Comparing different TBC out frame sizes?

[Aya_Rei]

Hello all, so after getting a new Cypress frame TBC from lordsmurf I've been testing it out, all around it's really good! Though from comparing my captures to how I did them before with a Kramer FC-400 as the frame TBC, I'm noticing a difference in the width of the frame and I'd just like to hear which is more 'correct' if that makes sense.

Two raw image samples are provided below, the first is from the Cypress CDM, while the second is from the Kramer FC-400. The rest of the workflow, being a Sony CCD Trv-87 with line TBC and DNR on and a Pinnacle 510 capture card remain unchanged, the only equipment change was to the frame TBC.

To me at least, I feel like the Cypress is correct, as with the Kramer the image is stretched a bit horizontally, with the black bars on the Cypress image looking to more of equal size for both the left and right side compared to the Kramer, which has a bigger black bar area on the left side of the image compared to the smaller right side.

Cypress
Cypress CDM 640 Raw Reference.png

Kramer

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.

[lordsmurf]

Too many factors, hard to say. But your instincts are good, Cypress is generally accurate with geometry.

If you want a real mindbender, there's no guarantee your camera accurately recorded the geometry when the tape was made. ("There is no spoon." ~ The Matrix) So be careful with assumption, or overly analyzing raw capture.

Remember that post-processing is where certain things are best addressed, not attempt to "fix it all" or "it all needs to be perfect" from the capture hardware. Capturing is just step 1 of the longer video journey. Make it as best "as possible", not undershoot (from cheap/lazy), or attempting to overshoot. Know limits.

But yes, observation is always a good thing.

[Aya_Rei]

I understand how there are many factors when it comes to video, and this isn't as simple as it seems, but thank you for your response. Happy to know that I have good instincts here when it comes to which is more accurate between the two.

I do have another quick question on the topic of frame sizes, somebody more well inversed with upscaling content for YouTube might chime in here, but it's relating to cropping the final AVI in Hybrid. Just want to make sure what I should do from now on when it comes to cropping by a certain amount. As always I do crop from a 720x480 raw AVI file

So the first image was cropped by 12 pixels from the right, 16 from the left, 6 from the top and 8 from the bottom before being resized to 1440x1080 with a PAR of 1:1.

With that, the left and right sides of the image have double the cropped pixels as the top and bottom (28 and 14 respectively), this was after viewing your settings guide and thinking that this is how I should handle cropping, to make the left and right amount double from the top and bottom amount

CropTestOne.jpg

The second image is instead just cropped by 8 pixels from the right, left and bottom, with the top being cropped by 6. Therefore the top and bottom have a total of 14 pixels cut while the left and right have only 16. I've been told it's best to keep the width cropped to 704 to make the image true 4:3

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.

[lordsmurf]

Back up.

Crop 720 to 704 as the starting point.
Then crop 4x3.
Then upscale, pad/matte black to 16:9.

See that plate on the wall? It's a circle. Probably. I normally don't like to base geometry off edges of images, due to lens distortion. Middle of frame best.

Always verify the raw facts are what you see with your eyes. Separate science from assumptions made.

[Aya_Rei]

Ah, alright, guess I got ahead of myself here. When it comes to figuring out how the final result should look as 4:3 instead of 3:2. I tend to set the Aspect Ratio of the raw AVI in VirtualDub to 4:3 and use it's result as reference, I guess that isn't always the best case.

[aramkolt]

Ah yes, the good old aspect ratio conversation haha.

Where I get confused is if the "head switching nose" area at the bottom is supposed to figure in. Ideally you'd mask that to black rather than crop I think, but no one really wants a YouTube video with a black bar at the bottom.

How do you otherwise find that the performance of the Cypress versus the Kramer is better? I have something like 8 different TBC models to pit against each other and I'm having trouble designing tests that expose different flaws or even how to feed them an identical starting feed as there can be some variation from playback to playback. Ideally you'd digitally record an analog composite waveform and play the analog waveform back still as composite to each, but I'm not aware of a device that actually does that. A distribution amplifier would work to capture the same analog signal simultaneously from a single VHS playback, but then you're stuck using different capture cards with each TBC and needing a lot of different computers do do the captures which adds a lot of variables...

