Experiment at higher frame rates

Forum for discussion of narrow-bandwidth mechanical television

Moderators: Steve Anderson, Dave Moll, Andrew Davie

Experiment at higher frame rates

Postby Andrew Davie » Sat May 12, 2007 4:23 pm

I converted a sample NBTV signal into 1.5x and 2x speed, using a free software program 'Audacity'. Much to my surprise, both of these signals play OK on my mechanical TV -- provided I allow the motor to run fast enough.

In both cases, if I had to make a call, I'd say that the picture was SHARPER than at 12.5fps. This is probably a result of the increased frame rate and less flicker that results. In any case, I can confirm that this hardware configuration (NBTV circuits, motor control) are quite happy at higher frame rates.

Apart from adhering to the standard, I really can't see any problem at all going with a higher frame rate (we still have the 20kHz limit applying, the only difference is 'apparent' vertical resolution). Basically, the picture looks heaps better at 18.75Hz.
Last edited by Andrew Davie on Sat May 12, 2007 7:15 pm, edited 2 times in total.
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1515
Joined: Wed Jan 24, 2007 4:42 pm
Location: Hobart, Tasmania, Australia

Postby Andrew Davie » Sat May 12, 2007 7:14 pm

I have further experimented, and note that my setup will happily display and synchronise to signals up to and including double-speed (that's as far as I went). I noted no image resolution degradation (I compared the smallest pixels in image 23 on CD#1). In fact, as I noted earlier, the images look sharper/clearer, not worse!
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1515
Joined: Wed Jan 24, 2007 4:42 pm
Location: Hobart, Tasmania, Australia

Postby DrZarkov » Sat May 12, 2007 7:32 pm

That is very interesting! I wonder where the technical limits are of CD-players, or maybe we can use PCs with a good soundcard, a little bit more than 20 kHz should be possible, maybe it is possible to add the left and right audio channel to get more bandwidth, if necessary. If you have such good results, I think 60 lines/25 frames could be possible? (I would then switch to landscape format.)

Don't forget that 20 kHz is the maximum of a common CD, but in reality we never transmit a checkboard with every pixel in black and white contrast, so I think 60 lines could be possible.

Of course such a monitor would be more difficult to make, but it is interesting to find the limits.
User avatar
DrZarkov
I think I've had a cranial implosion.
 
Posts: 1037
Joined: Mon Feb 19, 2007 11:28 pm
Location: Kamp-Lintfort, Germany

Postby Stephen » Wed May 16, 2007 2:58 am

DrZarkov wrote:That is very interesting! I wonder where the technical limits are of CD-players, or maybe we can use PCs with a good soundcard, a little bit more than 20 kHz should be possible, maybe it is possible to add the left and right audio channel to get more bandwidth, if necessary. If you have such good results, I think 60 lines/25 frames could be possible? (I would then switch to landscape format.)

Don't forget that 20 kHz is the maximum of a common CD, but in reality we never transmit a checkboard with every pixel in black and white contrast, so I think 60 lines could be possible.

Of course such a monitor would be more difficult to make, but it is interesting to find the limits.
Actually, 60 lines at a frame rate of 37.5 fps is very possible with no increase in bandwidth. Take a look at John Logie Baird's patent for increasing line count and frame rate at http://www.taswegian.com/NBTV/forum/viewtopic.php?t=163 . The specific example shown in Figures 1 through 4 is for a 24 line system, but all that you have to do for a 60 line system is substitute 20 square apertures for the 8 square ones shown and 10 rectangular ones for the 4 rectangular ones shown.

Mr. Baird actually describes such a 60 line, 37.5 fps arrangement on page 1, column 2, lines 1 through 84. The resulting image would be 60 pixels wide by 70 pixels high, therefore almost square. The bandwidth would be exactly the same as the standard Baird 30 line format, approximately 13 kHz.

The only difficulty with this embodiment is fabricating the discs. However, it would be worth the effort.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby DrZarkov » Wed May 16, 2007 4:46 am

Whatever we think about, Mr. Baird was always one step ahead! :D
Indeed a monitor with 60 lines would be very interesting, but i think 25 frames interlaced would be better, a converter for "normal" TV should be very easy, if we switch to horizontal scanning and an aspect ratio of 3:4. That would be 60x80 pixels.

The scanning disc would be very huge...
User avatar
DrZarkov
I think I've had a cranial implosion.
 
Posts: 1037
Joined: Mon Feb 19, 2007 11:28 pm
Location: Kamp-Lintfort, Germany

Correction!

