32-line song clip .wav file from 1930-vintage film

[chris_vk3aml]

Deep in the archives of early talkies - the delightful Nancy Carroll and Charles "Buddy" Rogers singing "We'd Make A Peach Of A Pair" in 1930. At last, a music video to play on your 32 line disc scanners and "Big Picture" software!

This stretches 32 line definition about as far as it will go - a duet with TWO people on camera means that they have to huddle close to be properly visible. Fortunately, in this scene, they did - and who wouldn't have been tempted to cuddle Nancy, one of the cutest gals of the early sound film era!

The cache limitations of postings on this website means that I'm limited to 2 mins 57.5 seconds of this at 44.1 KHz sampling and 16 bit. Please feel free to comment as I'm fiddling with the nbtv file with AdobeAudition3 after converting it from an .flv net file - boosting highs and lowering gamma to make it reproduce well on LED-disc receivers.

I'm working towards producing a 32-line program to exploit the medium's entertainment potential.

All the best,

Chris Long VK3AML.

[Klaas Robers]

Chris,
I downloaded both pieces of 32 line video wave-files and looked at the wave form. Then it struck me that the video is so large in amplitude. I attach 3 images:
1. colorbar.gif, this is the video signal of the colourbar in CCNBTV. It is a kind of reference wave form between Vic and me. You will see that the sync is 0,9 div negative, black is zero, also in the missing sync pulse and white is 2 div positive. This complies more or less to 1 volt video and 0,4 volt sync. (It is a little bit more)
2. shadow.gif, this is the video of your shadows file, and
3. clip.gif, the songclip in this thread
You will see that in both wav-files 2 and 3 the sync is 0,9 div negative, but the video goes almost to 4 div.

Why have you done that?
And how have you done that?

[chris_vk3aml]

Well, Klaas, I was not aware that there was any "standard" level required for these CD files. As sound cards rarely produce a standardised output voltage (at least my two sound cards do not - and my three stand-alone CD players definitely do not) it would seem only logical to use a level close to the maximum amplitude of which the digital recording system is capable, without clipping, to take advantage of the maximum possible bit depth.

The clip was modified after conversion from .flv file with AdobeAudition v3. By using the tabs under "effects" you can select the tab "special" and then the sub-tab "distort". This permits independent control of the amplitude and transfer characteristic of half-cycles of the NBTV waveform. One can control the level of the video (on the positive side of the datum) independently of the syncs (which are on the negative side of the datum). One can also place control points along the transfer characteristic curve, then spline them to change the gamma in any way you desire. By altering the zero point above the x-axis on this protocol, you can also set black level on an input file lacking really deep blacks - as many rough transfers from old films do.

This video level controlling facility on AdobeAudition v3 is impressive enough - but the parametric equaliser filter is even more useful, as it lets you set an h.f. boost curve to correct for the finite size of the scanning aperture. I discussed this on the postings that I made regarding "aperture correction". This is also hf boost WITHOUT phase shift - exactly as aperture correction theory demands. It is possible to set this up in a way that has little effect on syncs, but has a significant sharpening effect on the 32-line video.

The result of all this was applied to the 1930 "song clip" that I posted yesterday.

By comparison, I'll post a few sample seconds from the same clip as directly converted from the original .flv file - no blacks, gamma far from optimum, and almost too fuzzy to watch. I think you'll agree that the changes made were highly desirable.

All the best,

Chris Long VK3AML.

[chris_vk3aml]

...and for reference, here is a screenshot of the settings on Gary Millard's "BigPicture" program on which my "raised level" 32-line videos seem to give the best results - at least on my LCD screen...

Best wishes, feedback always appreciated, and thanks for that, Klaas,

Chris Long VK3AML.

[Steve Anderson]

Well, Klaas, I was not aware that there was any "standard" level required for these CD files. As sound cards rarely produce a standardised output voltage (at least my two sound cards do not - and my three stand-alone CD players definitely do not) it would seem only logical to use a level close to the maximum amplitude of which the digital recording system is capable, without clipping, to take advantage of the maximum possible bit depth. Chris Long VK3AML.

Absolutely correct Chris. There is no standard level either from CD players or the output of PC sound-cards.

In addition there is no standard POLARITY.

I did a survey of about eight CD players about a year ago, they were all different, even from the same same company, S**y take note! Both in signal amplitude and polarity.

In the .wav files I create I use all of the 8-bits but don't have the headache of syncs as I don't use them.

