The VideoEq -- A low cost external grayscale/gamma/cms tool - Page 3

What mainstream consumer source video has more than 8-bit 4:2:0?

Dave

None. That's not the question.


The JVCs are good as long as things aren't moving too much.

FYI, I've updated our VideoEq FAQ to include the following (pretty basic) questions:


This'll be ongoing of course. I'm following a pile of threads all over the place, talking to SpectraCal, and have questions emailed to me all the time. As I get asked questions, I update the FAQ.

A few (largish) images of the Pro version have also been added. See [b]here[/b.

Kal

I beg to differ.


That was the question. How could it be any more than 8 bit if there is no source for >8-bit?

Dave

The reason I was asking if the RGB signal was 8-bit was because I expect banding to be a possible problem with 8 bit sources based on the 8-bit in, transform, 8-bit out. If that signal were first padded to 10 bits, I also expect the 10-bit-padded in, transform, 10-bit out to have the banding either reduced or eliminated. Note: we are not talking about banding that may already be in the source -there is nothing you can do about that.

The fact that content is only 8-bit doesn't mean that 8-bit processing is sufficient. If you process in 8-bit, as has been explained to you for many years, you end up with significantly less than 8-bit because you can't remap things like that without smashing levels together and skipping treads. Hence banding.


It outputs 10 or 12-bit. It's padded to that if untouched, or if being processed then you can have the benefits of higher bitdepth.

There are many sources which will output in a higher bitdepth, as well as video processors, and this is hugely advantageous if you're doing video processing. The VideoEQ, being a video processor itself, then should take advantage of this need to preserve the 8-bit content's quality through to the display. If you're only processing in 8-bit or running through an 8-bit bottleneck after processing then you're throwing away bits from the original 8-bit content and ending up with <8-bit effectively, hence banding. That's the concern. We're trying to preserve the quality of the 8-bit source content without degradation and a reduction in the effective bit-depth and a consequent introduction of banding problems. Continuing to repeat that "content is only 8-bit 4:2:0" is like continually repeating the first sentence of a novel when everyone else in class has long since moved on in the book. We get it. Content is only 8-bit. That's kind of entirely beside the point and if you're trying to say that therefore we don't need anything more than 8-bit for processing or LUTs, then please create your own thread to argue that, and you can have a wonderfully productive conversation with yourself.

Padded with what? Zeros? Zeros are literally nothing. Padded with information'? Where and how did one get this information' to pad with if it wasn't ever in the source?

You could pad it with a 1000 bits of zeroes and it would still only have 8-bits of information and 220 valid Y steps if the original source material was 8-bit 4:2:0.

I'd expect zero-padded 8-bit video to behave exactly like 8-bit video. No reduction or elimination of any banding.

Dave

The file format for the VideoEq CMS has changed a bit to be easier to use:

Yes padded with nothing. It's exactly the same as 8-bit. The number of steps of actual information are exactly the same.

Amen.

Sorry...I wish I'd never asked.

Back OT the CMS being in a more familiar format looks to be an improvement over the earlier version. The pictures don't do it justice either. It looks nicer in the flesh, so to speak, and it's only slightly larger than the size of a packet of cigarettes, which surprised me when it came.

You can't just pad it, that won't work, you have to expand it. Unless the source and target have agreed that you are using a protocol that is padding. 8 bit references are 16-235, 10 bit is 64-940.

for instance 16 in binary (big endian)
0001 0000
If you padded the most significant bits to 10 bit
00 0001 0000
The value is still 16, but black is now suppose to be at 64.

If you padded the least significant bits to 10 bit
00 0100 0000
You actually do get 64 for black, but this doesn't work for white

White at 220 in binary
1101 1100
Pad the lead significant you get
11 0111 0000
Which is 880, but reference white in 10 is 940.

So if you want to maintain the integrity of the signal you're gonna have to do some floating point math.

Derek, can these numbers go above 100? Not that I expect to increase saturation, but the brightness value comes to mind: if for example, for some reason reducing saturation happens to also reduce brightness, and one needs to compensate by increasing brightness. If that does not happen, then I guess I have nothing to worry about.

Thanks

Yes you can go above 100 but at some point they really don't do much because of already being at the top of the range. If you lower say saturation you do then have some room to increase brightness above 100.

So far in testing the interaction between saturation and brightness is minimal.

As I mentioned earlier, I spent a lot of time this afternoon taking readings and adjusting the LUT (currently, that means editing a text file, uploading it to the box, waiting for it to change and measuring) at 0-15% in 1% steps, and 15-100% in 5% steps, with a total of 33-points. I've now put that into a spreadsheet to have it fill in the gaps to create a full LUT and uploaded it to the VideoEQ.


Note: It looks like I made the same mistake right next to 0 as I did the first time (having the last couple of points drop to 0 sharply rather than simply being cut off) so I'll have to correct that.

It also looks like there's some unevenness with the LUT in those 1% steps that might look better if they were simply removed/averagedI'll have to try it and see.

