Great synopsis of a complicated mixing of several elements into one stream of data. The engineers who figured this all out way back when it was just black and white were simply brilliant. Even more remarkable was the addition of color. I remember the first RCA color TVs (I was born in 1947, so you understand my point). They were horrible by today’s standards, but a marvelous new toy for the living room. I doubt I will ever delve into this, but enjoyed the lesson. Thanks for sharing.
As I understand only the side bands are sent as I and Q. They correspond to R-Y and B-Y. As I understand the oscalator vibrates and the difference in phase between those and the color burst are the colors and saturation by amplitude. After mutlplying the sine and cosine then you should get YIQ which the tv turns to RGB. Dot crawl is another common artifact where color is interpreted as luminance information, especially in quick changes of color. Its a beautiful though complex procedure. Makes me appreciate that CRT in the garage alot more the work it did for you was amazing
Perfect explanation John! Back in the old TV repair days, you could use a vector scope (oscilloscope with H input - 3.58 Mhz color-burst, V input - chrominance sub-carrier) and get a Lissajous pattern that made it easier to see the phase difference for the color hues.
Please note that US HDTV is still 59.94 including the two 720P progressive networks, ABC and Fox. The reason is NTSC compatibility. When HDTV was introduced in 1999, there were few HTDV sets out there. The gross majority of the viewers still had NTSC receivers. True HD shows were also down-converted to NTSC and NTSC material was up-converted to the HD standard in use. For multiple reasons in a broadcast facility, the signals must have the exact same vertical sync rate for simulcasting. This legacy problem still exists to this day. In the 4K/UHD standards, the finally dropped 59.94 for 60.00. Note that on the production side most episodic shows and film originated material is done at 1080P/23.98. This means the master tape (or digital files these days) is compatible with any world TV standard including the legacy NTSC and PAL variants. The broadcasters never see 23.98 or 24p as the release master is converted to the format the broadcaster uses - world wide.
The process of frequency interleaving the chroma into the luma is responsible for the color burst shifts forward 90 degrees not 180 degrees per field. After four fields it arrives back in phase. Hence there is a four field sequence. Note that there is a 180 degree reversal line to line!
yup, it'was called "color framing" and an annoying problem editing with direct NTSC color record broadcast VTRs. Often the edit point was a frame off from where you wanted it to maintain the color field sequence. If ignored there was a 140ns H shift in the image at the edit. But if the edit was not a match frame, the production industry often did a non color framed edit anyway as with a scene change, you would not notice the H shift. And then there was 3/2 pulldown from film sources - that's another story. Note that PAL had an eight field sequence!
Do I and Q start at 90 degrees? Then they are not in the colorburst? I'll my guess is it shows the start color for that line and that allows the circuit s to compare them to the color burst
I seem to remember something about the color burst changing phase 180 degrees in PAL but not NTSC. It seems that without the phase shift then the PLL that keeps the internal reference oscillator on frequency would allow the oscillator to be on frequency but slightly out of phase. Thus NTSC="Never The Same Color". By reversing the burst phase the internal reference oscillator is kept on frequency and in phase.
@@JohnsBasement Lol, you first heard that one now, in 2023? I got another one for you.. CP/M stands for "conspiracy to preserve the ministry".. because evidently Gary Kildall admired the arcane and eldritch OSes and job control languages of 60's mainframes and wanted to cultivate a similar priesthood of experts in the microcomputer world. I heard that one back in the 80's.
