How PsF video from the V1 is different than "p" or "F" video

[Steve Mullen]

Originally Posted by Piotr Wozniacki

Steve, simply with VLC streaming from Capture Device (uses DirecDraw). NO line twitter - just like with captured files. Now, what do you make out of this?!!

The i.LINK should be getting an exact copy of the bits on tape. Assuming you are using VLC with deinterlace DISABLED -- that would mean the 2 fields are going into 1 frame that is repeated twice per second. This is the correct way of presenting 25p.

So what's happening with the camera analog & HDMI output?

I assume it's sending out one field then the next field. Each field has only 540-lines. It is an interlaced signal which means a monitor will treat it as interlaced. Static video will have it's fields woven together (weave) into a frame and presented twice per second. No twitter.

What happens with motion video is based on the monitor's deinterlacer. It might bob, or do something much better.

If bob: the first field's odd lines go into a 1080-line frame. The 540 missing (even) lines are interpolated from the existing 540-lines. Now this frame is presented for 1/50th second.

Then, the second field's even lines go into a 1080-line frame. The 540 missing (odd) lines are interpolated from the existing 540-lines. Now this frame is presented for 1/50th second.

Herein lies the problem. Watch the top odd line: it goes from real to interpolated to real to interpolated.

A thin line will fall on a single row. Assume the line falls on top of real line. The next interpolated line CANNOT have this thin line. So when you watch, every 1/50th of a second it dissappears. On and off. On and off.

bob generates flicker. Hence the name bob.

I'll bet if you plug the V1 into an HDTV that does not do bob the twitter will go away. There are two levels of Bravia. One is a cheap unit typically sold in Asia, etc. The S2 if I remember.

Look for an XBR HDTV.

I think the reason we see this only on the E/P is because the CMOS sampling rate is higher than it is for the U -- yet the chips have the same number of rows. This causes more aliasing, but also captures more V rez. hence the E/P actually records finer vertical detail. Hence, the greater possibility of very "thin" lines.

If I'm correct, it is an unhappy combination of the V1's geater V. rez. plus its interlaced PsF plus a bob monitor's deinterlacer. A real bad combination.

Can anyone guess why the E/P might have a CMOS sampling rate is higher than it is for the U?

[Steve Mullen]

Originally Posted by Bob Grant

As I pointed out before, here's the problem. Vegas cannot print to tape 25PsF the same as the camera records. That's straight from the Vegas engineers. Also there's no way to PTT 25p as per Canon either.

So Vegas is a one way solution. That would be like editing 24p from the DVX100 and never being able to record 24p back to the camcorder.

Now I wonder if it also cannot record 30p and 24p back to the V1U.

Spot?

[Piotr Wozniacki]

Originally Posted by Steve Mullen

So Vegas is a one way solution. That would be like editing 24p from the DVX100 and never being able to record 24p back to the camcorder.

Now I wonder if it also cannot record 30p and 24p back to the V1U.

Spot?

Yes, from the timeline you can only print a 1080/50(60)i back to tape, but as I mentioned earlier you can print a 25p file from your HDD as well.

Steve, what would your "bottom line" verdict on the V1E's 25PsF be, based on what we established? Your theory on why the line twitter appears (and spoils the picture) under certain conditions boil down to how the video is watched and NOT to how it's produced - do I get you right?

[Piotr Wozniacki]

Originally Posted by Steve Mullen

Look for an XBR HDTV.

Steve, I cannot find XBR in the context of European Bravia models, and the local dealer never heard of it. Any hints?

[Steve Mullen]

Originally Posted by Piotr Wozniacki

but as I mentioned earlier you can print a 25p file from your HDD as well.

How do you write a 25p file from your HDD using Vegas to the V1?

Later today I'm going to burn your 25p and play it on my hd dvd play. This would be like your using a 50i movie.

lets see if the twitter stays gone.

I'm heading to bed.

[Piotr Wozniacki]

Sorry Steve, my bad. Yes you can pick a progressive movie file from your HDD and print it to tape from within Vegas, but it is recorded in the interlaced format anyway.

[Bob Grant]

Originally Posted by Steve Mullen

So Vegas is a one way solution. That would be like editing 24p from the DVX100 and never being able to record 24p back to the camcorder.

Now I wonder if it also cannot record 30p and 24p back to the V1U.

Spot?

No 30PsF back to tape either, only 24p. Keep in mind though that as far as I'm aware there's no way to send a frame down component.
One can pretty easily convert 25p to 24p in Vegas without any resampling, you can do a precise frame rate conversion. There's at least one utility to do this in FCP too.

As to are the bit clocks faster in the V1E.
I'd suspect they might be. In SD DV they're the same 13.5MHz, the lower PAL frame rate yields more pixels in the same bps. What happens in HD where there's the same number of pixels and a different frame rate?
My best guess is the bitclock runs faster and hence yields more res.

[Bob Grant]

Originally Posted by Steve Mullen

The i.LINK should be getting an exact copy of the bits on tape. Assuming you are using VLC with deinterlace DISABLED -- that would mean the 2 fields are going into 1 frame that is repeated twice per second. This is the correct way of presenting 25p.

So what's happening with the camera analog & HDMI output?

I assume it's sending out one field then the next field. Each field has only 540-lines. It is an interlaced signal which means a monitor will treat it as interlaced. Static video will have it's fields woven together (weave) into a frame and presented twice per second. No twitter.

What happens with motion video is based on the monitor's deinterlacer. It might bob, or do something much better.