Steve A.

[Steve Anderson]

Sometime in the last day or two someone mentioned that 8-bit .wav files kill the syncs in some manner. A while ago I when I was working in 32-lines with syncs I created a number of 8-bit wave files.

Here's a screenshot of how my PC handles the syncs which seems OK to me. Can't understand why they should get corrupted, it is after all only audio.

You'll also notice that the PC puts out a signal of 3.5V p/p.

Steve A.

P.S. The 8-Bit sync/wav problem was in a PM, not on the forum proper. No wonder I couldn't find it!

[gary]

The only possible disadvantage I can see with using 8 bits with sync is being limited to 128 greyscales - not usually much of a problem in video...?

[Steve Anderson]

The only possible disadvantage I can see with using 8 bits with sync is being limited to 128 greyscales - not usually much of a problem in video...?

It's actually around 190 video levels as the syncs only take up 30% of the available 256 values. 5-Bit video or stills (only 32 levels) can look amazingly good, there's a discussion in another thread regarding this.

I did think of breaking up the bytes into 4-bit nybbles, but at 16 video levels contouring becomes excessive. It's just becoming visible in the picture below, but it does depend on your monitor, it looks worse on the laptops LCD than the CRT monitor.

As for the syncs themselves, I got rid of them partly for the reason above, but mostly 'cos they're simply not needed.

Steve A.

[gary]

Yes if you allow negative going black levels the loss is only the depth of the sync, once it is AC coupling I suppose it makes no difference...

[AncientBrit]

Chris,

Cracking material, well done.

Regards,

Graham

[Klaas Robers]

Yes indeed there is no standard for analogue output levels of CD players and there is no kept standard for polarity. However it would be nice if the output level of the video signal is more or less unified from CD to CD. It is annoying when each CD has a completely different level and you have to readjust gain (contrast) and DC-level (brightnes) when you start another CD in your player. That is why I made the comment.

On the other hand, there IS a recommended output level and polarity for line outputs of CD players. At high end audiosystems you will hear the difference if you invert the polarity, so those guys wanted to have this fixed. And from the beginning of CD I experienced this also that way. Only during time manufacturers could save on one transistor by leaving the polarity inversed. For professional equipment this is better of course.

If you want to have a guideline, look into the standard of CCNBTV. There are decimal numbers given for top sync, black and white. This then still gives slightly higher video signals than the clubdiscs 1, 2 and 3. The amplitude has been changed for colour to take the maximum advantage of the 7 bits available.

Of course you know that there is no reason to maximise the video signal in a 16 bits surrounding. The quantisation noise is so extremely low that you don't need it. This becomes more important if you limit this to e.g. 7 bits, as was done in CCNBTV.

[Steve Anderson]

....However it would be nice if the output level of the video signal is more or less unified from CD to CD. It is annoying when each CD has a completely different level and you have to readjust gain (contrast) and DC-level (brightnes) when you start another CD in your player.

That is a very valid point Klaas, but there is an inbuilt reference level in every digital representation of our analogue world, it's called full-scale. If everyone recorded the video at full-scale amplitude then there would be no fiddling with knobs required once set. We're recording an anlogue signal onto a digital format, how do we know what the 'digital level' is? We don't really unless you examine the CD at hex level.

Digital AES audio has a reference that is set some way below full-scale and as ever there are four commonly used references, -16, -18, -20 and -24dbFS. There are others.

Digital video (CCIR601 for example) is fixed known quantity, sync-tips to peak white always produces 1V p/p in a 75 Ohm analogue circuit, the converse when recording. There is the issue of line and frame rate, but that's something else. Thankfully this has been carried over into domestic equipment, but in audio, no. Although in professional audio there is a fixed world-wide standard for analogue signals, but not in digital.

They had a chance to get it right and blew it.

The only way I see around this for NBTV is a standard-level disc, a test disc that has (say) a 50 Hz sine-wave on it at -12db. You set up your system such that you get 1.414V RMS out on playback and then you record a I.414V 50Hz signal and adjust the record gain such when you play your recording back it's also 1.414V RMS. Unity gain from a known reference which in an ideal world everyone would be using.

I chose 50 Hz 1.414V as not everyone has a scope and the response of most DVMs is inaccurate at higher frequencies, and a sine-wave as they also will respond differently to a complex waveform such as NBTV.