Going a bit deeper into NTSC theory. the chroma signal is NOT exactly a 3.58mhz sinewave. NTSC uses DSB-SC Double Sideband Supressed Carrier modulation for the color difference signals. The 3.58mhz AM carrier is not transmitted except for the color burst. Only the modulation sidebands are sent as the chroma signal. But the receiver (TV) needs a constant raw 3.58mhz carrier to demodulate the color difference signals. This is the reason for the color burst. Just as H & V sync frequency locks the receiver scan circuits, the color burst locks a local 3,58mhz oscillator for both frequency and phase in the receiver at each scan line. If the chroma signals were modulated as a basic AM radio signal, the picture artifacts would be horrendous. This was one of the problems encountered (there are a few more) with the spectrum interleaving of the chroma with the luminance signal. They had to get the energy level of the chroma signal lower and the use of DSB-SC obtained that. The two color difference signals are a complex mixture of the RGB signals and there are multiple standards to do this function, I & Q, R-Y & B-Y, but for the utmost simplicity think of the color difference signals as red and blue with the luminance a combination of RGB. Using basic algebra circuits we can extract the green signal, out of the luminance at the receiver.
The people that came up with these things and got them all working with the commodity technology they had at the time are geniuses! These days, I get to make pretty pictures with VGA and ONLY have to care about the hsync & vsync! (yeah... I know I am ignoring gamma correction. It'll look good enough for a terminal screen and some video games!)
5.5 MHz versus 4.2 bandwidth can just about make or break readible subtitles that are not tiring to watch a whole night, without too much dot crawl. Many PAL countries also happen to need to subtitle their foreign content.
Can you explain to me why i get the signal on different frequencies. I thought it was Amplitude modulated. It looks like fc - fsc for v_sync and fc for blanking and fc_fsc for white.
The color burst? It is both amplitude and phase modulated. You want to ask this on the Z80Retro discord channel. There are some very good experts on this subject that lurk in there!
I think I get the color burst thing a little better now. So for each color pixel you have a phase offset on the visible horizontal line. So how does it 'fit' that 3.58mhz for each pixel?
What do you mean by "fit?" There's just one thing changing here -- the voltage level over time. So at 54mV you see black, at 1.0V you see white. You only see color when you "wiggle" that voltage close to 3.58 million times per second. It's just one continuous signal as you stripe across the screen, and sometimes it wiggles (color) and sometimes it doesn't (pure black or white). "Pixels" aren't a real concept in analog video (a pixel is a digital concept -- I arrange a buffer of colors I want in a 2d array of pixels for example). The *effective* rate that you can change your voltage level on an analog signal is what loosely corresponds to a pixel. In other words, if you have a really slow, kind of ineffective video device (like say the Atari 2600), it can't change that signal as fast as the TV is able to, and so it's "pixels" are constrained by the speed at which it can generate a scanline with a given color wiggle, and thus those old devices looked "blocky" when they produced video. If you're wondering what the highest *theoretical* speed at which you could change that signal, he explained that in the video -- it's somewhere in the 300 - 650 times per scanline range (it varies because different tvs will act differently -- some old vacuum tube driven TV probably can't do better than 300 changes per scan line, but a brand new circa 1990 tube tv with lots of digital components might be able to show 650 changes per scan line). Above that 300 - 650 changes per line, your TV's analog circuitry won't be able to keep up, and it will smear them in some way (producing incorrect colors, or simply not displaying an individual change you had in your voltage stream).
How many 3.8MHZ time periods can fit into one active display scan line? Once you exceed changing the luma at that rate, it is likely that a TV can't differentiate between interpreting it as a color and as a brightness.
@@Calphool222 What I am trying to understand is how that 'wiggle' ( I assume a sine wave) determines the color. I get it has to do with the phase of that 'wiggle' but lets say I wanted to use a micro controller, like a Pi Pico to generate this signal, and I want to get at least 8 bit color up on the screen, I need to wiggle that signal at a phase difference from the color burst? I get that the color burst is telling the TV that a color signal is incoming so is that color burst frequency for each 'pixel' color just a phase difference from the original color burst? I do understand that with a micro controller you wont be generating a sine wave but a square wave which I imagine should work, no?
@@JohnsBasement Well... say you had a signal capable of 8 bit color, 300 or more 'pixels' on a single 15.7khz ...and lets round that 'pixel' count to a standard like 320, you will need to modulate that color burst at the correct phase for each 'pixel' at around 49.2(rounded up) times per scan line if my math is correct. So I am guessing that 'wiggle' per 'pixel' would have to fit in at the proper phase for each color 'pixel' within that 1/49th for each 'pixel'?
