parameters.The longer this arithmatic stays at high number resolution the more accurate the output will be. To record to tape this data must now be converted to the tape format whether it is DV, MPEG2 etc. Again the higher the number resolution for this conversion the better. IF internal processing was at 14bit say it will now be 'downconverted' to whatever the tape format requirments are, say 8 bit. In arithmetic terms the rounding errors will be less by staying in the higher bit representations for longer. An example of rounding as follows. Say the A/D created a value of 1.81645 which has to be multiplied by another measured value of 400.189 this is 726.92331 rounded to 5 places. However if we started with only two decimal places the values would have been 1.82 and 400.19 of the same numbers giving us a value of 728.35. If these values are being used say for luminance one picture would be brighter than the other. Just by a little but this was just one multiplication the actual encoding is a lot more complicated and could produce larger errors every time the result was multipied.
Encoders are lossy but they create the same file every time from the same source data. However low resolution A/D do not reproduce the same data every time because slight changes in the CCD charges could be magnified by the low resolution A/D.
More expensive cameras will have CCD's with high signal to noise ratios and low self noise characteristics resulting in a more repeatable output, have higher resolution A/D and internal processing again result in repeatablility and consistancy.

Ron Evans

Ron,


Completely agreed on what you said.

The large errors will always be in the A/D stages, which are much better as bits go up and CCDs get better.

When I refer to linearity I mean as compared to film curves. Analog video systems have less stops than film but are reasonably linear.



Carlos

Carlos :)

What does the gamma curve and dynamic range have to do with simple resolution? I don't see how that relates to what I was discussing.

Ron: you make some good points. Obviously higher bit processing in camera will produce superior results. That is, however, part of my point. I don't think the fact that an image is downconverted from a higher res has anything to do with a percieved quality change. It seems much more likely to be external camera specific factors.

The simplistic example I gave was to show that low bit resolution system would potentially give a different value to a high bit resolution system either absolutely or when downconverting to the lower bit resolution. When downconverting the inherent quality will be retained up to the level of the resolution of the downconversion. IN my example downconverting to 2 decimal places would mean the higher resolution processing would give a value of 726.92 and the low resolution would be 728.35. THe inherent accuracy of the high internal processing is retained just expressed to the limits of the lower resolution. So simply downconverting to a lower resolution should be better. This means the 14bit DXP is benificial both to HDV and DV. The issue is what else in terms of encoding takes place before visible on a display device. There have been several comments that the component outputs seem to look better than the DV downconversion/transcode. This may be true since this is a downconversion to analogue for display rather than yet another encode to DV as well as a downconversion, then a decode of DV to analogue to view on a normal TV. I would not be suprised to find the composite output from the FX1 camcorder is better than the DV output decoded by another camcorder /deck for viewing on the same monitor!! Hopefully my FX1 will come in the next week and I will be able to check everything out for myself.
I intend to use DV recording for 4x3 DV so that I can mix easily with my present or rented equipment and will experiment with HDV. My purchase of the FX1 is mainly to buy a 3CCD DV camcorder with some future proofing.
Ron Evans

<<<-- Originally posted by Aaron Shaw : Carlos :)

What does the gamma curve and dynamic range have to do with simple resolution? I don't see how that relates to what I was discussing.
-->>>


I thought we were discussing non-linearities instead of resolution. Even if resolution is certainly a part of it.

DV and apparently HDV are prey to several aberrations, particularly in dark areas, which we did not have in analog video.

So in my opinion it certainly relates.



Carlos

When I refer to high and low bit resolution I am NOT talking about TV lines of resolution. I am refering to the internal arithmatic used to translate the CCD output into a data stream to be encoded for recording to tape. The higher this arithmatic resolution the more points of measurement are used to translate a real scene into numbers. Having got this first set of data, transforms can be applied to bring out detail in the dark areas ( black stretch) or reduce the effects of over exposure ( knee curve to avoid clipping of whites) Blacks are sometime crushed and whites clipped to make sure there are enough "numbers " to bring out the detail in the middle exposure range when transcoded back to 8 bit to record to tape. This is why all the really Pro cameras have camera setting menus so that the desired output can be set as needed. Consumer camcorders do not normally have these controls and the user is left with what the design engineer decided!!! This contouring is the gamma curve response. Our eyes have a very large dynamic range and we can see detail in both very dark and very light areas at the same time. Cameras are not that good. But a 14Bit DSP camera will "see" more than an 8 bit DSP camera and has the opportunity to decided what will be seen when transcoded to 8 bit. With an 8 bit system there are only 256 levels available to be recorded to tape. Changing the gamma curve will redistribute how the original scene ( say at 14 bit data resolution with 16384 levels) will be encoded to 8 bit. And yes this is by its very nature a non linear translation. THis is why even for consumer cameras the camera output is considerably better than what is recorded to tape which is limited by the recording process. DV is quite capable of recording fine black detail, fine white detail and fine mid detail, but with only 8 bits you can't have them all at the same time!!!! But the choice is up to you if you have a Pro camera. With a consumer/Prosumer camera your choice is to overexpose or underexpose!!!
In the analogue world advantage was taken of the fact that lots of devices are nonlinear in response and by default did not produce the nasty responses of digital circuits to overload( they automatically produces a knee effect when they saturate). Analogue has its own and to some extent worse problems noise etc.

Ron Evans