Gamma!

Forum for discussion of electronic television. Generally, stuff to do with CRTs and not using mechanical displays.

Gamma!

Postby Panrock » Tue Oct 23, 2018 5:48 am

Guys,

It's time to come clean and 'out' with a big admission. I don't understand GAMMA. This is basic stuff, I know... No, really... I've been to various online sources to try to swot up on this and I'm now just getting more and more confused. I know you're a kindly and knowledgeable bunch here. Can any of you help? :roll:

I've put together the following understanding from the web. Some of it I may have got the wrong way round!

Gamma is the non linearity in light output that compensates for the non linear sensitivity of our eyes.

In most computer display systems and CRTs, images are fed in encoded with a gamma of about 0.45 and decoded by the device with the reciprocal gamma of 2.2.

When the gamma is 1, CRTs natively, as seen by the eye, look like they cramp to the whites with a linearly graded input. In other words, the brightness rises early in comparison with the volts in. The greys of a linear stepwedge should thus tend to look light.

Gamma correction is applied at the programme source stage.

Due to an odd bit of engineering luck, the native gamma of a CRT is 2.5 — almost the inverse of our eyes. Our eyes lift the darker greys up to look lighter than they really are. If our eyes were linear, a CRT display would look black crushed. This is the opposite of what I've said two paras above! Utterly confused!


With thanks,

Steve O
Panrock
Green padded cells are quite homely.
 
Posts: 870
Joined: Mon Feb 05, 2007 8:25 am
Location: Sedgeberrow, England

Re: Gamma!

Postby Klaas Robers » Tue Oct 23, 2018 7:38 pm

Steve, gamma started with the behaviour of CRT's. These tubes have close to the cathode a first grid in the form of a plate with a small hole in it. The voltage on the grid is made negative to prevent electrons to go through the hole.

Now, if we make the voltage less negative then the hole "opens". The diameter of the effective hole gets larger linearly with the voltage change. But it is the surface of the hole that defines how many electrons can escape through the hole. When the effective hole gets twice as large, this results in four times as much electrons and four times as much brightness on the screen. So the brightness on the screen is the grid voltage to the power of 2, .... aproximately.

This should be compensated for somewhere in the chain. Of course the total system should be linear. It was decided that this was done in the (B&W) cameras of 1936. There are many TV-sets and only one camera in the beginning. That is the origin of gamma, 1936, EMI.

So the correction circuit should boost the voltage change in the dark areas of the picture, because the electron gun of the picture tube boosts the brightness changes in the bright parts of the picture.

LCD panels and LED panels should behave in the same way as CRT's to have an over all compatibility of the evoluating system.

Our eyes are NOT non-linear. And if they are non-linear, this is compensated for in our brains. This is always how the human body works. We cannot measure the linearity of our eyes, as we have no access to the communication channel form the eyes to the brains.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Re: Gamma!

Postby Panrock » Tue Oct 23, 2018 9:02 pm

Klaas, thank you. I have never before seen the basic principle at work in the CRT explained as elegantly as that.

So, regarding my original question... Say a CRT grid just cuts off at -10v (negative). If you then put -5v on it, what sort of grey do you get, assuming linear measurement of the spot brightness? A lighter-than-mid or darker-than-mid grey? Although instinct supposes it will be lighter rather than darker grey due to the law of increasing squares, the opposite is actually the case! This is because there is even more expansion of the squares' values toward white, between -5 and 0v, than between -5v and -10v, meaning the mid grey is comparatively pushed down toward black.

And also... when you are looking at the grey spot, you are not 'linearly measuring' the spot brightness but are looking at it with your eyes! Your eye-brain system is non-linear, favouring expansion (better gradated perception) of the dark greys. So wouldn't this tend to counter the 'dark grey' rendering of a mid grey signal on a CRT, making the response look approximately linear, and restore the spot to looking mid-grey, without any gamma correction?

The above may be wrong. I am still making myself confused.

Steve O
Panrock
Green padded cells are quite homely.
 
Posts: 870
Joined: Mon Feb 05, 2007 8:25 am
Location: Sedgeberrow, England

Re: Gamma!

Postby ppppenguin » Wed Oct 24, 2018 4:23 am

A CRT natually has a gamma of between 2 and 3 as explained by Klaas. This is quite handy since the eye is more sensitive to noise near black level. So we want to devote more dynamic range (or numbers in a digital channel) to signals near black. Even if all imagers and displays were linear we would still want to impose somethng like gamma in the transmission channel for this very reason.

In an ideal theoretical world the transfer function from the televised scene to display would be linear. in practice this isn't always possible due to limitations in dynamic range. Also subjectively TV usually looks better with a gamma of slightly more than unity.

Gamma has very little to do with our eyes and everything to do with how TV systems work.
Jeffrey Borinsky www.borinsky.co.uk
User avatar
ppppenguin
Research Scientist
 
Posts: 44
Joined: Fri Nov 04, 2016 5:30 am

Re: Gamma!

Postby Panrock » Wed Oct 24, 2018 5:11 am

Nice to hear from you Jeffrey.

For me, there's something slippery and hard to pin down about the concept of gamma. For example... yes, our visual perception is most sensitive to distinguishing gradations (and thus noise) in the darker greys. Yes, the cathode ray tube progressively expands the difference between even gradations in signal level as it works up from black. However (as I currently understand it) this expansion effect gets ever greater as the signal approaches white... meaning in the overall scheme of things the darker greys appear crushed!

Steve O
Panrock
Green padded cells are quite homely.
 
