Bringing Up Blacks From Black Clipping???
 

I'm finding this issue a little difficult to describe so let me first give the facts of the shoot. It was a very bright high contrast scene. Bright white clouds, white cars, black tarmac, shadows in door way and close-up bushes in full shadow. I exposed about half a stop below clipping so have full detail in the clouds. I get same result as described here for all CINE gammas in default settings and in all sorts of scenes.

As shot the darkest part, the bushes show minimal detail as you would expect and I'm trying to bring out a little of this detail with the Color Curves but here is the problem. The strong blacks cut-off sharply just like they clipped. IE the left side of the histogram has a very high value at 0 (zero luminance) and the drops to nothing. The Waveform monitor has a thick line running all along the 0 bar. Trying to drop the end by the lowest blacks even with very fine adjustment seems to drop all the blacks, not just the very bottom, and produces a lot of noise.

What is happening here and why is the EX1 histogram in bright contrasty scene not like a bell curve that tapers at both ends but like a rectangle where blacks (and to some extent white) drop off like the side of a cliff? With HDV I never saw anything like this as I always had gradual drop-off at each end of the luminance scale. I'm really confused as to just what the camera is recording and to how Vegas seems it.
COULD this be due to under-exposing. The picture quality unprocessed is superb, I just want to soften contrast to be more film like.

There are a lot of pixels in the black region of the image and that is what the histogram shows; namely how many pixels are at each brightness. The camera histogram puts black a bit to the right of the apparent zero and in an underexposed image you get that cliff at the black end because that's where all the pixels are. If you over expose it becomes a gradual decline because there aren't many pixels that are dark.
In Photoshop your posted image gives the usual histogram I'd expect (there are a lot of blacks present and no cut-off -- apart from a number down at 0) and the amount of detail in the dark hedge is quite surprising (considering the brightness range in the subject). You can bring it up with levels, but that doesn't help the rest of the image. I would use masking to bring up only the hedge using levels, but noise will be a problem with such an underexposed area.
Looks pretty good, to me.

Thanks Serena. Isn't the exposure latitude of this camera amazing!

Hi Michael,

What did the left side of the EX1 histogram look like when shooting? Was it registering any "0" pixels. How were you guaging white clipping? If zebras, at what setting?

This grab looks very saturated for stock settings. Was that something you did in post? If so, could you post an unaltered out of camera shot?

Very interesting thread!

NO POST - .mxf straight from camera. Frame grab from Vegas 8 32-bit time-line.

Matrix was set to HiSAT. Otherwise straight CINE4 with black master -3 and black gamma -2. 1080x 24p

The EX1 histogram was NOT vertical immediate cut-off like I saw in Vegas.

Zebra set at 95 with zebras just gone. That's why I say it is a little underexposed.

Yes, the under exposure is clear in the Photoshop histogram. Working to the 109% limit offers better dynamic range. Or, if you like, less noise in darks. Apply "levels" in post.

Thanks for the extra info. If I understand correctly, Zebra set to 95 will start to register Zebra indications at 90. If you backed it off even more from there to be safe, that is likely the issue.

Not sure why the difference in the two histograms.

This is the disadvantage of using Zebra 1 only - it show a range rather then above threshold values, which may be confusing. Pity that Zebra 2 is not adjustable, but still I prefer using it alone - at least I know that "just gone" means I'm around 100 IRE.

Zebra 1 is excellent for checking e.g. human faces and alike, within the allowable max range.

My understanding is that zebra 2 also shows a range (95-105). Is that incorrect?

According to the (Euro) manual, page 40:

2 (Zebra 2): to display a zebra pattern for the video level over 100%.

Thanks Piotr,
I hadn't caught that. That is very helpful.

For Zebra 1, the manual is stating "To display a zebra patttern in the area of +/- 10% centering the video level set with Zebra 1 Level". This reads to me like if Michael had set to 95% that he would start seeing Zebra indications at 85%. Some had said before that it was a 10% range (5% up, 5% down), but this reads to me like 20%.

I think my 95 equates to 90-100 which in reality, when using it at the top end is 90. I realize now that Zebra 1 is meant for a RANGE in the lower areas as for skin and hence the default of 70. For clipping I now believe i was doing it wrong and must use the Zebra 2 set at 100. Today I will repeat my tests using Zebra 2.

You have zebras and you have the histogram, the latter being far more useful.

Serena; Did you read my latest post Zebra107= etc ... because if you really find the histogram good maybe my camera has a problem? As I said in that post MY histogram seems useless (except by re calibration) in that when I reach white clipping at 107 my histogram still has three more vertical bars to go! Tell me how your histogram is and how you use it so I can figure if I have a problem but I have never seen a histogram that looks like this one. My white clip, the spot meter says 100 and the Zebras say 107 and my histogram is only at about 80% max?

Regardless of how far to the right the histogram goes surely it's self evident when you hit clipping?
Start by underexposing and open the iris, the more of the graph that disappears to the right of the graph the more you are clipping. Same goes for clipping at the blacks. Remember that a completely black or completely white frame only shows as a single, thin vertical line on a histogram. Keep in mind that a histogram is a statistical plot of the population of pixel values and the old saying about statistics applies unless you take some care in how you read the results.

Bob: I'm saying my histogram is not disappearing at clipping! The histogram is not showing clipping at the same time the image does. When the last line start climbing the histogran is three lines past clipping. IF I want to use the histigram to set a clipping point I must open until the histogram starts to dissapear as you say, then back off until the right most three verticle lines have gone.

