nanoFlash Public Beta 1.6.226 Firmware Comments

[Piotr Wozniacki]

Originally Posted by Adam Stanislav

Is that Wawel?

That's right, Adam. BTW, this particular event took place at the Batory Courtyard - a place never available to the public before.

[Billy Steinberg]

Originally Posted by Piotr Wozniacki

Exactly, Peter:

4:4:4 ----> H:1/1, V:1/1

4:2:2 ----> H:1/2, V:1/1
4:2:0 ----> H:1/2, V:1/2

- which means the vertical resolution is exactly 2x higher in 4:2:2 than it's in 4:2:0. Both 422 and 420 have horizontal resolution 2x lower than the full 4:4:4 sampling (and this is probably what you meant).

Keep in mind that you're talking about chroma subsampling here, not luminance. While I'm not trying to dismiss the degradation that happens with chroma subsampling, it's a far cry from what most people think of as 1/2 the resolution.

Billy

[Piotr Wozniacki]

Thank you very much for enlightening me, Mr. Steinberg. It would have never occurred to me!

[Billy Steinberg]

Originally Posted by Piotr Wozniacki

Dear Dan,

Just to return to this thread main subject: what do you think is causing this slight audio delay in the present Beta?

FYI, it used to be even longer with previous firmware releases (I don't know about the 1.6.29, as I skipped it).

Do you have an easy means to check the sync right out of the camera's HD-SDI spigot (or HDMI spigot if that's what you use)?

The reason I ask is that the path the video/audio takes to the internal recorder is very different from the path it takes getting out of the camera. Just because the internal recorder is dead on sync while the nano recording is slightly out doesn't mean the delay is being introduced by the nano. It doesn't mean it's not the fault of the nano, particularly if early versions of the nano firmware were enough out of sync that it was obvious it wasn't the camera, but now that you're talking about minuscule sync errors, before you spend a huge amount of time determining if the nano is out of sync, you might want to check if the feed from your camera is out of sync.

Just a thought.

Billy

[Rafael Amador]

Originally Posted by Billy Steinberg

Keep in mind that you're talking about chroma subsampling here, not luminance. While I'm not trying to dismiss the degradation that happens with chroma subsampling, it's a far cry from what most people think of as 1/2 the resolution.

Billy

Right.
Luma is full resolution, whatever the format, from DV up.
rafael

[Peter Moretti]

Originally Posted by Rafael Amador

Peter,
The miss match is HUGE.
You simply can not cut between a picture with an aperture and the same picture with a different aperture.
The picture JUMP as when you were cutting between signal out of sync.
They have different sizes on screen.
This may be something originally designed for DV, but affects every single standard, format, size,..
Just open any QT file and switch between the different aperture options.
rafael

"Classic" and "Encoded Pixels" mode usually look the same. Many times there will be no difference between "Clean" and "Production" modes. One crops for frame size and one doesn't.

So are you saying that nano is using the wrong pixel aspect ratio for the codec, which would cause a differenc between "Classic" nano and "Production" native recording? If so, how come no one, AFICT, has noticed this? (It's hugely obvious but no one has complained about it?)

Finally, are you saying that the six dots around the center are caused by an Aperture Mode difference? You can see clear as day in the images by Piotr that the difference is caused by chroma bleeding in along the horizontal grey band of the color bars. That has to do with 4:2:2 vs 4:2:0 chroma sub-sampling. It has nothing to do w/ Aperture size. If you can show otherwise, then I will readily admit I'm wrong. I don't want to mislead anyone. But I don't see how you can say that difference is caused by Aperture mode. But, of course, I'll leave my mind open to being convinced otherwise.

[Adam Stanislav]

Originally Posted by Dan Keaton

While I understand a Vectorscope and Histogram, I do not consider myself an expert in using these tools.

It is really very simple. The vectorscope plots the chroma on Cartesian coordinates, with Cb being the x-axis and Cr being the y-axis. So, it calculates the Cb and the Cr of every pixel in the image and places a dot in the 2D plane based on those coordinates. It does not tell you how many pixels each dot represents, only that there is at least one pixel in the bitmap with those Cb and Cr values.

