Shape of the output signal on TIA pin

[Windless]

Hi,

 

  Is there a documentation showing the shape of the signal on the COL pin ? I think LUM0/1/2 show 0/+5V (because they a reconverted to an analog value via aresistor ladder), SYNC probably do the same, but what about COL ? Does it output the 4.43361875 MHz modulated carrier ? Or does it just output a squarish signal (with a delay varying with the color, and staying 0 if color is black) that is later multiplied with the 4.433 signal ?

 

thanks,

Windless.

 

PS: i tryed opening a subject in the hardware part of the forum but maybe the subject will get more attraction here ?

[+DrVenkman]

Interesting question. If no one gets you an answer in the next few days, I'll pull my daily driver 4-switch Woody apart and take some scope traces for you. Pretty sure someone will come along with the answer before I can do it, however.

[Windless]

I wanted to try this myself but :

-I'm not very skilled with scopes, I used mine only twice

-My 2600 is SECAM, which means it has the B&W pin grounded, so nothing on COL if I understood correctly, so I either need a game that ignores B&W or to code something but for this i need some devcart. I could make one with a bluepill I guess.

 

The game I own are :

-Tennis

-Circus Atari

-Bump'n'Jup ("P" sticker)

-Real Sport Soccer (P)

-Defender (P)

-Real Sport Boxing

-Pole Position

-Jungle Hunt

-Off the wall

-Centipede (P)

-Forest

-Ms. Pac-man (p)

 

Is there a B&W ignoring game in this list ?

[alex_79]

20 hours ago, Windless said:

  Is there a documentation showing the shape of the signal on the COL pin ? I think LUM0/1/2 show 0/+5V (because they a reconverted to an analog value via aresistor ladder), SYNC probably do the same, but what about COL ? Does it output the 4.43361875 MHz modulated carrier ? Or does it just output a squarish signal (with a delay varying with the color, and staying 0 if color is black) that is later multiplied with the 4.433 signal ?

I only have very basic knowledge about electronics, so I can't be of much help on this.

What I know is that some SECAM boards revisions use a PAL TIA, some other a NTSC one. The colors are generated based only on the 3 LUM pins by extra circuitry on the board, the TIA COL pin is not used.

 

In theory you could convert your SECAM console into a PAL or NTSC one (depending on what TIA is installed on your system), but that requires replacing the crystal oscillator (and for PAL you need a second oscillator for color) and recreating the video circuitry using the PAL or NTSC schematic as reference.

CX2600P_Official_schematic.pdfCX2600-CX2600A_Official_schematics.pdf

 

In this thread, there's some info on the COL pin that might be useful:

Quote

[...]

Also, if everything else is buffered, you might as well buffer COL as well. I read a post where someone asked why this wasn't done, and I think LHE said it was an analog signal, but it is not - it is actually a digital signal, but delayed so it will be out of phase as the phase of the signal determines the color. [...]

 

Quote

[...] I looked at the TIA schematics, and it is most certainly an open drain digital output. It is asynchronous and fairly high-speed, however, so a gate delay might mess up the phase. [...]

 

Tia schematics (NTSC) can be found here.

 

 

5 hours ago, Windless said:

Is there a B&W ignoring game in this list ?

From that list, Bump'n'jump, RealSports Soccer, Realsports Boxing, Pole Position, Jungle Hunt, Centipede, Forest and Ms. Pac-man are unaffected by the TV TYPE switch and display always in color in PAL and NTSC consoles (that is, you should have a signal on the COL pin of the TIA chip).

"Off The Wall" uses the TV TYPE switch to pause the game. On a Secam console you can't access the pause feature in this game, but it is playable. It also displays always in color on PAL and NTSC systems.

 

Edited July 17, 2019 by alex_79

[Windless]

Thank you Alex

 

I did a few tests with Jump'n'Bump on pin 9 of my PAL TIA on SECAM 2600, and failed to observe a signal. I did find something on LUM2.

 

Possible problems are :

-Wrong use of the DS1054Z (frequency, mode, ... ?)

-Maybe a lacking signal on PAL-S / PAL-I pins or something unexpected on DEL causes the output on COL to be absent ?

