Prototype set is here

[tubesrule]

Originally Posted by Penthode

Note that CPA was NTSC!

I guess I only consider CBS and today's analog color standard as NTSC since those were the ones ratified. CPA, dot sequential and the rest were only lab exercises or presentations to the NTSC.

I am envisioning an issue with CPA and compatibility with the SCRF converters. Since these converters don't have frame memories they can't do frame rate conversion. This means the line and frame timing will match modern standards so there will not be a exact relationship between the line timing and sub-carrier frequency. We would need to either shift the sub-carrier frequency slightly, or accept a non-integer relationship. Not sure if the later would be a problem as I have not seen how the set determines which field phase it's in. The more expensive WC converter can do whatever we need but is overkill.

Darryl

[Penthode]

Originally Posted by tubesrule

I guess I only consider CBS and today's analog color standard as NTSC since those were the ones ratified. CPA, dot sequential and the rest were only lab exercises or presentations to the NTSC.

I am envisioning an issue with CPA and compatibility with the SCRF converters. Since these converters don't have frame memories they can't do frame rate conversion. This means the line and frame timing will match modern standards so there will not be a exact relationship between the line timing and sub-carrier frequency. We would need to either shift the sub-carrier frequency slightly, or accept a non-integer relationship. Not sure if the later would be a problem as I have not seen how the set determines which field phase it's in. The more expensive WC converter can do whatever we need but is overkill.

Darryl

Darryl,

The DOT sequential system is RCA. But the CPA test came under the auspices of the NTSC. I believe Bernard Loughlin refers to CPA in the context of NTSC.

Can you explain your concern over compatibility with the SCRF?

It would seem to me that the current NTSC frame rates are acceptible for CPA since the NTSC frame rates ensure proper chroma interleaving with the presence of the 4.5MHz sound carrier.

Hence if the current scan rates are used, the standard NTSC subcarrier relationship (odd integer multiple of half the horizontal rate would suffice for the nominal 3.89MHz. But CPA means that with only a half line offset, the Y and R-Y spectra will align to form a vertical dot line up and resulting dot pattern. Hence I believe a quarter line offset is required.

The nominal 3.89MHz results in:

3890000/fH/2 = 494.4622...

The closest quarter line offset multiple is therefore 493.75

Therefore the CPA fsc = 493.75 x fH/2 = 3.884396MHz

I believe this CPA subcarrier can be used with current NTSC scan rates and provide the best interleaving of the chroma.

What do you think?

Terry

[tubesrule]

Terry,

The CPA chroma frequency was defined as exactly 495/2 * Fh. Using the original Fh of 15,750.00Hz yields exactly 3,898,125Hz. This also means that there was an exact relationship between the phase of the chroma carrier and the line frequency (i.e. the phase and position of the chroma carrier will be exactly the same on every other line). From what I have read there was no half line offset. That's something that came later. Of course anything can be done now, but I would recommend putting the set back to as close to original as possible, warts and all, to verify what they saw during the original tests. This will result in a stationary, visible pattern but that's what they would have seen.

So using the current NTSC line frequency of ~15,734.27Hz and the same 495/2 yields ~3,894,230.77Hz. This means that to maintain the same locked phase relationship between the chroma carrier and the line, we would have to use this new frequency. If we wanted to use the original 3,898,125Hz, we would loose the locked phase relationship between the chroma carrier and the line. This will change the visible dot pattern.

The issue with the SCRF converter is there is no frame memory, so rate conversion cannot be performed. The line output timing would match the line input timing, or 15,734.27Hz. With the WC converter the output is complete independent of the input and anything can be done.

So is there a design reason to have the chroma locked to the line, or does it even matter with this set? Do we want to observe the set exactly as it would have appeared during the tests? Since a crystal will need to be cut it could easily be cut for either frequency, or both could be cut and tried. To the converter it's just firmware so it doesn't matter. This will ultimately be Nik's/Steve's call assuming they take the set all the way to operation on CPA.

Darryl

[Penthode]

Originally Posted by tubesrule

Terry,

The CPA chroma frequency was defined as exactly 495/2 * Fh. Using the original Fh of 15,750.00Hz yields exactly 3,898,125Hz. This also means that there was an exact relationship between the phase of the chroma carrier and the line frequency (i.e. the phase and position of the chroma carrier will be exactly the same on every other line). From what I have read there was no half line offset. That's something that came later. Of course anything can be done now, but I would recommend putting the set back to as close to original as possible, warts and all, to verify what they saw during the original tests. This will result in a stationary, visible pattern but that's what they would have seen.

So using the current NTSC line frequency of ~15,734.27Hz and the same 495/2 yields ~3,894,230.77Hz. This means that to maintain the same locked phase relationship between the chroma carrier and the line, we would have to use this new frequency. If we wanted to use the original 3,898,125Hz, we would loose the locked phase relationship between the chroma carrier and the line. This will change the visible dot pattern.