The placement of each dot around the circle (i.e., the angle of the imaginary line connecting the dot and the center of the coordinate system) represents the hue of the pixel(s) the dot represents. The distance of the dot from the center (i.e., the length of said imaginary line) represents the saturation of the pixel(s) the dot represent.

So, any pixel with no saturation (black, white, or any gray value in-between) will be represented by a dot in the center (Cb = 0, Cr = 0). In the case Piotr has presented, the EX image had a lot of pixels with very little saturation (so the dots were scattered near the center), while those same pixels had no saturation in the nF image (one dot in the center). Neither is necessarily “better,” they are just different.

Final Cut Pro 7 - Scopes - Vectorscope describes how to interpret the vectorscope in Final Cut Pro, but it really applies to any software displaying a vectorscope.

The histogram calculates the Y (luminosity) of each pixel and shows you how many pixels in the bitmap have that Y value. In this case, the x-axis represents the luminosity, the y-axis the number of the pixels with that particular luminosity. The x-axis also shows you the darkest (left extreme) and the lightest (right extreme) colors in the bitmap.

At any rate, when color grading, the vectorscope helps you adjust different clips to about the same hue and saturation, the histogram helps you adjust them to about the same range of luminosities.

[Dan Keaton]

Dear Adam,

Thank you very much.

I was thinking of the single dot in the center of the nanoFlash vectorscope as being Black.

Is this correct?

[Adam Stanislav]

No, the single dot in the center means any type of gray, from black to white. In other words, anything with no chroma regardless of its luminosity (or luminance or brightness).

I have just taken a second look at those graphs and realized they were of the SMPTE bars. In that case I take back what I said one was not necessarily better than the other. The nF image is clearly superior to the EX image because the EX has some chroma for the grays, which clearly should be achromatic, just as the nF has them.

There are very few dots in those vectorscopes because there are only a few colors in the SMPTE bars. And as I pointed out, the vectorscope simply shows how many colors are in an image, not how many pixels are of each color.

Neither is perfect due to the MPEG compression, but the nano is considerably better: It shows four different yellows, one green, three cyans, two blues, two magentas, and two reds. There really should be just one or two of each. The white, black, and the grays are all without any chroma (as they should be), so that is excellent.

The EX shows many more colors because it compresses more, so it produces more intermediate colors at the edges of two colors. And, unfortunately, it adds some chroma to the grays, which is rather bad.

The nano is a clear winner here.

[Piotr Wozniacki]

Originally Posted by Adam Stanislav

The EX shows many more colors because it compresses more, so it produces more intermediate colors at the edges of two colors. And, unfortunately, it adds some chroma to the grays, which is rather bad..

Agreed - this is exactly what I meant supposing earlier:

Originally Posted by Piotr Wozniacki

[...] that the "extra information" would probably be just the EX's lower color resolution garbage.

Thanks Adam!

[John Richard]

Maybe Adam, who is more versed than I in scopes, could chime in on the potential effect of software scopes vs. hardware scopes in this discussion.

Most software scopes are only using parts of the image to display results. Some software scopes will only look at and display the results of every 6th or 7th scan line - NOT the whole image as is done with hardware scopes. The only sofware scope I know that attempts to display the vectorsope results of the WHOLE image is the one in Apple's Color (and even this software scope uses some manipulations to achieve a measurement representative of the whole image).

It would seem to me that we are looking at some very fine quality detail judgements with a software vectorscope that is only measuring a very small percentage of the scan lines in the image. I have no idea what the actual ratio of total scan lines that the scopes in Sony Vegas use, but I would bet that it isn't measuring but a small percentage of the scan lines in the image as do most software scopes. I think we could be using the wrong type of vectorscope to make such critical technical quality decisions?

[Rafael Amador]

Originally Posted by Peter Moretti

"Classic" and "Encoded Pixels" mode usually look the same. Many times there will be no difference between "Clean" and "Production" modes. One crops for frame size and one doesn't.