Keep in mind the histogram is also measuring the quantity of pixels at that level. If you have only a few peaking pixels you may very well not see it at all on the histogram.

Point the camera at a large object with a major amount of pixels and you should see that last bar grow tall.

I really think the histogram is not the best toll for measuring peaks since that's not its purpose. That's what Zebras and the center numeric are for. Histogram is good for measuring contrast/latitude. I'll often use it as a second check to see if i need to tweak iris a bit for a bit better "spread" (depending on the look I'm going for).

How do you know the image is clipping?
If you're using the Vegas scopes in Studio RGB and assuming 100IRE is clipping I think you're looking at this incorrectly. There does seem to be something a bit wierd going on. I deliberately overexposed a piece of white card to the point that the camera is going into "Too Bright" alarm and the scopes in Vegas only read 95% (in ComputerRGB).

If I switch to 32bit/2.2 processing then the scopes show a 100% flat line! Really looks like Ive clipped it completely into digital oblivion.

If I add a ComputerRGB to studioRGB levels FX in the Preview window I'm back to what I had before with the scopes reading 95%.

Ill post some screen shots tomorrow. If I increase exposure until the last line of the histogram appears then render that image and play it on a monitor (to remove the Vegas scopes from the equation) half my image will be gone to white. Maybe tomorrow I'll drop to 760p and connect DVRack up via firewire and compare scopes.

Just my opinion, but I think you guys are putting way too much attention on the histogram. It just doesn't tell you that much.

I have seen that on a waveform the IRE indicator in the viewfinder was off maybe as much as 10 points. I can't remember which way but I'll look at it again tomorrow. I'll also check the zebras against a waveform.

Michael, to use firewire you must "drop" to SP (1440x1080), not 760p. The latter is still HQ, and the i.LINK is cut-off!

Michael,
I took the liberty to do a CC of the image you posted. Using Color Finesse in FCP I tried to open the shadows, soften the mid tones and add some saturation. I think that the sky is a bit too saturated for my liking...
I think the possibilities offered by this camera are incredible.

Hi Michael,

I think I know what is going on for you. The histogram will not reach the far right of the graph with certain PP's selected, as they are not producing any pixels of that brightness. An easy test. With your given settings, try to blow the highlights. Aim directly at a light bulb at close range so that it fills the screen. Look at the exposure meter and your histogram. Try again withh PP off. Report back.

Randy, that "The histogram will not reach the far right of the graph with certain PP's selected" is obvious, but - as always - provided one isn't heavily overexposing. You can drive the histogram towards the very RH end with ANY PP (with sufficient light, of course).

So, if Michael really means it that whatever he does, he NEVER sees his EX1's histogram filled at 100 (or is it really 108, the far RH end?), something is wrong with his histogram/camera. But frankly, I do not really think so; it just takes some practice to learn all the aspects of a quite complicated machine that the EX1 certainly is.

I can make my histogram behave in a quite predictable fashion, even though it took me several days to master it.

And one more thing: it's certainly more difficult to clip the whites with Cine4 engaged than it is with Cine1, not to mention STD3.

I believe CineGamma 2 prevents peaks from going beyond 100. It's meant for "broadcast safe" shooting.

I took the liberty to grade your image to my preference and to shou how much information is still contained in the blacks. You can see the detail in the bushes as well as the guy under the porch: pinon_graded.jpg.

I can't attach images to my post (yet), so no preview... I've used 601 output levels, so it is a little 'flatter', but I like that ;)

In-camera a different PP and especially gamma setting should get you something like this from the start.

George/

Hi Piotr,
I do not think that is what is occurring for Michael, but maybe he can answer better. With PP2 selected, the histogram will never reach over 100. With some of Bill Raven's Cine based profiles for instance, When you are at the high side of proper exposure the histogram will start stacking up a few bars short of full right. It takes a major exposure boost (way out of poper exposure range) to blow past this barrier.

Lets get a basic definition out of the way. The histogram is in a box. Is there any significance to the position of the left and right side? Are the extremities of the the box at clipping? If so then I can say my histogram never approaches the sides that I have seen. What I see as clipping as per the Zebras or the Spot meter only uses the center 70-80% of that box.

Are you saying that if you set everything to manual, pp to off, gain to 18, iris to open, shutter to the lowest possible, ND filter to none, and aim at a bright light, that your histogram does not move to its extreme right?

Randy: NO. I am NOT saying that. Under the conditions you describe the histogram will go to the right but in this situation we are way past clipping.

What I am saying is the histogram is a long way from the right and not even to its maximum to the right as possible under the shooting conditions, when real clipping occurs. I am defining clipping as when the Zebras go to 108 and the spot meter goes to 100 and all white detail disappears.

IE to find the clipping point O need use Zebras or the spot meter or my eye. The histogram tells me nothing except a general indication of the dynamic range and this disappoints me.

mike

Mike,
When I brought the image in the CC software, there was a lot of headroom at the right of histogram in the highlights area: a comfortable one stop. The same info was shown in the waveform. Will try to post a screen grab of the histogram and wave form displays.

Valeriu: What is CC? Sounds like you under-exposed.

Mike,
I was refering to the image you posted on the forum. I have downloaded it and brought it in the Final Cut Pro and did the colour correction (CC) using ColorFinesse plug-in. I have uploaded the 'corrected' image and you can see the results in the attached file from my previous post.