@@GORF_EMPIRE You kind of have to understand a little bit about AC signals to grasp what "phase difference means." Imagine any AC waveform. If you say it is at 0 degrees, that means it starts at position 0 and swings up and down at whatever its frequency is. If you say it is at 90 degrees, then it will start part way through a swing (not at 0). When you compare that waveform to another waveform at the same frequency, they either overlap (their phases are the same) or one precedes the other. You can *measure* how much one precedes the other, and get a "phase difference." Now, imagine a color wheel. In the center is gray, and all around the outside are all the colors of the rainbow. The gray bleeds into the colors. They are most radiant at the outside and most gray in the center. Any color in the NTSC scan line signal can be represented as a point in this color wheel. The *phase difference* tells you the angle on the color wheel, and the *amplitude* tells you how far out toward the outside edge (where the color is most vibrant) of the color wheel to go.
Born in 99 trying to get into analog video, this is the single most educational resource I’ve found
Glad I could help! 😀
Great synopsis of a complicated mixing of several elements into one stream of data. The engineers who figured this all out way back when it was just black and white were simply brilliant. Even more remarkable was the addition of color. I remember the first RCA color TVs (I was born in 1947, so you understand my point). They were horrible by today’s standards, but a marvelous new toy for the living room. I doubt I will ever delve into this, but enjoyed the lesson. Thanks for sharing.
Yeah... I agree. Some seriously smart people designed this stuff!
@@JohnsBasement All with nary a DSP block - or transistor - in sight 🙂
This didnt happen in a snap. Like almost all technology, its a gradual improvement. Im not downplaying the fact that its isnt impressive though.
As I understand only the side bands are sent as I and Q. They correspond to R-Y and B-Y. As I understand the oscalator vibrates and the difference in phase between those and the color burst are the colors and saturation by amplitude. After mutlplying the sine and cosine then you should get YIQ which the tv turns to RGB. Dot crawl is another common artifact where color is interpreted as luminance information, especially in quick changes of color. Its a beautiful though complex procedure. Makes me appreciate that CRT in the garage alot more the work it did for you was amazing
Perfect explanation John! Back in the old TV repair days, you could use a vector scope (oscilloscope with H input - 3.58 Mhz color-burst, V input - chrominance sub-carrier) and get a Lissajous pattern that made it easier to see the phase difference for the color hues.
A prior employer threw one out. It was on its last leg. I looked at it and thought "should I?" Never thought I'd need it. Till now. 🙄
Thank you, John! Best video on the subject.
Excellent video and a must watch! Thanks for sharing
Thanks for saying!
Thank you for sharing your expertise. More please.
Thanks. I'll try not to let ya down!
Please note that US HDTV is still 59.94 including the two 720P progressive networks, ABC and Fox. The reason is NTSC compatibility. When HDTV was introduced in 1999, there were few HTDV sets out there. The gross majority of the viewers still had NTSC receivers. True HD shows were also down-converted to NTSC and NTSC material was up-converted to the HD standard in use. For multiple reasons in a broadcast facility, the signals must have the exact same vertical sync rate for simulcasting. This legacy problem still exists to this day. In the 4K/UHD standards, the finally dropped 59.94 for 60.00. Note that on the production side most episodic shows and film originated material is done at 1080P/23.98. This means the master tape (or digital files these days) is compatible with any world TV standard including the legacy NTSC and PAL variants. The broadcasters never see 23.98 or 24p as the release master is converted to the format the broadcaster uses - world wide.