Test's I've come up with so far:
1. VCR blue screen
2. White noise of an unrecorded/blank tape
3. Tape with visible vertical jitter (usually these are EP tapes recorded on a different machine than the one playing back)
4. Physically slowing down the video head with your finger during playback
5. Loosen Video head drum mounts so that it is allowed to vibrate a little (this can create some pretty extreme vertical roll - found that out while trying to fix a dynamic drum that wasn't fully seated haha)
6. Macrovision removal - though this technically isn't a TBC "required" function - it's just vertical blanking, but is desirable for VHS transfers anyway - most rack mount units have a setting to enable or disable.
7. Grey ramp and some other test patterns to try to determine the true chroma subsampling and color bit depth.
8. Jitter and other measurements using a VM-700T (composite only)
9. Vectorscope/Waveform monitor looking at input and output levels to see how they change upon passthrough and if see where clipping starts to occur at illegal levels using a proc amp before the TBC.
10. Fine detail test patterns to determine any loss of vertical or horizontal sharpness.
11. Might attempt some way to measure the video delay added by the device, but not exactly sure how to do that yet - imagine it involves using a two probe oscilloscope on the input and output and subtracting the difference in time when a frame goes from all black to showing something.
12. Closed captioning preservation - Not sure how many people try to preserve that, but since it's right next to the macrovision signal, it can sometimes be wiped out depending on the device.

As far as your screenshots go, there seems to be some more chroma noise on the cypress most noticeable on the far right of your image on the refrigerator door edge. Could just be that the chroma levels are higher with the cypress though.

[Aya_Rei]

I was more so wondering about the difference in frame size, but all those tests do sound interesting.

On the topic of chroma noise, I can see where you are coming from, I've set the Cypress settings to their default values and the chroma levels do look to be a bit higher in some areas compared to the Kramer, which never had any proc amp controls.

Overall the Cypress is more pink-ish than the Kramer, while the white section on the left side appears to be more bright with the Kramer, and a bit darker with the Cypress. I could try to adjust the Cypress values to match how it looked on the Kramer. I just left the settings at their defaults because I didn't want to mess something up by leaving the contrast or brightness too high or too low for example.

-- merged --

So on the topic of comparing the Kramer and Cypress' outputs, especially in comparison to the colors. I'm a bit, concerned over the results. As it looks like the Cypress might be introducing more chroma noise? To me it looks like the main difference in the two is the whites being toned down, while the reds are instead brighter. I don't know exactly if the Cypress is introducing more chroma noise than the Kramer, or if it's just that the reds are brighter in general is what is throwing me off here..

Attached are two example images, the only difference in the workflow would be the change in frame TBC. The Cypress' proc amp settings have been set at their defaults, which is what I always go with

Kramer
KramerRedExample.png

Cypress

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.

Kramer
CypressRedExample.png

Cypress

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.

[aramkolt]

Hmm. Tough to say- still seems to me that there's some sort of scaling going on with the Cypress? I am aware there are supposed to be 16 horizontal pixels TOTAL (some before the line and some after) in each line of a 720x480 capture, but just seems visually you can see that it is more than 16 pixels total just having looked at a fair number of other captures and compared to your Kramer which does appear to have roughly the right amount of total black bars on either side of the image. I opened the image with a program called ImageJ that is used to "measure" things on images and it shows there's about 32 pixels of black on either side of the Cypress, so it's almost like the player is giving the 16 black pixels, but then the TBC is adding another 16 pixels on top of that. ImageJ is often used in labs to measure things on image captures of slides from microscopes where the scale/magnification is also known so that pixels can be translated into actual scaled distance, but it'll tell you how many pixels a line just drawn over the image was as well.

As for chroma noise, it's hard to say which one is being more true to the original source. Could be that the Kramer is doing some form of chroma noise reduction and the Cypress isn't? Or could be the cypress is adding noise and the Kramer isn't.