In theory you could convert your SECAM console into a PAL or NTSC one (depending on what TIA is installed on your system), but that requires replacing the crystal oscillator (and for PAL you need a second oscillator for color) and recreating the video circuitry using the PAL or NTSC schematic as reference.

Instead of recreating the composite signal, one could also probe the COL and LUM0/1/2 with a raspberry pi zero (1gHz), and have an HDMI output. It would be interesting to see if the 1x, Ex and Fx color give no COL signal, or a COL signal that is out of scope (if the later is true,  it would be possible to use the full NTSC palette on PAL and SECAM consoles)

[Windless]

Thank you Alex

 

I did a few tests with Jump'n'Bump on pin 9 (COL) of my PAL TIA on SECAM 2600, and failed to observe a signal there nor on PAL-S or PAL-I. I did find something on LUM2.

 

Possible problems are :

-Wrong use of the DS1054Z (frequency, mode, ... ?)

-Maybe a lacking signal on PAL-S / PAL-I pins or something unexpected on DEL causes the output on COL to be absent ?

In theory you could convert your SECAM console into a PAL or NTSC one (depending on what TIA is installed on your system), but that requires replacing the crystal oscillator (and for PAL you need a second oscillator for color) and recreating the video circuitry using the PAL or NTSC schematic as reference.

Instead of recreating the composite signal, one could also probe the COL and LUM0/1/2 with a raspberry pi zero (1gHz), and have an HDMI output. It would be interesting to see if the 1x, Ex and Fx color give no COL signal, or a COL signal that is out of scope (if the later is true,  it would be possible to use the full NTSC palette on PAL and SECAM consoles)

 

Here is a small sample of the signal I find on LUM (with settings in txt file) :

NewFile1_png.txt

 

should the same setting show something that does not look like 0.5V noise on COL or PAL-I / PAL-S ?

[Windless]

Something I do not understand : SECAM uses 50Hz, one frame lasts as long as 262*228 pixels, so one pixel lasts 1/2.986.800 seconds, or 0.335µs, or the signal here appear to have a period of 2*3.4µs... Did I miss something ?

[ChildOfCv]

The internals of the TIA color circuit are a delayed-shift register which is around 14 bits long (however many hues the Atari is capable of).  PAL may be slightly different but that's basically what it is.  The input is fed a 4.43MHz clock.  If the color delay is set correctly, it will propagate halfway down the row of registers before a state change on the input.  Color output is selected by sampling at the point of the color index you want.  Except that on color burst, the clock itself is sent out.  Black is special because it leaves the color output disabled.

 

So yeah, color is just square waves that get sent through a bandpass filter before mixing with the luma DAC output.

[Windless]

3 minutes ago, ChildOfCv said:

The internals of the TIA color circuit are a delayed-shift register which is around 14 bits long (however many hues the Atari is capable of).  PAL may be slightly different but that's basically what it is.  The input is fed a 4.43MHz clock.  If the color delay is set correctly, it will propagate halfway down the row of registers before a state change on the input.  Color output is selected by sampling at the point of the color index you want.  Except that on color burst, the clock itself is sent out.  Black is special because it leaves the color output disabled.

 

So yeah, color is just square waves that get sent through a bandpass filter before mixing with the luma DAC output.

Ok, thank you for this information a!

 

aIs there a documentation I missed where this is explained ? (I found some info on the biglist archive but nothing that precise)

[ChildOfCv]

1 hour ago, Windless said:

Something I do not understand : SECAM uses 50Hz, one frame lasts as long as 262*228 pixels, so one pixel lasts 1/2.986.800 seconds, or 0.335µs, or the signal here appear to have a period of 2*3.4µs... Did I miss something ?

The 25Hz frame rate is for a total of 625 lines.  Some of those lines will be vertical blanking, and it will also be interlaced.  So that's 15625 lines per second (visible or not).  Horizontal resolution is 160, so that's roughly 2.5M pixels per second, not counting sync time.

[ChildOfCv]

7 minutes ago, Windless said:

Ok, thank you for this information a!

 

aIs there a documentation I missed where this is explained ? (I found some info on the biglist archive but nothing that precise)

I'm not sure about documentation, but I looked at the TIA schematics.  They're posted in some threads on this board.  You may also find them at the Atari Museum.