So are you saying that nano is using the wrong pixel aspect ratio for the codec, which would cause a differenc between "Classic" nano and "Production" native recording? If so, how come no one, AFICT, has noticed this? (It's hugely obvious but no one has complained about it?)

Finally, are you saying that the six dots around the center are caused by an Aperture Mode difference? You can see clear as day in the images by Piotr that the difference is caused by chroma bleeding in along the horizontal grey band of the color bars. That has to do with 4:2:2 vs 4:2:0 chroma sub-sampling. It has nothing to do w/ Aperture size. If you can show otherwise, then I will readily admit I'm wrong. I don't want to mislead anyone. But I don't see how you can say that difference is caused by Aperture mode. But, of course, I'll leave my mind open to being convinced otherwise.

Hi Peter,
No.
No relation between the "6 dots" and the aperture.
The "6 dots" are clearly generated when the 422 stuff is down-sampled as 420.
When the line on the base of the color bars are merged in with the top line of the gray below.
The 420 blocks generates those pixels, with the same tone, but desaturated.
No mystery.

About "Aperture", as you can read in the fallowing post, I've asked CD to address this issue.

http://www.dvinfo.net/forum/converge...-aperture.html

I've saw long ago that NANO pictures do not match with the same pictures on the SxS cards.

I have to apologize because I've pointed Dan about "Aperture" as the reason of the NANO/SxS files miss matching on the Piotr test.
I'm trying to find out if there is something similar to "Aperture" on the MXF files.
I understand that the need of different "presentations" for digital video, is not exclusive of the QT files.
However I'm very limited to try to dig on the MXF format.
Cheers,
rafael

[Piotr Wozniacki]

Originally Posted by Rafael Amador

About "Aperture", as you can read in the fallowing post, I've asked CD to address this issue.

http://www.dvinfo.net/forum/converge...-aperture.html

I've saw long ago that NANO pictures do not match with the same pictures on the SxS cards.

Rafael,

Yep - I remembered your post immediately after the "Aperture" term came into play in this thread. I've been wondering the same - is there a similar setting somewhere, for the MXF format?

So far, I found nothing.

Cheers

Piotr

[Piotr Wozniacki]

Originally Posted by John Richard

Maybe Adam, who is more versed than I in scopes, could chime in on the potential effect of software scopes vs. hardware scopes in this discussion.

Most software scopes are only using parts of the image to display results. Some software scopes will only look at and display the results of every 6th or 7th scan line - NOT the whole image as is done with hardware scopes. The only sofware scope I know that attempts to display the vectorsope results of the WHOLE image is the one in Apple's Color (and even this software scope uses some manipulations to achieve a measurement representative of the whole image).

It would seem to me that we are looking at some very fine quality detail judgements with a software vectorscope that is only measuring a very small percentage of the scan lines in the image. I have no idea what the actual ratio of total scan lines that the scopes in Sony Vegas use, but I would bet that it isn't measuring but a small percentage of the scan lines in the image as do most software scopes. I think we could be using the wrong type of vectorscope to make such critical technical quality decisions?

John,

I think you're right that the software scopes are not as accurate (and unbiased) as good, properly calibrated and used, hardware instruments - but I believe that for the sake of this simple comparison, they're enough.

What they show is (at least for me) clearly the same as what I'd expect - and Adam's post explains the details very well.

Nevertheless, being knowledge-hungry - I'm still hoping someone will join the party here with even more details, especially on the Aperture setting (whether it has its counterpart in the PC-based NLE world, etc.).

Cheers

Piotr

[Adam Stanislav]

Originally Posted by John Richard

Maybe Adam, who is more versed than I in scopes

Oh, I am no expert. I just understand the mathematics behind it. As for the difference between hardware and software, it is hard to tell without knowing about the specific software used. Some software will only look at portions of the image, some at the entire image. And some will look at the entire image of the current frame but only at portions of the image while playing the video. Add to it that software manuals do not always disclose this information.

Nevertheless, in this case we were comparing the vectorscope of the same image as compressed by two different devices, presumably using the same software for the scope, and the differences were quite clear.