PDA

View Full Version : Frequency/bandwidths of analog signals and their limits


fellaw
24th March 2004, 07:12
In another thread (http://forum.doom9.org/showthread.php?s=&threadid=73102) I made some thoughs about how much resolution is really needed when capturing from an analog PAL source. However this didn't really fit into the thread, that's why I created this one.

I was just wondering about the signal was build in a CRT and the effect on the recording of it.

Considering you have 1152 black and white lines to be drawn, You would need a black and a white line to be drawn on each of the 576 lines. But the ray can only be in either one state, black OR white. So 576 lines is the maximum resolution you'll get.
Same goes for the width. I was using 768 columns, because there aren't any pels in the analog world and there's still the 4:3 aspect ratio. The PAR and those 720 pels only come into play when digitising.

Now calc: 768 / 2 x 52µs = 7,4 MHz. This is the maximum frequency used for the width!

Of course it's no bandwidth like we're used to in the digital world, but it's kind of. When you need more information in the analog world, the frequency has to be higher. So you can consider the frequency of bandwidth.

The degree of the signal during the display caused the CCIR to limit the frequency because of the costs. Who needs high frequency if you can't display the information it carries anyway?

Why should one intend to use more than those 5/5.5/6 Mhz to record an analog signal if he needs nothing more than that to fulfill the specs? I don't know what really happens inside the studios, but I'm pretty sure they do their tricks and nothing more.

Of course, the technology has improved a lot. But when I have a look at some of my channels(e.g. MTV), I don't see any needs to go above 520 x 374, because the signal is even worse than that. Even when digitising, I can't get any more information than the signal gives me. All above is only noise. Some channels seem to use the full 7,4 MHz or at least 6.5. But some like MTV only use the minimum given by the CCIR.

I hope you can understand my thoughts now.

trevlac
24th March 2004, 17:41
Originally posted by fellaw
But when I have a look at some of my channels(e.g. MTV), I don't see any needs to go above 520 x 374, because the signal is even worse than that. Even when digitising, I can't get any more information than the signal gives me. All above is only noise. Some channels seem to use the full 7,4 MHz or at least 6.5. But some like MTV only use the minimum given by the CCIR.



In the grand theoretical scheme of things, you are kinda right. However, you are mixing many things ...

First the height.

A PAL signal is bcast as 576 active picture lines. These lines are no different from pixels. They ARE digital. The picture height has already been 'sampled' by the camera. So... regardless of kell, or how you want to talk about resolution, if you resample the 576 to 374, you are limiting the resolution. Granted, an all black signal does not contain much detail, but I would bet that broadcasters use as much vertical resolution as they can. Why? Bandwidth has nothing to do with vertical resolution. The signal always contains 576 active picture lines. No gain on less resolution. All analog PAL sources can contain this much resolution. All PAL TVs can display it.

Width
You are on the right track. A 5MHz signal is limited to 520 samples of resolution. There are some practical limits to what you do that don't make this a big issue.

- To get a proper displayed picture, you want to follow a standard
- Capture cards always sample at 1 rate and resize
- Contrast make more of a difference than bandwidth as far as how sharp the picture looks


Be happy to discuss more, but this has been covered in detail across many generations of this forum. :)

This is a nice read on the topic:
http://members.aol.com/ajaynejr/vidres.htm

fellaw
24th March 2004, 18:59
A PAL signal is bcast as 576 active picture lines. These lines are no different from pixels. They ARE digital. The picture height has already been 'sampled' by the camera. So... regardless of kell, or how you want to talk about resolution, if you resample the 576 to 374, you are limiting the resolution.

I agree with this. But do all of those 576 active lines carry different analog information from line to line? Signal lines next to each other probably contain basically the same information when being displayed on a tv, because of the Kell, don't they? Additionaly even the 576 could be influenced by noise caused due to signal losses. However, this depends very much on the source, the camera and the post-processing of the signal during playback(VCR, DVD, whatever). If PAL is made to be watched on a tv, why would one do better than 374 if you can't tell the difference between that and 576 anyway when using a tv? But I'm refering to the worst case, a black to white change each new line. Maybe I'm wrong with usual material and regarding digising.

