Gamma correction. Do we need it?

Forum for discussion of narrow-bandwidth mechanical television

Moderators: Dave Moll, Andrew Davie, Steve Anderson

Gamma correction. Do we need it?

Postby Steve Anderson » Tue Jun 26, 2007 7:46 pm

Being the sort of guy I am I thought I might just stir things up a bit....

Now, before I go any further, please read and digest the following article that was published in Vol. 32 No. 1 of the newsletter which I have scanned in. Please read it first before you return here......




I cannot agree with Garth more. Period. There is no need for Gamma correction in NBTV except to undo the correction that is applied in Camcorders, off-air signals and the like. The signal-to-noise ratio thing (unless you're on HF) is irrelevant.

I did publish an item on Gamma correction in Vol. 28 No. 4. which at the time I took to be a de-facto standard, but it appears not.

Even a CRT is linear if it's current driven as opposed to voltage, have a look at the DG32-7 page also attached, if that's not a straight line, I don't know what is. It's sort of intuitive, and applies to LEDs as well. With all the discussion on resistors and LEDs, what are we trying to do? Turn a Voltage into a current.

So along with Garth (who doesn't seem to be a member of this forum) I do this time propose that we dump Gamma correction.

Steve A.

(I'm going to leap over the fence and watch the fireworks display!).
Attachments
gamma 1.gif
gamma 1.gif (132.34 KiB) Viewed 15856 times
gamma 2.gif
gamma 2.gif (275.01 KiB) Viewed 15856 times
DG732_8.JPG
DG732_8.JPG (75.74 KiB) Viewed 15856 times
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5361
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Ban the Gamma!

Postby Stephen » Tue Jun 26, 2007 11:33 pm

I agree wholeheartedly, Steve. There are enough issues with NBTV without needing to fiddle with gamma correction, which only has meaning with respect to the response of cathode ray devices to electrical potential. In my opinion it is better all around to have a nice linear NBTV signal for both display and synchronisation purposes.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Re: Ban the Gamma!

Postby Steve Anderson » Wed Jun 27, 2007 12:52 pm

Stephen wrote:I agree wholeheartedly, Steve.


Thanks for that.

You'll also notice where Garth refers to the timing of sync pulses, back porches and the like that microseconds (µs) have got turned into milliseconds (ms).

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5361
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby AncientBrit » Thu Jun 28, 2007 5:50 pm

Steve,

I agree phosphor output is linear wrt to beam current.

But if you look at grid volts against light output, this approximates to a square law because of the triode characteristics of the gun assembly.

And we normally drive a CRT from a voltage source.



Graham
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Currently I'm being driven.

Postby Steve Anderson » Thu Jun 28, 2007 10:21 pm

AncientBrit wrote:...And we normally drive a CRT from a voltage source.

Graham


Yes, very true, but there's no reason not to current drive a CRT, it's just been difficult (nay impossible) to do up until the advent of transistors and opto-couplers. I'm working on it now for my SSTV stuff, but will apply equally to NBTV.

You get automatic Gamma correction, in the sense that now the tube has a Gamma of 1.0, no different to a LED.

Bye bye Gamma!

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5361
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby AncientBrit » Thu Jun 28, 2007 10:25 pm

Very neat Steve, certainly a new approach.

Graham
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Grounded grids.

Postby Steve Anderson » Fri Jun 29, 2007 1:29 pm

AncientBrit wrote:Very neat Steve, certainly a new approach.

Graham


I don't claim originality for it athough I have never seen any reference to it. It's a modification of a grounded-grid amplifier except the cathode of the CRT is driven with a current, not voltage.

But I keep thinking at the back of my mind, "There must be something fundemental I'm missing", but I'll find out soon enough.

And I think I have just tripped across it.....

I was doing a drawing to attach to this message and as I was I saw the flaw, it's amazing how doing a drawing focusses ones' mind. Note the italics.

OK, so you bolt the control grid to -570V (in this case) and with the current drive the cathode can go from approximately that Voltage (maximum current and brilliance) to -470V (cut-off) which is limited by a Zener. All fine and well.

But what about the focus electrode? That needs to be at a consistant voltage wrt the cathode, which now it's not. So the focus will vary with modulation.

Time for plan 'B'.

This in some ways is easier than the above. The idea is to PWM the CRT, either it's producing a stream of electrons or it's not...just one focus point, no problems with lineararity, and we get our Gamma of 1.0.

For NBTV the PWM frequency would have to be at least in the order of 40kHz chopping each line up into 100 pulses, 100kHz would be better but at this stage I'm not sure if the CRT will respond to this degree of modulation. It is just a scope tube, not a TV tube which would be fine.

