How do I optimise capture levels?

[gsymeoni]

I have recently acquired a Panasonic HS960 and I'm experimenting with different captures with the AIW 7500 on VDub. I've made some comparison captures to help me analyse differences in levels, comparing different capture workflows and/or capture settings.

I'd like to read your thoughts/analysis of the screen captures and their histograms, and what you would use for best capture.

VCR Settings:

EDIT: OFF
COLOUR: SOFT
3D DNR: OFF
TBC: ON

Questions:

• Capture (1) extends way out of the luminance range, shouldn't the Panasonic output legal levels only?
  
• In line with sanlyn's guide, for capture (2) I've tweaked the AIW procamp to capture everything within legal levels. Should I be tuning the luminance right on the edges of the brown border? I've seen some of sanlyn's examples, and he gives it some slack sometimes, leaving a bit more "space" between the lowest/highest values and the brown border. If so, why?
  
• On capture (3) it's obvious that the Sony DVDR is doing some processing. It is definitely correcting levels (badly?clamping?), but something else is also at work, looking at the other histograms. What are those spikes, and is this ultimately destructive for capturing a master?
  
• Why is capture (3) more saturated, is this something I can tell by comparing the histograms? What is the Sony doing there?
  
• What does the y-axis signify? is it a broader colour range, a stronger signal perhaps? What does it mean when there's a larger area in the graph?
  
• I know it's hard to tell with resizing/compression/screenshot/etc/ but are there any glaring issues that my newbie eyes cannot see with the capture in general? (excluding levels)

Note: I've set the capture to the worst-case scenario, you can see some sun spots and the dark back-stage, this is just to help me learn handling the levels without worrying if I'm going to be crushing anything from tweaking a milder scene

Here are the captures:

1. VCR->AIW (No Proc Amp tweaks)
raw.jpg

2. VCR->AIW (Proc Amp tweaks)

You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.

3. VCR->SONY DVDR->AIW (No Proc Amp tweaks, letting the Sony do its thing)
proc.jpg

[EDIT: Just realised that the old Sony capture below adds borders, so I cut those to make the histogram comparison fair. See new screeshot above]

You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.

[Feedbucket]

The Y represents how much of the value on the X is present in the scene (black/white, yellow/blue, cyan/red).

I've found that spikes occur when these spectra are compressed and then re-expanded, usually through processing that results in a reduced amount of data like compression, lower-accuracy digital conversion, or denoising. The areas between the spikes represent potentially absent/lost data. This could be mitigated during/post re-expansion through some smoothing process but at that point it's just interpolation and guessing.

[gsymeoni]

Thanks for your insights, Feedbucket

Re. your first point, seems the y-axis scale is relative to the peaks and not a fixed number, that's what threw me off. Seeing that large spike at the right end is probably what has caused the graph to "compress", not a lower value per se

Interesting points on the spikes, noted