Prototype set is here

[Penthode]

Originally Posted by tubesrule

Hi Terry,
Certainly obtaining the video directly from a digital source is preferable but not practical for this level of converters. In my professional design work, everything is SDI, quickly becoming 3G SDI but that is unheard of outside the broadcast industry. Consumer level HDMI is hampered by HDCP making it useless unless you want to try to obtain a license which would be cost prohibitive for these low volume converters.

For my low cost converters I use a video front end with a 4 line adaptive comb filter, so while not quite as good as component, it has proven to be more than adequate for the capabilities of these early sets. These parts have the capability to support S-Vid, so this input can be added to these converters. My broadcast WC converter has CVBS, S, and component YUV/RGB inputs but is overkill for this purpose. It uses 10 bit ADC/DAC's with 14 bit minimum internal processing which considerably raises the cost.



I don't think it would be quite this easy in analog using baseband component. You need to generate the 3.8MHz carrier, preferably correctly locked to the Hsync (assuming the Hsync from the source is even stable enough although you could free run like consumer vcr's), bandwidth limit and quadrature modulate the R-Y and B-Y and then add back into Y. Potentially an off the shelf part might be able to do at least a portion of this, but it's still a fairly involved design.

The nice thing with an fpga design is you are not limited to the physical hardware. If you want to add a filter, change the carrier frequency or phase, etc. it's just a respin of the vhdl. I have added stuff to clients designs that were not even in the original spec all without touching the hardware.

Darryl

Hi Darryl,

I certainly see your points. But the issue at the time was for CPA to optimize luma and chroma resolution. For one to first hand understand whether the abandonment CPA was justified requires a signal with at least the chroma resolution the early NTSC was trying to achieve. CPA was initially intended to convery full bandwidth chroma for both color differ signals of 1.5 MHz. NTSC constricted the Q channel to 500kHz. Secondly the use of a four line adaptive decoder compromises the diagonal luma resolution as well as reducing the vertical chroma resolution. In other words concatenating the two codecs will not give a proper representation of CPA. However, this will only be worth the bother if the display CRT focus and bandwidth to the CRT are up to the job of showing the extra resolution that was intended by the use of CPA.

I also believe the thought of the NTSC at the time was if putting the subcarrier at 3.89MHz, would the subcarrier into luma interference pattern less obvious and allow the luma bandwidth to be somewhat greater? The IQ NTSC standard with the subcarrier at 3.58MHz meant that the luma trap limited luma bandwidth to only about 3.0MHz and with the subcarrier at 3.89MHz, the luma response was effectively extended another 400kHz.

Component video from a DVD player will have reasonable stable sync. I agree that locking the subcarrier to the sync will be a challenge. VHS tape as a source will be totally out of the question since VHS does not fulfil the mathematical subcarrier to Hsync relationship and VHS simply does not have the bandwidth. DVD playback does and is stable.

I would think the reason I would persue CPA encoder without the IQ concatenation would be to see how well CPA would work. I think we would all agree that component 525 video looks very decent. (In fact playback of a good quality component DVD authored from 35mm film looks better than much (if not most) broadcast video purported to be HD).

I agree that a dozen transistors will not account for even the sync divider. But I think a few ASICs may be adapted to work in a CPA encoder. I trust the restoration of this set (if it is indeed a CPA Prototype) will come to fruition and the opportunity presents itself to compare CPA encoding schemes. Besides I would like to see how well an FPGA solution will work.

All the best,

Terry

[miniman82]

Originally Posted by Penthode

For one to first hand understand whether the abandonment CPA was justified requires a signal with at least the chroma resolution the early NTSC was trying to achieve.

To this end, I don't honestly need to see it to understand what the engineers saw back in 1953. I have a few papers that detail some of the issues they themselves saw when using CPA, and their reasons for abandoning it's use. The flicker was seen as objectionable, though their notes indicated that it was not too bad at low brightness levels. Additionally, as was found by the later tests, I/Q sets produced pictures that were acceptable for the general public and I doubt anyone pitched a fit or even noticed a difference when sets went to narrowband shortly thereafter. Let's remember that the average person back then (and I'd argue even now) wasn't so concerned about picking apart every little detail as some of us are. I'd assume that they were enthralled by simply having a color picture on the screen, much less being able to afford one as expensive as they were.

I can tell the difference in picture quality between a wide bandwidth set like my 21CT55 and a narrow bandwidth CTC-4, but most people would be hard pressed to see the difference. All that said, it would be important to feed a CPA set with signals having the correct bandwidth in order to evaluate its performance. But since this set doesn't appear to have a CPA switch in it, I don't think we will ever know.