Both of the above ideas are at this stage conceptual, but I intend to make a start very soon. If anyone can see any flaws beyond the one above, please let me know your thoughts.

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5361
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby AncientBrit » Fri Jun 29, 2007 5:44 pm

Steve,

Jeremy Jago demonstrated an interesting concept at a recent convention.

He assumed that the scope being used as a display device had no access to Z mod.

Instead he used PWM but deflected the image to one side with a duty cycle depending on modulation.

Net effect were two displaced images, one positive, one negative.

Grey scale seemed okay.

Not sure how this would tie in with gamma though.

Over to you!

Cheers,


Graham
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Thanks are due....

Postby Steve Anderson » Fri Jun 29, 2007 6:07 pm

AncientBrit wrote:Jeremy Jago demonstrated an interesting concept at a recent convention.

Graham


You know I had completely forgotten about that article, even though I do have a copy of that particular newsletter!

And I have to admit it's a better method than mucking around with grids and cathodes.

The Gamma should come out at 1.0 assuming the PVM is strictly proportional which Jeremy implied in the article. It's simple, intuitive and makes my ruminations redundant.

Thank you both, Graham and Jeremy, you've just made things a lot simpler for me, and equally applies to my SSTV stuff too.

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5361
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby AncientBrit » Fri Jun 29, 2007 10:40 pm

No problems Steve.

As to failing memory I suffer from that. Maybe it's charge migration....


Graham (I think)
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Postby Klaas Robers » Sat Jun 30, 2007 4:44 am

Gamma correction came into TV because CRT's have a squared response from Vg1 to beam current. This is a very simple mechanism: The diameter of the spot is lineairly dependent on the voltage, but the surface is the square of the diameter, so the beam current is the Vg1 squared.
Because there were many more monitors (TV-receivers) than cameras the correction was done in the (only) camera.

But there is more as came out afterwards:
Our eyes are sensitive to the radio of light. So the cantrast of brightness 1 in respect to 2 is equal to the sunjective contrast between 4 and 8. And also between 0,1 and 0,2 !! This implies that noise in dark areas is much more visible than noise in bright areas.
Gamma correction does the opposite. It amplifies the dark video signals and attennuates the bright video signals. Then it is transmitted. Equal noise is added. The picture tube attennuates the dark video signal, so attenuates the added noise and amplifies the bright video signals. That amplified noise is no problem at all because our eyes are insensitive to it.

The same is for digitalised video. Now noise is added due to quantisation. With gamma correction B+W video can be digitized into 6 bits, without gamma corection you need 3 to 4 bits more. Yes CD has bits enough, in black and white. But because of gamma correction some of you can store NBTV in 8 bits wave files and we could do the trick of CCNC.

Allas, even for Garth it should not be too difficult to build a gamma corrector in his Nipkow monitor, should it?
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Stephen » Sat Jun 30, 2007 6:32 am

Klaas Robers wrote:Our eyes are sensitive to the ratio of light. So the cantrast of brightness 1 in respect to 2 is equal to the sunjective contrast between 4 and 8. And also between 0,1 and 0,2 !! This implies that noise in dark areas is much more visible than noise in bright areas.
Gamma correction does the opposite. It amplifies the dark video signals and attennuates the bright video signals. Then it is transmitted. Equal noise is added. The picture tube attennuates the dark video signal, so attenuates the added noise and amplifies the bright video signals. That amplified noise is no problem at all because our eyes are insensitive to it.
Similarly, our ears are logarithmically sensitive to sound, being able to distinguish subtle differences in intensity at low levels but hardly at all at higher levels. Likewise for audio systems, spurious noise due to the recording or transmission medium is more noticeable than if our ears had a linear response curve.
Since at least the 1930s, various compressor-expander, or “compander” schemes have been tried to improve the signal-to-noise ratio of audio recordings and radio channels. They may be very effective in improving signal-to-noise ratio. This is an exact analogy to using “gamma correction” for video, that is, compressing the video level in a logarithmic or square law fashion on the recording or transmission side and expanding the video level in a complementary manner on the reproducing or reception side.
The problem with any compander system, and the reason it never has been used for serious audio transmission and recording purposes, is that the companding process is inherently non-linear. A compander system may compress and expand a single sine wave signal perfectly. However, with a complex waveform that comprises a number of sine waves of different frequency and intensity, the square law characteristics of the compression and expansion circuits inherently create modulation products because they are by definition square law modulators.
Now the modulation products that result from the audio companding process are not necessarily objectionable depending upon the source material, but anyone with any sort of ear can hear the difference. The gamma factor circuitry will generate the same sort of modulation products in a video signal.
For instance, if you have a test pattern that generates a video signal that has a complex waveform comprising a 4000 Hz sine wave and a 500 Hz sine wave, the gamma circuitry is going to modulate the two waves to generate 3500 and 4500 Hz modulation products as well. These modulation products are going to show up on the display as some sort of interference pattern. Whether or not they are going to be objectionable is another matter and of course will change depending upon the combination of frequencies, but in any case you are not going to get out what you put in. That is the nature of the compander process.
In the case of cathode ray television there was not much choice, because we had to deal with the reality of the non-linear characteristics of the cathode ray pickup and display devices. With NBTV of the optical scanning type we have a choice.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby Klaas Robers » Sun Jul 01, 2007 11:59 pm