[Windless]

I play a bit with my oscilloscope that I almost never used, and got a bit better.

 

-On PAL-I seems to show a 890kHz signal :

 

 

NewFile2_png.txt

 

I found this thread :

stating that on PAL consoles, the PAL-S provides a signal to keep the crystal oscillating, the output of this crystal entering in PAL-I. Is this noisy 890kHz used to make sure a crystal oscillate ? There is a crystal but not 4.43,it is a 4.453125MHz, I did not follow the track to see if it is somehow connected to PAL-I, but PAL-I seems to have only a very noisy 1gHz signal :

 

I really don't know enough about electronic to understand how the PAL TIA is used there, but I did not find any output on COL.

 

Here are a few pictures of the crystal on the SECAM 2600 :

 

Is there any chance to make a modification so I can get some kind of ouput on the COL pin ?

 

Thank you,

Windless.

[ChildOfCv]

Looking at the 2600 PAL schematic, PAL-S is for synchronization.  Apparently they want the PAL signal to be synchronized with something, at least on occasion.  This feeds the PAL-I which should be oscillating at 4.453MHz every time PAL-S "rings the bell".

 

But yes, for PAL it should be 4.43MHz.  But until you figure out why the oscillator isn't outputting anything useful at the moment, that's a moot point.  Are you using an auto-timing mode?  You should make sure it's doing about 100ns per division in order to see whether there's anything at the level you're interested in.

 

What does the signal look like at the common node that connects R237, Y201, C241, and L204?

Edited July 18, 2019 by ChildOfCv

[Windless]

Quote

But yes, for PAL it should be 4.43MHz.

My console is SECAM with a PAL TIA, not PAL. That might be the reason.

I was writting " But should this explain the fact that there is no output on the COL pin ? " but then it occurred to me... that COL is unconnected on this board. The schematic of the PAL 2600 shows a 1K pull up... That may be all I need to read the signal on COL ^^

 

Quote

Are you using an auto-timing mode?

In the capture on my previous message, I used a trigger at +100mV and a manual 500ns and 5us per division.

 

Quote

What does the signal look like at the common node that connects R237, Y201, C241, and L204?

The 4.453125MHz crystal is labelled Y202 not 201.It took measurements on both side of the crystal. The side which is up on the last picture of my previous post (on the small smallest of the two yellow components) have a signal with the same 890kHz hat is found at PAL-S, but cleaner and with a superposed sinusoïd :

 

On the other side, the expected 4.45MHz is found. It is not very clean but this I removes the EMI shield and I know the current from the power outlet is far from clean here (bad ground + electrical water pumps around) :

(on the picture the oscilloscope guessed 6.99MHz, but most of the times it shows 4.45)

 

Still no change on pin 8 / PAL-I.

 

For reference, a picture of the points I used to measure (maybe I made a mistake ?)

 

[ChildOfCv]

Yeah, I gather that your goal is to convert to PAL?

 

Anyway, the crystal can be replaced with the correct value but that circuit seems to be failing.  What's the resistance to ground at the point that has 4.45MHz?  (Power disconnected for this measurement of course)

[Windless]

Quote

Yeah, I gather that your goal is to convert to PAL?

Converting to PAL is an option, another is to read COL and LIM with a pi zero and have an HDMI output, maybe being able to select NTSC / PAL / SECAM palette with a switch. For now I learn using the oscilloscope, and start understanding a few things about analog signals and RF, that's my main goal for now.

 

6 minutes ago, ChildOfCv said:

What's the resistance to ground at the point that has 4.45MHz?  (Power disconnected for this measurement of course)

4.680kR on the 4.45MHz side of the crystal, nearly 0 on the pin 8 of the TIA.

[Windless]

One thing I did not mention but may be of importance : I use an old 4:3 LCD monitor with a tuner. The first time I switch this console on, I hade to on/off several time so the game would start. The TV did not find the color signal everytime and often displayed B&W.

 

Now that I have open the console and removed the EMI shield (and move to another power outlet), the picture is always B&W with lot of snow unless I put the switch on B&W.

[ChildOfCv]

11 minutes ago, Windless said:

4.680kR on the 4.45MHz side of the crystal, nearly 0 on the pin 8 of the TIA.