You are on the right track. A 5MHz signal is limited to 520 samples of resolution. :D :D :( :(

OK, so it doesn't make any sense to capture bcast PAL with more than 520? Of course a higher resolution is better for post processing, but you can't add any real information with this if it's not there anyway.

I'll think I'll switch to either Half D1 or SVCD res for bcast capturing, depending on my focused media.

The article explained a few things, but most I knew already. Thx anyway!

trevlac
24th March 2004, 20:59
Originally posted by fellaw
If PAL is made to be watched on a tv, why would one do better than 374 if you can't tell the difference between that and 576 anyway when using a tv?

What is 374 vs 576? I guess 374 is a resolution measure and 576 is scan lines. How are you comparing these? As i said, 576 is the full resolution frame height. If you have less, you have less possible resoultion.


OK, so it doesn't make any sense to capture bcast PAL with more than 520?


I'll think I'll switch to either Half D1 or SVCD res for bcast capturing, depending on my focused media.

I'm not sure what you use, but my cheap BT878a card always captures over 1500 (luma pixels) when it samples the line. Cards sample higher than the bandwidth for many practical reasons like; line sync and ADC aliasing filtering to name 2.

So when you say 520, you are simply resizing (resampling) down to that size. You can do that in post or as you capture. Depends on what does a better job.

As far as SVCD and CVD frame sizes, BCast PAL can contain more detail than those formats hold. But to each his own. If you don't see it, there could be many reasons why.... bcast source, capture card pre filters, capture card resize filter, playback device, playback monitor.


.... but most I knew already. Thx anyway!

Glad you already knew. ;)

BTW: Just because I argee with 1 of your conclusions, does not mean I agree with how you got there.

fellaw
24th March 2004, 21:28
What is 374 vs 576? I guess 374 is a resolution measure and 576 is scan lines. How are you comparing these? As i said, 576 is the full resolution frame height. If you have less, you have less possible resoultion.
I meant the lines you can finally differ on a tv. IIRC, the Kell also affects the height, regardless how many lines had been sampled!
I'm not sure what you use...
I'm using a Terratec Cinergy 400, Philips SAA7134. I'm not exactly sure how much it samples really. Didn't find any tests so far and had no time to create a test DVD.
So when you say 520, you are simply resizing (resampling) down to that size. You can do that in post or as you capture. Depends on what does a better job. Cards sample higher than the bandwidth for many practical reasons like; line sync and ADC aliasing filtering to name 2.
OK, then I made a mistake. I thought the pels you select are the number of samples the card uses. Now I understood that the card samples at a fixed rate and the driver resizes the picture afterwards.
As far as SVCD and CVD frame sizes, BCast PAL can contain more detail than those formats hold.
Of course, because the res./sample frequency is higher and there are obviously no compression artifacts. :cool: :cool:

I think I am through with that topic now...

Next I'll have to work on the proper res. and PAR/DAR calculation for the card. Maybe I'll post a howto for the Terratec in this form when I'm done with it.

edit: Just deleted a few empty lines I didn't recognize...

Arachnotron
24th March 2004, 23:38
I'm using a Terratec Cinergy 400, Philips SAA7134. I'm not exactly sure how much it samples really. Didn't find any tests so far and had no time to create a test DVD.

I'll leave the rest of this discussion to Trevlac, but because I own the same card I will provide the info for you :)

This card operates at 27 MHz. This gives 1728 samples in the complete 64 µs PAL scanline. This is cropped to 52.15 µs/1408 samples

From these 1408 samples, whatever number of pixels you set as horizontal resolution are interpolated.

[edit] so if you want to go DVD with this card, cap 704x576 and you are ok as far as AR is concerned.

trevlac
25th March 2004, 03:59
Originally posted by fellaw
I meant the lines you can finally differ on a tv. IIRC, the Kell also affects the height, regardless how many lines had been sampled!


Here is my point. PAL bcast contains 576 lines. If you reduce it, you have less lines and less resolution. Regardless of how you view it. TV, PC CRT, PC LCD, Plasma, blah blah blah ....

So, if you wish to resample your ?x576 source to ?x374, you will loose resolution.

But if you tire of my babble .... let me know. :)

fellaw
25th March 2004, 17:52
@Arachnotron

Thx, that's I was looking for all night long yesterday. It seems 704 would be suitable, because 1408 mod 704 = 0 :D

@trevlac

I think I've got enough for now :)