The issue with the SCRF converter is there is no frame memory, so rate conversion cannot be performed. The line output timing would match the line input timing, or 15,734.27Hz. With the WC converter the output is complete independent of the input and anything can be done.

So is there a design reason to have the chroma locked to the line, or does it even matter with this set? Do we want to observe the set exactly as it would have appeared during the tests? Since a crystal will need to be cut it could easily be cut for either frequency, or both could be cut and tried. To the converter it's just firmware so it doesn't matter. This will ultimately be Nik's/Steve's call assuming they take the set all the way to operation on CPA.

Darryl

Hi Darryl,

The shift from the horizontal rate of 15,750Hz to 15750 x 1000/1001Hz was only to accommodate the 4.5MHz sound carrier. Hence shifting the subcarrier to accommodate the slightly lower scan rates should be inconsequential. I trust you agree the frame rate conversion is not necessary.

The half line offset I referred to as an odd number multiple of half the line rate and is what is used for the current NTSC system.

PAL uses a quarter line offset so that there is a quarter cycle less of subcarrier per line. This means that when the R-Y axis reverses line by line, there will not be the chance of vertical dot line up.

I have not found any information that CPA used a quarter line offset but I included in my calculation anyway.

My thoughts are this: a fixed relationship between CPA subcarrier and the scan rates is desirable but not essential. The frequency response of VHS tape is restricted so that there is no line locked relationship to fsc in VHS tape and it looks pretty good. If however wider luma bandwidth is desired (through comb filtering the source video) then maintaining a locked frequency relationship.

The quarter line offset (as used with PAL) I believe applies to CPA. But is this correct? (It requires some thought) And would the visual field averaging require the quarter line offset? Perhaps it was not used or was not necessary?

Perhaps the NTSC in 1952 was preoccupied with the edge flashing artifacts and so did not get around to thinking about CPA interleaving in greater detail? If that is the case, I would proceed with CPA fsc = 495 x 15750/2 x 1000/1001.

[tubesrule]

Hi Terry,

Originally Posted by Penthode

Hi Darryl,

The shift from the horizontal rate of 15,750Hz to 15750 x 1000/1001Hz was only to accommodate the 4.5MHz sound carrier. Hence shifting the subcarrier to accommodate the slightly lower scan rates should be inconsequential. I trust you agree the frame rate conversion is not necessary.

I believe shifting the CPA subcarrier to match the current line rate is the right thing to do. While it won't use the original subcarrier frequency, it will use the original subcarrier to line rate relationship and will provide for the original viewing experience.

Originally Posted by Penthode

The half line offset I referred to as an odd number multiple of half the line rate and is what is used for the current NTSC system.

PAL uses a quarter line offset so that there is a quarter cycle less of subcarrier per line. This means that when the R-Y axis reverses line by line, there will not be the chance of vertical dot line up.

I have not found any information that CPA used a quarter line offset but I included in my calculation anyway.

I had included in my prior response a section on how current NTSC and PAL formats are constructed but later removed it. This gets back to my previous comment about what is to be accomplished. (see below)

Originally Posted by Penthode

My thoughts are this: a fixed relationship between CPA subcarrier and the scan rates is desirable but not essential. The frequency response of VHS tape is restricted so that there is no line locked relationship to fsc in VHS tape and it looks pretty good. If however wider luma bandwidth is desired (through comb filtering the source video) then maintaining a locked frequency relationship.

The quarter line offset (as used with PAL) I believe applies to CPA. But is this correct? (It requires some thought) And would the visual field averaging require the quarter line offset? Perhaps it was not used or was not necessary?

Perhaps the NTSC in 1952 was preoccupied with the edge flashing artifacts and so did not get around to thinking about CPA interleaving in greater detail? If that is the case, I would proceed with CPA fsc = 495 x 15750/2 x 1000/1001.

I think your last comment is pretty close to how I feel. While we can certainly use some of the "tricks" applied to current NTSC and PAL to make the image look better, and while this would be a worthwhile exercise to see what could have been, my feelings are the initial design goal would be to recreate CPA exactly as it would have been during the trials.

From my point of view none of the specifics matter other than what the final target is. Since it's just VHDL coding and requires no physical hardware modifications to the converter, we can play with the specifics ad infinitum. (actually the FLASH memory is only rated for 100K write cycles )

Some impact would be imparted on the set as specific crystals would need to be cut for different trials. This is not expensive or difficult, but needs to be considered. Since the crystal is socketed it could be quickly replaced if you want original performance or "enhanced" performance.

If you could work through the specifics of what the best enhanced format would be, I can get it coded and then we would have both available for testing.

Darryl