Hmmm, well 0 is definitely not a good number.  There are a couple of possibilities:  1) The SECAM mod jumpers pin 8 to ground and this circuit goes elsewhere.  2) The TIA is dead.  3) Q203 (the transistor that should be amplifying that RC tank) is beyond dead.

[Windless]

I'll try making a schematic of what I find on this part of the board tomorrow.

Unfortunately I won't understand much of what I will see

 

Quote

2) The TIA is dead.  3) Q203 (the transistor that should be amplifying that RC tank) is beyond dead

not completely impossible I made shortcut and blow something, but I'll try on other tuners to see if I can have my color back.

[ChildOfCv]

7 minutes ago, Windless said:

I'll try making a schematic of what I find on this part of the board tomorrow.

Unfortunately I won't understand much of what I will see

That's okay.  You could use the PAL schematic as a crutch anyway.  That may give you a style guide, if nothing else, and some or all of the circuit may be the same.

[ChildOfCv]

So, I'm going to make an educated guess about how the SECAM mod works.

 

1) Pin 8 is connected to ground.

2) The 4.53MHz crystal is the oscillator source for an FM modulator for the color.

3) The "luma" output is instead used as the color and goes to a DAC resistor ladder as usual, but the output of that feeds the FM modulator's input.

4) The crystal's weird frequency is probably chosen because there's some genius circuit that can use it to hit both of the required subcarrier frequencies for the red and blue lines.

 

It should be possible to covert this to PAL, but you may need to add a daughter-board to complete the required signal paths.

 

Edited July 18, 2019 by ChildOfCv

[Windless]

Ok, the track fom pin8 of the tia (PAL-I) quickly reaches pin 2 on the 6507 (Vss).

 

I tryed to understand from the schematic of the TIA if this should give no signal on COL or a signal that would not have the expected shape/timing on the NTSC / SECAM ouput but that's a bit complicated for me.

 

I order DIP40 sockets so I can lift pins without too much deterioration of the IC.

[ChildOfCv]

Yep, Vss is ground.  So they clipped the PAL-I's wings.  The oscillator circuit is obviously doing something else, and my strong suspicion is that it is modulating the colors.  Unlike PAL and NTSC, SECAM uses FM modulation instead of QAM for color.  It alternates on each line between sending red and blue, so there will be a selector that listens to one pin or the other.  The selector will also choose the base frequency, since red and blue are sent with different carrier frequencies.

 

So there is a sizable amount of work to do if you want to make the system PAL-compatible.  First thing would be to cut the line from pin 8 to ground.  Then build the circuit as seen in the schematic to connect PAL-S to PAL-I.  At that point, you should actually see output on the color pin (that is, once a pull-up resistor is attached).

Edited July 19, 2019 by ChildOfCv

[Windless]

Do you think I can use a 22pf on the output of PAL-S to filter, read it with a blue pill (72MHz Cortex 0 board you can buy for $1.5), then generate a 4.43MHz-ish square signal to PAL-I with the Cortex ?

 

This would introduce a delay of about 3 cycles (<50ns) to detect the crossing of the 890KHz signal, then I could generate a 16 cycles signal to PAL-I (4.5 instead of 4.43MHz). This is a 1.6% variation, which needs to be added to the variation of the crystal on the cheapo cortex M0 board. Could this be used to test to see if a signal appear on COL, even if the output is not good enough to get real secam colors ?

 

Random guess : could the PAL-I signal be used only to modulate the output of COL signal, hence I could just tie PAL-I to +5V and get a square signal instead of 4.43MHz signal on COL ? I need to find some documentation about how subcarriers work

Edited July 19, 2019 by Windless
typos

[ChildOfCv]

TBH, using a Cortex chip to generate a square wave seems like gross overkill, considering that the board would be able to emulate the entire Atari by itself.  The circuit on the schematic is simple enough, can be built on a breadboard, and should give you the required input.

 

If you just want to see what the signal might look like if it had a clock, you could jumper PAL-I to the 3.58MHz clock on board.  It won't be synchronized by PAL-S, but it should give you something to oooh and aaaah at on the COL output.  Of course to see different colors in play, you'd have your scope take output from the COL output while syncing on the PAL-S input.