The process of frequency interleaving the chroma into the luma is responsible for the color burst shifts forward 90 degrees not 180 degrees per field. After four fields it arrives back in phase. Hence there is a four field sequence. Note that there is a 180 degree reversal line to line!
yup, it'was called "color framing" and an annoying problem editing with direct NTSC color record broadcast VTRs. Often the edit point was a frame off from where you wanted it to maintain the color field sequence. If ignored there was a 140ns H shift in the image at the edit. But if the edit was not a match frame, the production industry often did a non color framed edit anyway as with a scene change, you would not notice the H shift. And then there was 3/2 pulldown from film sources - that's another story. Note that PAL had an eight field sequence!
Do I and Q start at 90 degrees? Then they are not in the colorburst? I'll my guess is it shows the start color for that line and that allows the circuit s to compare them to the color burst
I seem to remember something about the color burst changing phase 180 degrees in PAL but not NTSC. It seems that without the phase shift then the PLL that keeps the internal reference oscillator on frequency would allow the oscillator to be on frequency but slightly out of phase. Thus NTSC="Never The Same Color". By reversing the burst phase the internal reference oscillator is kept on frequency and in phase.
It changes in NTSC.
I misspoke when discussing WHEN the NTSC color burst changes. It is supposed to shift 180 degrees each FRAME. Not on each FIELD.
PAL stands for "People Are Lavender", due to the inaccurate color reproduction.
@@activelow9297 🤣
@@JohnsBasement Lol, you first heard that one now, in 2023? I got another one for you.. CP/M stands for "conspiracy to preserve the ministry".. because evidently Gary Kildall admired the arcane and eldritch OSes and job control languages of 60's mainframes and wanted to cultivate a similar priesthood of experts in the microcomputer world. I heard that one back in the 80's.
The French SECAM was "system essentially contrary to American methods", and another one for PAL - "peace at last".
Going a bit deeper into NTSC theory. the chroma signal is NOT exactly a 3.58mhz sinewave. NTSC uses DSB-SC Double Sideband Supressed Carrier modulation for the color difference signals. The 3.58mhz AM carrier is not transmitted except for the color burst. Only the modulation sidebands are sent as the chroma signal. But the receiver (TV) needs a constant raw 3.58mhz carrier to demodulate the color difference signals. This is the reason for the color burst. Just as H & V sync frequency locks the receiver scan circuits, the color burst locks a local 3,58mhz oscillator for both frequency and phase in the receiver at each scan line. If the chroma signals were modulated as a basic AM radio signal, the picture artifacts would be horrendous. This was one of the problems encountered (there are a few more) with the spectrum interleaving of the chroma with the luminance signal. They had to get the energy level of the chroma signal lower and the use of DSB-SC obtained that.
The two color difference signals are a complex mixture of the RGB signals and there are multiple standards to do this function, I & Q, R-Y & B-Y, but for the utmost simplicity think of the color difference signals as red and blue with the luminance a combination of RGB. Using basic algebra circuits we can extract the green signal, out of the luminance at the receiver.
The people that came up with these things and got them all working with the commodity technology they had at the time are geniuses!
These days, I get to make pretty pictures with VGA and ONLY have to care about the hsync & vsync! (yeah... I know I am ignoring gamma correction. It'll look good enough for a terminal screen and some video games!)
5.5 MHz versus 4.2 bandwidth can just about make or break readible subtitles that are not tiring to watch a whole night, without too much dot crawl. Many PAL countries also happen to need to subtitle their foreign content.
I look at NTSC today and can't believe how bad it looks after 20 years of 1080p. How did we EVER survive??? :-D
Amazing explanation 👍👍👍👍👍
Glad you liked it
Great tutorial !...cheers.
Thanks!
Can you explain to me why i get the signal on different frequencies. I thought it was Amplitude modulated.
It looks like fc - fsc for v_sync and fc for blanking and fc_fsc for white.
The color burst? It is both amplitude and phase modulated.
You want to ask this on the Z80Retro discord channel. There are some very good experts on this subject that lurk in there!
I think I get the color burst thing a little better now. So for each color pixel you have a phase offset on the visible horizontal line. So how does it 'fit' that 3.58mhz for each pixel?