Postby Stephen » Wed May 16, 2007 5:06 am

Stephen wrote:Actually, 60 lines at a frame rate of 37.5 fps is very possible with no increase in bandwidth.
Sorry, after I thought it through, I realised that this is not true. The particular combination of 60 lines at 37.5 fps to give a 60 by 70 pixel image with Mr. Baird's embodiment would require a bandwidth of about 39 kHz. However, if each line represents 40 instead of 70 pixels, then the bandwidth of such a 60 line 37.5 fps system with a 60 by 40 pixel image would fall to about 22 kHz.

One could modify Mr. Baird's embodiment for 25 fps simply by using two spirals instead of three. For instance, if each spiral has 32 square apertures and 8 rectangular ones, and each line represents 40 pixels, then the bandwidth for a 64 by 40 image at 25 fps would be 20 kHz. This would be little different than the 37.5 fps system because of the "multiplying" effect of the extra spirals.

In any case, both of these examples would provide a 4:3 "landscape" aspect ratio with vertical scanning. Of course as you mentioned, Volker, the discs would have to be rather large to have two or three spirals of apertures.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby DrZarkov » Wed May 16, 2007 5:54 am

My soundcard does 44 kHz, I think a video driver for something between 20 and 40 kHz should not be more difficult than those we are already using. I think mechanical skills are more important than electronic skills for "HD-NBTV".
User avatar
DrZarkov
I think I've had a cranial implosion.
 
Posts: 1037
Joined: Mon Feb 19, 2007 11:28 pm
Location: Kamp-Lintfort, Germany

Postby Stephen » Wed May 16, 2007 7:18 am

DrZarkov wrote:My soundcard does 44 kHz, I think a video driver for something between 20 and 40 kHz should not be more difficult than those we are already using. I think mechanical skills are more important than electronic skills for "HD-NBTV".
I always tend to think in terms of vertical scanning, but let us consider lateral scanning instead. If one modifies the Baird system as shown in the patent for lateral scanning with a 4:3 aspect ratio and 64 lateral lines at 25 fps, that would provide a laterally scanned 80 by 64 pixel "landscape" image. The bandwidth would be 40 kHz.

The disc would have two spirals, each with 40 apertures, 32 of which would be square and 8 of which would be rectangular. This means an 80 aperture disc. Of course, to get a sixty four line image using ordinary progressive scanning would require a 64 aperture disc, so this system with two spirals only requires an additional 16 apertures. The difference is that using ordinary progressive scanning an 80 by 64 pixel image at 25 fps would require 64 kHz, so the Baird system is a big improvement even with only the two spirals.

Another lateral scanning possibility is 60 lines at 37.5 fps with three sets of spirals, each spiral having 20 square apertures and 10 rectangular ones, a total of 90 apertures. This would produce a laterally scanned 80 by 60 pixel landscape image at 37.5 fps. The bandwidth would be 45 kHz, basically the same as the upper limit of your sound card. As a comparison, an 80 by 60 pixel image at 37.5 fps with ordinary progressive scanning would require a bandwidth of 90 kHz, so the Baird system is a huge improvement in this case.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

HD-NBTV.

Postby Stephen » Wed May 16, 2007 9:34 am

Another possibility is to use my variation of Mr. Baird's system in the patent that does not require special apertures. See my article about this at http://www.taswegian.com/NBTV/images/SDS.pdf , and in particular Figure 5. Instead of the regular Baird 30 by 70 line vertical scan format, one could substitute a 112 by 63 pixel scanning area, laterally scanned with 63 lines. An optical system would project an 84 pixels wide by 63 pixel high image onto the left side of the scanning area. To the right of that would be three smaller images, each 28 pixels wide by 21 pixels high, so that they line up in a column along the right side of the scanning area. The three smaller images would be image-reversed. The combination would then all fit within the 112 by 63 pixel scanning area, that is 112 or 84 + 28 pixels wide by 63 or 21 + 21 + 21 pixels high.

This arrangement would use a 63 aperture scanning disc with a single spiral and square or round apertures. It would provide an 84 by 63 pixel landscape image at a frame rate of 50 fps. The bandwidth would be 44.1 kHz, basically the same as your sound card.

This variation is actually more efficient than the embodiments that Mr. Baird described in his patent, and it uses a standard scanning disc configuration. It does require an optical converter system on the camera and the display to generate or combine the large and small images though.