Posts: 870
Joined: Mon Feb 05, 2007 8:25 am
Location: Sedgeberrow, England

Re: Gamma!

Postby Klaas Robers » Wed Oct 24, 2018 10:33 pm

Panrock wrote:So, regarding my original question... Say a CRT grid just cuts off at -10v (negative). If you then put -5v on it, what sort of grey do you get, assuming linear measurement of the spot brightness? A lighter-than-mid or darker-than-mid grey? Although instinct supposes it will be lighter rather than darker grey due to the law of increasing squares, the opposite is actually the case! This is because there is even more expansion of the squares' values toward white, between -5 and 0v, than between -5v and -10v, meaning the mid grey is comparatively pushed down toward black.


Indeed Steve, if -10 volts is cutoff, black on the screen, a brightness of 0 ,
and -0V is a fully open grid hole, white on the screen, a brightness of 1 ,
then -5 volts is a half open grid hole. half the diametre, 0.25 of the circle surface, so a brightness of 0.25 .

Panrock wrote:And also... when you are looking at the grey spot, you are not 'linearly measuring' the spot brightness but are looking at it with your eyes! Your eye-brain system is non-linear, favouring expansion (better gradated perception) of the dark greys. So wouldn't this tend to counter the 'dark grey' rendering of a mid grey signal on a CRT, making the response look approximately linear, and restore the spot to looking mid-grey, without any gamma correction?

Steve O


No Sreve, what happens in our eyes and brains is always in "the chain". There is no difference in looking to a flower directly or on the screen of a TV. At least that is how we want it to be. I reality we even don't know what happens in a human eye and what happens in the brain. This even might be different for different persons. But because they are fully used to the way THEIR vision system works, every body assumes that some one else sees the same as you. We even don't know if everybody sees RED as red (what ever that may be) and GREEN as green. We only call it by the same name because our mother told us so.

In first instance consider the gamma correction in a TV camera just as a compensation for the no linear behaviour of the CRT, without which we should not have television at all. This compensation costed one dual tube extra, which was too expensive to build it in every TV. So just only one in the (/ each) camera. In total the television channel is linear between white and black (= very dark grey).

And indeed, the correction of the gamma in the camera had the advantage that the (radio) transmission channel was better adapted to the equal visibility of noise / digital quantisation in dark areas compared to bright areas. This saved us in transmitter power a factor of 64 (in watts) and in bits 3 bits in word length. In a linear digital channel we should use 11 bits words to keep quantisation in dark parts invisible. But that is not how gamma correction was born.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Re: Gamma!

Postby Panrock » Thu Oct 25, 2018 3:00 am

Klaas Robers wrote:Indeed Steve, if -10 volts is cutoff, black on the screen, a brightness of 0 ,
and -0V is a fully open grid hole, white on the screen, a brightness of 1 ,
then -5 volts is a half open grid hole. half the diametre, 0.25 of the circle surface, so a brightness of 0.25 .


Right. So a half-level signal into a CRT will result in a dark-grey rather than mid-grey spot. Thanks.

Klaas Robers wrote:No Sreve, what happens in our eyes and brains is always in "the chain". There is no difference in looking to a flower directly or on the screen of a TV. At least that is how we want it to be. I reality we even don't know what happens in a human eye and what happens in the brain. This even might be different for different persons. But because they are fully used to the way THEIR vision system works, every body assumes that some one else sees the same as you. We even don't know if everybody sees RED as red (what ever that may be) and GREEN as green. We only call it by the same name because our mother told us so.


Indeed. This matter is dealt with in philosophy where I understand such subjective experiences are called 'qualia'.

Klaas Robers wrote:In total the television channel is linear between white and black (= very dark grey).


So whereas the CRT 'crunches together' the darker areas of the picture, the camera channel 'spreads them out'. Opposite gamma corrections. Right?

Klaas Robers wrote:And indeed, the correction of the gamma in the camera had the advantage that the (radio) transmission channel was better adapted to the equal visibility of noise / digital quantisation in dark areas compared to bright areas. This saved us in transmitter power a factor of 64 (in watts) and in bits 3 bits in word length. In a linear digital channel we should use 11 bits words to keep quantisation in dark parts invisible. But that is not how gamma correction was born.

Interesting. Thanks.

Steve O
Panrock
Green padded cells are quite homely.
 
Posts: 870
Joined: Mon Feb 05, 2007 8:25 am
Location: Sedgeberrow, England

Re: Gamma!

Postby ppppenguin » Thu Oct 25, 2018 5:38 pm

It can be tricky to see that gamma is used in 2 ways in TV.

1: To ensure that the overal transfer function is linear. Gamma = 1. Since CRTs have a gamma of around 2.5 it made sense to do it once at the TX, rather than in every RX.

2: Because of visibility of noise it makes sense to devote more dynamic range (or bits) to dark areas than light ones.

It happens, by chance, that the 2 requrements coincided rather well at the start of TV.

There is an interesting problem in the design of a gamma corrector at the camera. In theory you need infinite gain at black level to change gamma from 1 (for a plumbicon tube) to about 0.4 for transmission. in practice you increase gain up to a certain amount as you approach black and then revert back to linear. Makes no practical difference if you choose the transition point sensibly.

Then there's the whole business of constant luminance in colour systems. That's been a contentious subject over the years. It's very much a problem related to gamma correction.
Jeffrey Borinsky www.borinsky.co.uk
User avatar
ppppenguin
Research Scientist
 
Posts: 44
Joined: Fri Nov 04, 2016 5:30 am


Return to Electronic NBTV

Who is online

Users browsing this forum: No registered users and 12 guests