I'm actually wondering if, because it's a CRT, I should just use two pointsone at white and one as close to black as I can accurately measure, as it seems like the LUT is mostly a straight line. (the non-linearities may simply be errors in the measurements) Something to experiment with anyway

As it is though, it is a significant improvement over the LUTs created from measuring in 10% steps. (really needed more data near black to get it right)


I should point out that my setup is a worst-case scenario: 8-bit in (so 8-bit out) and I'm using a CRT which does not posterise the image like a flat-panel/projector would, so any loss in precision is visible whereas a flat panel may mask some of it with its own imprecision.

With this new LUT (which I am sure can be improved further) I would say that posterisation is not an issue 90% of the time. 9% of that last 10 is visible but not problematic, and 1% of the time ugly posterisation is noticeable.

What I mean by ugly' posterisation is when you have very obvious discolouration, rather than simply banding in the image. Banding is unwanted but tolerable to some degree, whereas discolouration is not. An example of ugly posterisation is the red/yellow blotches in this image. (of course nothing looks that bad)


I have tested a number of films/scenes that I remember being problematic on many of the digital displays I have used, and I can confidently say that while there is posterisation introduced into the image, it is much less noticeable, but more importantly, there is a lot less ugly' posterisation than on any of them.

There are two things I need to mention though that may be affecting this:

1. My player is dithering which will be helping to mask the posterisation. (HTPC using madVR)
2. These comments are all relating to Blu-ray. (which is 99% of what I watch these days) With DVD, posterisation is more noticeable as a result of the lower resolution and higher compression. It's still generally showing as banding rather than bad discolouration, however.

So, is it worth it? Probably, if you are willing accept that there may be some posterisation introduced into the image, and can spend the time to get a LUT properly adjusted. This should be a lot easier once AV Foundry release a GUI tool, and if the CalMAN v4 automation works well. (at the very least I expect it will do 90% of the work)

Aside from the fact that it is introducing posterisation into the image (again, may not be noticeable/problematic depending on your display/setup/LUT) I would say that it is a significant upgrade in my systemand I haven't made use of the CMS at all yet. (though my display is fairly accurate already)

Having the greyscale & gamma as close to being linear as possible makes a massive difference to the image. Most displays do not even have the controls necessary to get gamma flat, let alone getting it flat at the correct level. (2.35±0.15)

With my CRT this has allowed me to hit 2.35 down to the lower limits of my meter, and keep black at zero luminance.

The most noticeable effects of this on image quality are:

• Contrast is up significantly. Previously to get anything close to a flat greyscale/gamma I had to raise the brightness, reducing most of the black-level advantage CRT has. It also had to be ≈2.22 to be flat, rather than 2.35 so combined with the raised black level, the image looked washed-out.
• There is a lot more pop' in dark scenes. There are some scenes in films that, no matter what I did, always looked dim/flat/murky/drained of all colour. With the controls on the CRT, you could either raise the black level to make the detail more visible, or lower it to have them looking appropriately dark at the expense of it. Neither situation was ideal, and shadow detail was never well definedit all seemed to blend together.
  
  With the LUT, these scenes now look the way they always should have done. Even though they are dark, shadow detail is clearly visible (but not because it's been exaggerated the way some displays do) there is a lot more colour in the scene. The image just pops' in a way I've never seen before. I can't think of any way to better describe it.
  
  This may just be a CRT thing though, as the black level on most flat panels is too high to run a flat 2.35 gamma.

Compared to other devices/software I have used, this produces much better results. Anything that has simply offered options to raise/lower gamma has washed out the image near black, and compressed highlights.

Anything offering finer control has either introduced a lot more posterisation or simply hasn't had enough control for the loss in precision to be worth it. (it's also cost a lot more)

For the money, the VideoEQ seems like it's a great device. As good as the results are now though, it really makes me wish that it could output 10-bit regardless of input bit-depth, and really, the more precision the better. Maybe version 2. (hopefully this will sell enough to make that possible)

Certainly, I was just oversimplifying for the sake of clarity. You still have the same number of actual utilized values as in 8-bit, the values are just entirely different and in a new range. The benefits only accrue after you start doing any kind of processing within the new bitdepth. We are in agreement.

From reading this thread, it seems that this is a much better solution to grayscale/gamma and CMS correction for a JVC RS2 (Pioneer Elite FPJ1) than using a Lumagen HDP/Q, correct? I like my Lumagen/Pioneer setup, but I do get some banding that I would love to eliminate. Would the VideoEq Pro help with the banding as well?

It might have been me, but as I haven't really got going yet with it I can't comment. I planned to use the HDQ for V-stretch, DVD upscaling and electronic cropping to trim my slightly overspilt image, but I'm reconsidering my options after the 8 bits debate as it won't pass more than 8 bits being DVI.

(I wouldn't leave your HDQ on the shelf gathering dust, as there is still a used market for them, for now anyway ).

Questions for Derek...do you see any issues with older Displays that may prevent the VideoEQ from being fully utilized ? With the Pro version , can we setup a LUT for an SD and an HD Gamut ?

Scott................

I don't see why not as long as the display can at least hit rec709.