This suggestion doesn't use the full 16 bits, but with an audio CD it's more than good enough. With 8-bit .wav files I used the whole range available.

Then there is the issue of polarity....

Steve A.

[Klaas Robers]

The full scale digitally spoken is a good point, but I happen to know that in digital video there is always some headroom and some feetroom taken, if I am right white is F0, so eventual overshoots because of crispening are not clipped off.

In the NBTV CDs I tried to make the video signal not too loud, more or less comparable with the sound. That is why not the total range from 8000 to 7FFF is used. For CCNBTV Vic and me wanted to make the decoding in hardware not too difficult, and to be as close as possible to the existing club CDs, so the limits were choosen as they are.

In that case FOR CD I think it wise that the video signals when coming from different discs should match. And I was hoping more or less that Chris was preparing NBTV clips for making a CD. But I might be completely wrong and this is only ment to play back directly from a computer via the sound blaster. In that case no unification is needed at all of course.

And for the polarity: It looks very clumsy that two CDs were recorded in oposite polarity. Once you found the correct polarity for your player and monitor it should stay that way.

[chris_vk3aml]

You've just explained to me, Klaas, the reason that I had to rip the NBTVA club's CD's to hard disk and then normalise every file before playing them on Gary Millard's ORIGINAL NBTV software player. On my original computer, at least, the club CDs would not play with a full range of grey scale before the tracks were greatly increased in audio level.

I take your point on standardising the level on the CDs, but as there seems to be no standardisation of the input and output levels of sound cards, I favor the simple expedient of normalising (ie. maximising) the level of the CD - or keeping it within a dB or so of maximum.

Between 1996 and 2003 I ran an audio CD business issuing vintage recordings from the 1920s and earlier. Our policy was to normalise most tracks after recording, only holding back the overall peak level a few dB if a few tracks SOUNDED significantly different in level to the others. For example, a pre-electric recording with telephonic bandwidth would sound much lower in level than a full-bandwidth electrical recording, although the peak levels would be the same. Sometimes we had to pull the levels of the electrical recordings back 3 or 4 dB if they were issued on the same CD with pre-electric items. Normalising levels to a maximum that the system would handle was generally the last operation you would perform before "locking off" the CD mastering.

You would be surprised at the extreme variation of levels across commercial audio CD releases, and in some cases the commercial issues never were normalised, resulting in huge level variation from track to track. That can be really annoying if you're running an unattended CD in the background for guests at a dinner, for example.

If there is "a recommended output level and polarity for the line outputs of CD players" as you say, Klaas, I have yet to find that to be the case with ANY CD player that I have ever owned - and that would probably amount to a dozen samples so far.

And yes, Steve, when you go to 8-bit the greyscale remains fine, the audio is almost acceptable, but for some reason the sync does jitter on all of the software viewing packages that I've tried. I don't know why - perhaps the quantisation level interpolations become flaky on the high-slew leading and trailing edges of the sync pulses when their bit depth is reduced.

I think this whole argument of standardisation is getting a bit to the level of hair-splitting. My NBTV tapes have EVERY IMAGINABLE line rate, and between 1970 and the present I have used:

8 lines 1:1 ratio
16 lines 1:1 ratio
24 lines 1:1 ratio
48 lines 1:1 ratio
30 lines 3:7 ratio
64 lines 3:4 ratio

...all with field rates ranging from 1 f/s to 25 f/s

...plus I never have managed to find really powerful synchronous motors capable of reliably driving camera discs, so they have always either been DC motors, AC series motors or shaded pole motors with slip below sync speed. If the viewing system can't take account of these variations, I don't want to know about it, really.

Perhaps the viewing software should trend to being more adaptable, rather than applying standards where the levels, frequencies and sync systems become more rigid. This is, after all, EXPERIMENTAL television - and it always was.

All the best,

Chris Long VK3AML.

[Steve Anderson]

This is, after all, EXPERIMENTAL television - and it always was. Chris Long VK3AML.

A very good point. One I totally agree with. There of course is validity in having an agreed standard where we wish to connect your camera to my display where commonality is obviously required. But otherwise where's the beef?

The chances of me ever bringing some hardware to the convention amount to zero. So I'm going to stay with my 'excessive' signal levels, peculiar line rates, different aspect ratios and scanning directions.

If there is anyone else in this country who has this interest, I would be very surprised! So the chances of me ever hooking up to someone else's gear is most unlikely.

Steve A.