[Aya_Rei]

Quote:

Originally Posted by aramkolt

Hmm. Tough to say- still seems to me that there's some sort of scaling going on with the Cypress? I am aware there are supposed to be 16 horizontal pixels TOTAL (some before the line and some after) in each line of a 720x480 capture, but just seems visually you can see that it is more than 16 pixels total just having looked at a fair number of other captures and compared to your Kramer which does appear to have roughly the right amount of total black bars on either side of the image. I opened the image with a program called ImageJ that is used to "measure" things on images and it shows there's about 32 pixels of black on either side of the Cypress, so it's almost like the player is giving the 16 black pixels, but then the TBC is adding another 16 pixels on top of that. ImageJ is often used in labs to measure things on image captures of slides from microscopes where the scale/magnification is also known so that pixels can be translated into actual scaled distance, but it'll tell you how many pixels a line just drawn over the image was as well.

As for chroma noise, it's hard to say which one is being more true to the original source. Could be that the Kramer is doing some form of chroma noise reduction and the Cypress isn't? Or could be the cypress is adding noise and the Kramer isn't.

Ah I see, actually I do apologize with that sample as the original tape did have noticeable black edges, more so than other tapes and even with the FC-400.

Now I tried to test out another comparison, this time between using the CDM 640 and not using it (So the camcorder is plugged straight into the Pinnacle 510) To me this does show off the scaling the 640 is doing as it's squishing the image a bit horizontally

With Cypress TBC
With Cypress CDM 640.png

Without Cypress TBC

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.

[aramkolt]

Yeah, it seems to be adding more black side bars in that sample as well? I don't own a cypress to know if this is a known thing, but it seems to be adding it intentionally, and maybe can be switched off?

[lordsmurf]

Never assume the AR from the VCR is accurate. Never assume the TBC is accurate. Never assumes VCR+TBC is accurate. It's analog, there is slop and leeway. Nothing is perfect.

Slight geometry should be handled with mask/crop during processing. So it's a non-issue with such slight differences. Yes, those are tiny, maybe 8px total?

Cypress is definitely legal-izing some values there. That's good.

Cypress default chroma is just richer than Kramer. Cypress has proc amp, use it as needed.

I'm amused here because you're starting to get conflicting output from simple tests. Welcome to TBCs. You cannot do simple A/B testing. It takes a large body of work to isolate trends, The variables are the source tapes, sometimes VCRs, and those can massively skew results. See also: why this sort of testing can only be done with a wide variety of VCRs (dozens, not 2), and a wide variety of tapes with known knowns. It's more like A/B/C/E/F/G/H/I/etc testing needed to arrive at any understanding.

What you're doing in great, but don't overthink it, resist making shoddy hasty conclusions.

[Aya_Rei]

Quote:

Originally Posted by lordsmurf

Never assume the AR from the VCR is accurate. Never assume the TBC is accurate. Never assumes VCR+TBC is accurate. It's analog, there is slop and leeway. Nothing is perfect.

Slight geometry should be handled with mask/crop during processing. So it's a non-issue with such slight differences. Yes, those are tiny, maybe 8px total?

Cypress is definitely legal-izing some values there. That's good.

Cypress default chroma is just richer than Kramer. Cypress has proc amp, use it as needed.

I'm amused here because you're starting to get conflicting output from simple tests. Welcome to TBCs. You cannot do simple A/B testing. It takes a large body of work to isolate trends, The variables are the source tapes, sometimes VCRs, and those can massively skew results. See also: why this sort of testing can only be done with a wide variety of VCRs (dozens, not 2), and a wide variety of tapes with known knowns. It's more like A/B/C/E/F/G/H/I/etc testing needed to arrive at any understanding.

What you're doing in great, but don't overthink it, resist making shoddy hasty conclusions.

That's probably what it is, getting so worked up over something that isn't an issue after all. This difference in frame size would just require me to crop off more pixels from the sides for when I inevitably do some restoration work to it. And you're right with the difference being around 8px as I'd need to crop off 6 pixels from the right side to hide the black bar for the image that didn't have the Cypress TBC used, while for the image that did have it used, I'd have to crop off 14 pixels from the right side.

Though I suppose at the end of the day this isn't that big of a deal, and I'll know how to deal with it.