I feel like I'm taking crazy pills. I don't know about 8 bits, 10 bits. 4 bits a dollar

I like the idea of doing this myself. After having a ceritifed ISF calibrator come over twice and then leaving with I'm not sure but I think it looks better. come on. But looking at CALman and the videoEQ, I find the following;

CALman Home - $674
VideoEQ - $599
total $1273

CALman Enthusiast - $974
VideoEQ PRO - $999
total $1973
(and these are special introductory pricing!)

Are you going to be disappointed if you get the Home and standard videoEQ? After all, my RS2 does not have a CMS.

And as a fellow member said;

"That would be great if the home version supported the automation in the software for the VideoEQ's, otherwise it is a few hundred dollars more to jump to the Enthusiast.
Lets face it most home users will use this device on one display, set it up and never touch it again except to recalibrate every few months. I think I have more invested now than my projector is worth."

And my understanding is you have to get the DPG 1200 to generate the necessary displays? Does the calMAN enthusiast include this ability or am I loading a disk to load each one as needed?

According to the website;

The Enthusiast License supports the new CalMAN DPG-1000
The Enthusiast license allows you to automate whole measurement sequences using the DPG-1000, the first Digital Pattern Generator. Then what of the automation talked about with CalMAN version 4?

And if I buy another projector with CMS, then what of my videoEQ pro or not.

So any RS2/HD100 pioneer eq. These projector's got rave reviews out of the box. Is this worth it?

I think you can get the gray-scale nailed with just the RS2. Of course you'll need a meter and software. Accurate colors would require something like the VIdeoEQ Pro.

The RS2 is an excellent projector, but the native primaries of the RS series units are heavily oversaturated and inaccurate. Unless you have one of the models with on-board CMS (the RS2 does not), then an outboard CMS is the only way to achieve an accurate color gamut on the JVC projectors. This is one option to accomplish that, and more affordable than a fully featured video processor such as the Radiance at $3.5K retail, if you don't need the additional abilities of a fully featured video processor.


Except for the native color gamut which uncorrected is not accurate.

Whether it is worth it to you depends on whether 1) you want an accurate Rec709 or SMPTE C color gamut and 2) whether the cost of achieving that over your current inaccurate gamut is more important than spending that money on other things. Nobody can answer those questions for you because they are individually subjective. Some people are extremely picky and adamant about color gamut, many people are not and don't know what we're talking about.

This product looks interesting. While I would love to buy it bundled with CalMAN 4, I'm a home user with only two displays, a sense of frugality, and am not in a hurry. My question is how effective I could be, by simply using the Windows LUT-generating GUIs that are supplied with the VideoEQ Pro (I haven't seen anyone comment yet on how easy to use these are to use). If the convergence to, say, REC709 with HCFR and the VideoEQ Pro was easy, that would be one thing. If they interacted like the first rev of the RS20 CMS, with all but a few experts baffled by the interactions of the controls, that would be fatal for me. If, on the other hand, the video corrections were orthogonal and a neophyte could easily use the product to converge to the desired goal, that would be... a sale. Comments are solicited.

Kevin

I have the VideoEQ Pro on order but am also very disappointed that the Calman Home software (which I have) will not be able to control the VideoEQ at v4. The VideoEQ is for my personal use on a single TV, I have one meter and don't see anything in the Calman Enthusiast software (multiple meters etc) that is relevant to me.

Hopefully the bundled Windows software will be more than adequate, but by not including the functionality in the home version of Calman4 SpectraCal are missing the opportunity to more closely lock me into their software path. I hadn't used Calman for a year or so as I found HCFR more straightforward, but had switched back to Calman in anticipation of v4 and the VideoEQ. Now I won't get the functionality I was really looking for, so it throws the door open again.

I wouldn't worry too much about which software you are using with the VideoEQ Pro: I'm using Chromapure as I find it is really easy to use for colour gamut work (compared to HFCR which I started off with). Unfortunately I am unable to do the CMS work just yet as I'm upgrading to the Pro model from the EAP version I have now so I'm without any VideoEQ as of today.

With the Pro model you can control the CMS settings in real time using a remote (though I don't think on is supplied so you need to use a Pronto or Harmony - I have the later). Or using the USB port and a PC to control the settings via supplied software, again in real time AFAIK.

Sure the fully automatted process would be slicker, but given that the VideoEQ Pro doesn't seem to have the interaction between colours in the CMS, then I expect the process to be relatively straightforward (compared to early HD750 CMS for example): Assuming the greyscale and gamma is already adjusted to satifaction, I'd adjust each colour in turn for hue, saturation and brightness first, then recheck. I actually wonder if there will be much need for a second pass (though I'd maybe recheck at 75% if I calibrate at 100% or visa versa just to achieve consistant results).

As I've already got my greyscale and gamma sorted using just the HD350's controls, I will probably just concentrate on using the VEQ for CMS work, perhaps saving custom settings with different gamma curves, but sharing the same CMS settings, so I can select between perhaps 2.2 gamma and maybe 2.4/2.5 gamma depending on the content.

I can't wait for my 'Pro to come now.