[tubesrule]

Hi Terry,

Originally Posted by Penthode

I certainly see your points. But the issue at the time was for CPA to optimize luma and chroma resolution. For one to first hand understand whether the abandonment CPA was justified requires a signal with at least the chroma resolution the early NTSC was trying to achieve. CPA was initially intended to convery full bandwidth chroma for both color differ signals of 1.5 MHz. NTSC constricted the Q channel to 500kHz. Secondly the use of a four line adaptive decoder compromises the diagonal luma resolution as well as reducing the vertical chroma resolution. In other words concatenating the two codecs will not give a proper representation of CPA. However, this will only be worth the bother if the display CRT focus and bandwidth to the CRT are up to the job of showing the extra resolution that was intended by the use of CPA.

I also believe the thought of the NTSC at the time was if putting the subcarrier at 3.89MHz, would the subcarrier into luma interference pattern less obvious and allow the luma bandwidth to be somewhat greater? The IQ NTSC standard with the subcarrier at 3.58MHz meant that the luma trap limited luma bandwidth to only about 3.0MHz and with the subcarrier at 3.89MHz, the luma response was effectively extended another 400kHz

All valid points. To see CPA as it was originally envisioned would take higher chroma bandwidth than available with I/Q NTSC. This has also been brought up with the converters when displaying to a CBS sequential set which is one of the reasons I added the component inputs to the WC converter. As Nick pointed out however, it takes a trained eye to see any difference on the few CBS sets when using a composite versus component input.

Keep in mind that we are not using a strict NTSC signal as a video source. When coming from a set top box, DVD, DVR, basically any digitally sourced device, the video encoder is probably not doing Q limiting per the spec. This along with the fact that the source is clean and we are connecting directly to the set, many other limiting factors have been eliminated or greatly reduced.

Originally Posted by Penthode

Component video from a DVD player will have reasonable stable sync. I agree that locking the subcarrier to the sync will be a challenge. VHS tape as a source will be totally out of the question since VHS does not fulfil the mathematical subcarrier to Hsync relationship and VHS simply does not have the bandwidth. DVD playback does and is stable.

This is why I prefer digital solutions. The WC converter has a built in 8 frame memory that acts a TBC, so regardless of the source, the output will always be stable and correct. While that is not as important in this application, it makes a world of difference for the CBS sets. Once you get those wheels up to speed and in sync, you don't want the sync phase changing

Originally Posted by Penthode

I would think the reason I would persue CPA encoder without the IQ concatenation would be to see how well CPA would work. I think we would all agree that component 525 video looks very decent. (In fact playback of a good quality component DVD authored from 35mm film looks better than much (if not most) broadcast video purported to be HD).

Whenever a new format comes up, I always add it to all my converters, so since it will also be in the WC converter, we would be able to compare the effect of the different inputs.

For an input source on any of these comparisons we really should be using 10bit uncompressed video, or at least a minimally compressed one like DigiBeta. While DVD's can look good, as a design engineer the 8 bit quantization and mpeg2 compression are horrendous as evidenced in fast motion and quantization noise in dark scenes.

Great discussion Terry. I guess we need to see what Nick discovers as he documents the set.

Nick, it's disappointing to hear the set does not have the CPA switch and may be too far modified from it's origins to put it back.

Darryl

[miniman82]

Originally Posted by tubesrule

Nick, it's disappointing to hear the set does not have the CPA switch and may be too far modified from it's origins to put it back.

Regardless, it's a very interesting chassis to look at even if it never works. It's a great historical piece, since it's obvious they were trying different things with it. It has chassis stampings labeled 'CPA' and 'DELAY', and there are a pair of 12AU7's not connected to anything that look as if they may have had something to do with CPA at some point but now serve no purpose. Looking at the local oscillator, it has a pair of phase shifting networks on it's output which then go directly to the R-Y and B-Y demodulators. It resembles an NTSC circuit more than anything else, so that's what I'm putting my money on. Still not clear if it's I/Q, but since it was what they eventually decided on it's highly probable. Nailing down the years it was worked on would help, but there are no identifying marks on it.

[Penthode]

Yes, it is disappointing that the CPA switch cannot be found. Despite the sketchy information that is to be currently found on early NTSC development, the CPA decoder could possibly be reconstructed.

I also appreciate the insight to the FPGA design converter. Thanks, Darryl.

This set nevertheless is a very unique part of television history and I am glad it has a good home.

Terry