Oh no Stephen, gamma correction is totally different from compression/expansion. With compression/expansion the amplification changes (slowly) with the loudness of the audio on a certain moment. On a certain moment the amplification of the compressor and/or the expander is almost constant. This is to eliminate intermodulation.

In the gamma corrector the amplification is changing instantaneous with the voltage of that moment. This causes enormeous intermodulation, but if you do the opposite process after that (dégamma) then the intermodulation is cancelled out. More or less. For audio this could be done and looks than as soft clipping. I think in analogue satellite radio something like that is done, I think it was name something like Panda Wegener. In the receiver the opposite process is performed.

Intermodulation generates frequencies (sum and difference) that were not there originally. Our ears are very sensitive to that because they are organised as parallel filter banks. So the ears work in the frequency domain.
Ears are very sensitive to non linear distortion, but won't almost observe differences in delay time for different frequencies (phase errors).

Our eyes work in the time domain, pixel for pixel. They are not sensitive to (spatial) frequencies at all. In a video signal spatial frequencies map to normal frequencies and reverse. But our eyes won't see that.
Eyes won't observe non linear distortion in the video signal, but are very sensitive to differences in delay time or phase errors.
Last edited by Klaas Robers on Mon Jul 02, 2007 4:07 am, edited 1 time in total.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Stephen » Mon Jul 02, 2007 12:48 am

Klaas Robers wrote:Oh no Stephen, gamma correction is totally different from compression/expansion. With compression/expansion the amplification changes (slowly) with the loudness of the audio on a certain moment. On a certain moment the amplification of the compressor and/or the expander is almost constant. This is to eliminate intermodulation.
Yes, Klaas, I agree that more sophisticated compander systems have a slow-responding amplification gain change to minimise intermodulation distortion. The DBX system is an example. However simple companders are simply non-linear amplification systems. For instance, a super-regenerative detector has such an instantaneous compression characteristic to provide an "automatic volume control"-like action. That is why super-regenerative detectors always have a high level of background noise. They amplify low levels much more than high levels.

Also, thermionic valves and FETs have a square-law characteristic and such amplifying devices have a highly non-linear effect that can do the same thing. That is why any high quality amplifiers that employ such devices have to have negative feedback circuitry. It seems to me that the gamma factor circuitry by its very nature has to introduce such intermodulation distortion into a video signal for the same reason. As far as the visual effect, it seems to me that would have to depend upon the particular mix of sine wave frequencies in the complex waveform involved.

For instance, the subcarrier frequency of a colour television signal is critical to avoid visual interference between the subcarrier frequency and the baseband luminance channel--because of intermodulation of the two. I believe that the subcarrier frequency has to be a multiple of half of the line frequency rate to minimise, but not eliminate, the visibility of such interference. Any other frequency causes visible interference.

This is one example of visible intermodulation interference. I just have trouble accepting that intermodulation products due to non-linear amplification would not have visual effects. I may be completely off-base, though.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby Klaas Robers » Mon Jul 02, 2007 4:05 am

The idea is that when the phase of the subcarrier (for a certain coloured area) inverts line by line that where in one line the voltage of the subcarrier is highest, so it gives a brighter dot on the screen, in the next line it is minimal and gives a darker dot on the screen. This should give a checkerboard-like pattern on coloured larger areas and should minimise visibility of the colour carrier itself. And then the next frame it should be inverse. This gives somewhat difficult correlations between the carrier frequency and the line / frame frequencies.

However, if your eyes are following soemthing moving on the screen this trick works just half. And even then, because of the gamma of the CRT the effect of brightening dots is stronger than the effect of darkening dots. A nett brightening effect remains.

But because the viewer doesn't know which brightness is the correct brightness he doesn't observe it. If the B+W TV has a switcheable notch filter that supresses the subcarrier sine wave you will see that with the notch filter these areas become somewhat darker. I have seen that in the past. Impressive.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Next

Return to Mechanical NBTV

Who is online

Users browsing this forum: No registered users and 31 guests