Post by Myrtone November 2, 2007 (1 of 9)
|
|
I have had this idea for a long time, actually since I first heard about SACD. Digital video, like digital audio has been until recently, is always linear PCM (compressed formats like MPEG 2, are encoded from PCM, and decoded back to PCM prior to digital to analogue conversion). Let's start with a comparison of digital audio and video; In digital audio, quantisation error's cyclycal behavior can manifest itself as added frequency content in the wave heard disturbing high frequency noise, in images, it can result in an similar artifact known as paint by numbers or colour banding. Other things that still puzzle me about digital video, I know that steep filters (especially in the analogue domain) can cause phase shifts and ringing, and the like. I figure that if it adds audible distortion to signals representing sound, similar horizontal filtering would introduce visible distortion in their video counterparts too, similarly I have never heard of jitter in images, though still, I wouldn't be surprised it it does occur, or even introduce visible artifacts. More about dither in both applications at http://en.wikipedia.org/wiki/Dither
Anyway, I would like to know what others think of the application of DSD to video, possibly as a digital archival medium for film stock and analogue video recordings. I have had quite a few ideas over the years. I first thought of just applying the usual DSD-audio technique of reducing noise in band and increasing noise above the bandwidth, but I have had other ideas since. Look over at http://commons.wikimedia.org/wiki/User:Gerbrant/Dithering_algorithms and you will find many, some are actually more like noise shaping, if DSD can be more analogue like with audio, would it produce more film like results with video? Note that in the audio domain, DSD can be converted to analogue by means of an analogue lowpass filter, I realise, because of that, that DSD-video could be converted, either by a spatial light modulator (such as DLP), plus a blur filter (can they all be called Gaussian filters, or are their other types?), similar to what is found in a digital camera.
|
|
|
Post by tommwi November 2, 2007 (2 of 9)
|
|
Myrtone said:
Anyway, I would like to know what others think of the application of DSD to video
First I think we should be clear about what DSD is. It is a trademark of Sony!
Sometimes, actually very often, DSD is used to describe what is called a "digital signal" or "digital format". Well, it is all about definition of course, but DSD uses a generalised technique called PDM (Pulse Density Modulation). There is a special case of PDM, called PWM (Pulse Width Modulation). This technique is already at use in the laser disc application. So it cold be argued that "DSD" is already operating in this disc format. In principle there is no difference.
|
|
|
Post by Myrtone November 2, 2007 (3 of 9)
|
|
tommwi said:
First I think we should be clear about what DSD is. It is a trademark of Sony!
Sometimes, actually very often, DSD is used to describe what is called a "digital signal" or "digital format". Well, it is all about definition of course, but DSD uses a generalised technique called PDM (Pulse Density Modulation). There is a special case of PDM, called PWM (Pulse Width Modulation). This technique is already at use in the laser disc application. So it cold be argued that "DSD" is already operating in this disc format. In principle there is no difference.
DSD was not invented by Sony or Phillips but was developed by Ed Meitner. The difference between PWM and PDM, is that in the former, the carrier frequency is constant irrespective of the balance in duration between that two states, but in the latter, it varies inversely with it. Laserdisc (apparently, some wish this were still around) is, in fact, analogue, rather than digital, and the modulation is only horizontal for each line, LD uses an FM carrier with other carriers superimposed over the top, clipped to form a square wave, which is frequency modulated by the video signal and pulse width modulated by all others. There is a difference here, I am talking about digital applications here.
|
|
|
|
|
|
Correct. LaserDisc (the analog, double-sided 12" disc -- what an interesting analogy with vinyl) uses Pulse-Density Modulation which is very similar to DSD. See http://en.wikipedia.org/wiki/Direct_Stream_Digital (esp. the last paragraph)
Now I just wonder: since LaserDisc is an analog medium (the pits and lands on the optical disc don't represent 0 and 1 but the top and the bottom of the block wave and can, in principle, have any random length) does it have infinite resolution?
I guess there is a minimum length to the pits and lands and this defines the resultion.
Any ideas?
|
|
|
Post by Aleph November 2, 2007 (5 of 9)
|
|
The Seventh Taylor said:
Now I just wonder: since LaserDisc is an analog medium (the pits and lands on the optical disc don't represent 0 and 1 but the top and the bottom of the block wave and can, in principle, have any random length) does it have infinite resolution?
I guess there is a minimum length to the pits and lands and this defines the resultion.
Any ideas?
Well... I guess a certain George Mann decided to use the Laserdiscs as the basis to his newly promoted "ultimate analogue disc". An interview about the system can be found analoglovers.com, Teresa's web-site.
http://www.analoglovers.com/page17.html
|
|
|
Post by Myrtone November 2, 2007 (6 of 9)
|
|
The Seventh Taylor said:
Correct. LaserDisc (the analog, double-sided 12" disc -- what an interesting analogy with vinyl) uses Pulse-Density Modulation which is very similar to DSD. See http://en.wikipedia.org/wiki/Direct_Stream_Digital (esp. the last paragraph)
Now I just wonder: since LaserDisc is an analog medium (the pits and lands on the optical disc don't represent 0 and 1 but the top and the bottom of the block wave and can, in principle, have any random length) does it have infinite resolution?
I guess there is a minimum length to the pits and lands and this defines the resultion.
Any ideas?
As I said above, it is quite different, the wikipedia article does say that the technique is similar, but seems inaccurate claiming that it is decoded in the same way, I certainly don't think so.
Anyway, I think my idea could be used to improve DLP projectors. They are often called "digital" light amplifiers and rightly so in one sense, because each pixel is basically an electro-photonic switching amplifier. Current practice is to use temporal PWM to control the intensity of each pixel, perhaps a very poor technique, I can think of a better one. How about increasing the pixel resolution to 64^2 or 256^2 (^2 because it is spatial and has two dimensions) times the pixel resolution of the incoming video, and applying noise shaping, then individual pixels could be rendered invisible by means of a Gaussian filter, resulting in a very film like image. By also multiplying the incoming frame rate and adding motion blur filters, we would no longer need to rely of persitance of vision to convey the illusion of motion.
You might want to read http://www.microsoft.com/whdc/archive/ovrsamp.mspx , though, what's not surprising for a paper written in the digital era is that it does seem somewhat inaccurate about analogue TV (I'm using analogue in the true sense here, not the refer to picture formats and scanning systems that happen to play role in analogue). For one, image resizing is possible in the analogue domain and was done between, for example the Brtish 405 line system and the French 819 line system. Also, even analogue television can be progressively scanned, I know Baird's early Mechanical television standards were, though I haven't heard of an all-analogue electronic television system like this, maybe Baird's later electronic television standards were, I'm not sure. I know he wanted high definition very early on. I supposed that if he had had lived longer, the idea of spatial oversampling would have occurred to him.
Anyway, vertical filtering is quite feasible in the analogue domain and was used in older PAL and SECAM encoding.
|
|
|
|
|
|
Myrtone said:
I would like to know what others think of the application of DSD to video, possibly as a digital archival medium for film stock and analogue video recordings.
Perhaps we could call the format Super Video CD (SVCD)? ;-)
|
|
|
Post by Myrtone November 28, 2007 (8 of 9)
|
|
|
|
|
|
|
|
|
I know, hence the smiley. Thanks anyway.
|
|