I have attached a figure of the modified arrangement in Figure 5A. Figure 5A assumes left to right, top to bottom lateral scanning. The large image for fine detail would be 84 pixels wide by 63 pixels high. Each of the small images for coarse detail would be 28 pixels wide by 21 pixels high. They are each image-reversed compared to the large image. As I explained in the article, with red, green and blue filters over each of the small images at the camera and display, the system can show coarse detail in full colour in a field-sequential manner with no increase in bandwidth. I have shown each of the small images in colour to illustrate this. The large image for detail remains shades of grey, of course.
Attachments
Figure 5A.jpg
Figure 5A
Figure 5A.jpg (12.7 KiB) Viewed 6043 times
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby DrZarkov » Wed May 16, 2007 4:36 pm

Phantastic! Nothing for a beginner with low mechanical skills (the monitor is relative easy, if it is possible to get a professional made disc, but the camera is horror...), but very convincing. I think you should really go to the maximum of this technology (what I mean is 48 x 48 pixel, as you described instead of just club standard), to get the "ultimate" NBTV monitor.
User avatar
DrZarkov
I think I've had a cranial implosion.
 
Posts: 1037
Joined: Mon Feb 19, 2007 11:28 pm
Location: Kamp-Lintfort, Germany

HD-NBTV.

Postby Stephen » Wed May 16, 2007 11:33 pm

The other nice thing about this scheme is that is applicable to more than narrow-bandwidth TV. For instance, one could project the fine and coarse detail images onto the face of a monochrome CCD camera sensor and monochrome LCD to convert a monochrome standard definition system to full colour with no additional electronics. All you need is a camera optical system to project the images on the face of the CCD sensor and a display optical system for the LCD to combine the fine and coarse detail images into a single full colour image.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby AncientBrit » Thu May 17, 2007 11:11 pm

Stephen,

In your discussion on the proposed 84x63x50 format you mentioned a B/W of 44.1kHz as being the same as that of a sound card.

In fact whilst the sound card samples at 44.1kHz, the actual signal B/W after filtering is nearer to 20kHz.

The theoretical upper limit is 1/2 of sample frequency, but for practical purposes it's normally safer to consider a value closer to 1/3 of sample rate.

Regards,

GL
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Postby Stephen » Thu May 17, 2007 11:26 pm

AncientBrit wrote:Stephen,
In fact whilst the sound card samples at 44.1kHz, the actual signal B/W after filtering is nearer to 20kHz.
Thanks, GL. Actually, I was wondering the same thing about Volker's sound card. Volker, do the specifications for the sound card indicate a frequency response up to 44 kHz, or is that the sampling frequency? As GL points out, if the bandwidth of the sound card is 44 kHz, the sampling rate would have to be just over twice that at a minimum.
Stephen
User avatar
Stephen
Anyone have a spare straightjacket?
 
Posts: 427
Joined: Fri Jan 26, 2007 7:00 am

Postby gary » Fri May 18, 2007 1:36 am

Most soundcards these days are AC'97 codec based. These codecs appear to have a 'native' sample rate of 48kHz, but support a number of other sample rates up to 96kHz by using on-board sample rate conversion.

Unfortunately, as my many experiments have shown, this SRC is quite poor in relation to temporal quality, i.e. when converting to other sample rates the time base is inaccurate such that over a period of time the number of samples per second is slightly out. This has ramifications when recording NBTV at rates other than 48kHz as the number of samples per line (even if a fraction as in 44.1kHz) varies over time. I suspect the same is true when playing out, thus CDs, when played out of a soundcard will play a little fast or slow. However this may be alleviated if the items are recorded at 48kHz, or, if playing out, sample rate converted to 48kHz using a reasonable quality software convertor.

So, if using the soundcard for recording or playing out NBTV, depending on the codec used, you should get better quality by sticking to 48kHz (or an integer multiple thereof).

And yes, of course, the effective bandwidth is, by Nyquist-Shannon theorem, at best, half of the sampling rate.
User avatar
gary
"Fester! Don't do that to 'Thing'"
 
Posts: 2303
Joined: Sat Jan 27, 2007 11:29 am
Location: Bundanoon, Australia

Sample rates.

Postby Steve Anderson » Fri May 18, 2007 2:23 pm

This has ramifications when recording NBTV at rates other than 48kHz as the number of samples per line (even if a fraction as in 44.1kHz) varies over time.


Well, you've made up my mind. I have been dithering over the past few days as to what sample rate to use for the .wav files I intend to create. Now it's a very definate 48kHz.

I often wondered how PCs generated some of these 'awkward' frequencies, like 44.1kHz, I wonder why it was chosen in the first place? Wouldn't 44kHz (exact) been easier. I guess Sony and Philips had their reasons.

48kHz means we get an exact 400Hz line rate on playback, plus an easier samples per line number of 120. It does sadly make the files 9% larger.

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 3965
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Next

Return to Mechanical NBTV

Who is online

Users browsing this forum: No registered users and 4 guests