What do you mean by "fit?" There's just one thing changing here -- the voltage level over time. So at 54mV you see black, at 1.0V you see white. You only see color when you "wiggle" that voltage close to 3.58 million times per second. It's just one continuous signal as you stripe across the screen, and sometimes it wiggles (color) and sometimes it doesn't (pure black or white). "Pixels" aren't a real concept in analog video (a pixel is a digital concept -- I arrange a buffer of colors I want in a 2d array of pixels for example). The *effective* rate that you can change your voltage level on an analog signal is what loosely corresponds to a pixel. In other words, if you have a really slow, kind of ineffective video device (like say the Atari 2600), it can't change that signal as fast as the TV is able to, and so it's "pixels" are constrained by the speed at which it can generate a scanline with a given color wiggle, and thus those old devices looked "blocky" when they produced video.
If you're wondering what the highest *theoretical* speed at which you could change that signal, he explained that in the video -- it's somewhere in the 300 - 650 times per scanline range (it varies because different tvs will act differently -- some old vacuum tube driven TV probably can't do better than 300 changes per scan line, but a brand new circa 1990 tube tv with lots of digital components might be able to show 650 changes per scan line). Above that 300 - 650 changes per line, your TV's analog circuitry won't be able to keep up, and it will smear them in some way (producing incorrect colors, or simply not displaying an individual change you had in your voltage stream).
How many 3.8MHZ time periods can fit into one active display scan line? Once you exceed changing the luma at that rate, it is likely that a TV can't differentiate between interpreting it as a color and as a brightness.
@@Calphool222 What I am trying to understand is how that 'wiggle' ( I assume a sine wave) determines the color. I get it has to do with the phase of that 'wiggle' but lets say I wanted to use a micro controller, like a Pi Pico to generate this signal, and I want to get at least 8 bit color up on the screen, I need to wiggle that signal at a phase difference from the color burst? I get that the color burst is telling the TV that a color signal is incoming so is that color burst frequency for each 'pixel' color just a phase difference from the original color burst? I do understand that with a micro controller you wont be generating a sine wave but a square wave which I imagine should work, no?
@@JohnsBasement Well... say you had a signal capable of 8 bit color, 300 or more 'pixels' on a single 15.7khz ...and lets round that 'pixel' count to a standard like 320, you will need to modulate that color burst at the correct phase for each 'pixel' at around 49.2(rounded up) times per scan line if my math is correct. So I am guessing that 'wiggle' per 'pixel' would have to fit in at the proper phase for each color 'pixel' within that 1/49th for each 'pixel'?
@@GORF_EMPIRE
You kind of have to understand a little bit about AC signals to grasp what "phase difference means." Imagine any AC waveform. If you say it is at 0 degrees, that means it starts at position 0 and swings up and down at whatever its frequency is. If you say it is at 90 degrees, then it will start part way through a swing (not at 0). When you compare that waveform to another waveform at the same frequency, they either overlap (their phases are the same) or one precedes the other. You can *measure* how much one precedes the other, and get a "phase difference."
Now, imagine a color wheel. In the center is gray, and all around the outside are all the colors of the rainbow. The gray bleeds into the colors. They are most radiant at the outside and most gray in the center. Any color in the NTSC scan line signal can be represented as a point in this color wheel. The *phase difference* tells you the angle on the color wheel, and the *amplitude* tells you how far out toward the outside edge (where the color is most vibrant) of the color wheel to go.
The Apple ][ famously used the NTSC color anomalies to its advantage - giving it more available colours than would have been possible.
Go Woz!
😃
So that's how they hacked colors with black and white. Sneaky!
Yep. And it looks pretty decent for the camera technology of the time. It is unfortunate that crisp vertical edges are so troublesome.
Trying to understand NTSC as a Z-Gen... Ironic
@@katarsisAX 😂
I give normies a pass when it comes to USB since everybody ends up flipping more than once a lot. I suspect USB A exists in quantum superposition.
😂
Hi. Could you help to make an analog video signal scrambler?
Sorry. That is not my thing.