If bob: the first field's odd lines go into a 1080-line frame. The 540 missing (even) lines are interpolated from the existing 540-lines. Now this frame is presented for 1/50th second.

Then, the second field's even lines go into a 1080-line frame. The 540 missing (odd) lines are interpolated from the existing 540-lines. Now this frame is presented for 1/50th second.

Herein lies the problem. Watch the top odd line: it goes from real to interpolated to real to interpolated.

A thin line will fall on a single row. Assume the line falls on top of real line. The next interpolated line CANNOT have this thin line. So when you watch, every 1/50th of a second it dissappears. On and off. On and off.

bob generates flicker. Hence the name bob.

I'll bet if you plug the V1 into an HDTV that does not do bob the twitter will go away. There are two levels of Bravia. One is a cheap unit typically sold in Asia, etc. The S2 if I remember.

Look for an XBR HDTV.

I think the reason we see this only on the E/P is because the CMOS sampling rate is higher than it is for the U -- yet the chips have the same number of rows. This causes more aliasing, but also captures more V rez. hence the E/P actually records finer vertical detail. Hence, the greater possibility of very "thin" lines.

If I'm correct, it is an unhappy combination of the V1's geater V. rez. plus its interlaced PsF plus a bob monitor's deinterlacer. A real bad combination.

Can anyone guess why the E/P might have a CMOS sampling rate is higher than it is for the U?

Your assumption about static video at least on the Barvia V series would seem incorrect unfortunately. With the V1P connected HDMI to the display and a locked down camera with nothing moving in the frame it most certainly can twitter.

Thanks for your explaination of how bob works, from what I've seen on the Bravia it looks exactly like what you'd get if that's what it's doing, as you've rightly said it shouldn't be doing that, it should be weaving.

Is this just a sloppy display device or is it something the camera is sending it fooling it into bobbing when it should be weaving?

I need to go back and very carefully recheck this but I'm 80% certain that my 25p footage shot at 1/25th on a bright sunny day shows no twitter on the Bravia. With the camera connected directly via HDMI to the Bravia and pointed into our flat well lit office, no line twitter in 25p. Move the camera around to point at a dimly lit page with fine horizontal lines, twitter.

How is the display being fooled, my best guess, noise.
I should check this by simply pointing the camera at the same problematic subject and changing the light level while letting the camera raise gain.

[Piotr Wozniacki]

Originally Posted by Bob Grant

I need to go back and very carefully recheck this but I'm 80% certain that my 25p footage shot at 1/25th on a bright sunny day shows no twitter on the Bravia. With the camera connected directly via HDMI to the Bravia and pointed into our flat well lit office, no line twitter in 25p. Move the camera around to point at a dimly lit page with fine horizontal lines, twitter.

How is the display being fooled, my best guess, noise.
I should check this by simply pointing the camera at the same problematic subject and changing the light level while letting the camera raise gain.

I'm afraid I cannot confirm this observation Bob. When on HDMI or Component, the line twitter is present on ALL contrasty, nearly horizontal lines - also in those picture areas where there is virtually no noise at all.

[Bob Grant]

Originally Posted by Piotr Wozniacki

I'm afraid I cannot confirm this observation Bob. When on HDMI or Component, the line twitter is present on ALL contrasty, nearly horizontal lines - also in those picture areas where there is virtually no noise at all.

Thanks for that.
I'll still do the tests later this week.
On this very matter though here's what has me confused.

With a locked off shot, no camera or subject motion. What the camera is sending to the Bravia will be the same whether the camera is in I or P, two fields with no temporal offset. Now Steve has said the resolution is the same in I or P, OK. So why does the Bravia not display any twitter in I and yet it does in P?

In theory in I a camera is supposed to employ line averaging to avoid twitter but from what Steve was saying some time ago, this camera doesn't do that. I'm not so certain this is correct. The noise levels are lower in I and the twitter is missing from the display.

[Piotr Wozniacki]

Originally Posted by Bob Grant

The noise levels are lower in I and the twitter is missing from the display.

Confirmed here 100%, but the twitter aside, the noise in 25p I can live with as the price for the ultra-sharp picture I'm getting with sharpness levels up to 12 (I didn't test higher). And there are many situations (depending on contents etc) where sharpness of 4 or 5 is enough, and the noise (dancing pixels) are negligible then.

If we could nail down the true nature of the line twitter and devise an optimal method of delivering 25p without it, I'm sold on the V1E.

[Steve Mullen]

Originally Posted by Bob Grant

Your assumption about static video at least on the Barvia V series would seem incorrect unfortunately. With the V1P connected HDMI to the display and a locked down camera with nothing moving in the frame it most certainly can twitter.


Is this just a sloppy display device or is it something the camera is sending it fooling it into bobbing when it should be weaving?

How is the display being fooled, my best guess, noise.
I should check this by simply pointing the camera at the same problematic subject and changing the light level while letting the camera raise gain.

I think you are onto something here. Besides twitter, I find that as S is increased the whole picture gets "busy" on fine details. This dancing dot noise is aliasing. So is the barber pole banding you see on thin objects. All this is "noise" heading toward the display.

So it could be that even on static VIDEO, the changing noise accross fields is enough to trigger the deinterlacer into not weaving.


"Steve was saying some time ago, this camera doesn't do that ..."

I've worried about this. Last night I rechecked the stuff from Japan. The cmos signals pass through a signal-procesing chip on the way to the eip. I now think that in interlace mode a flicker filter is engaged which will also slightly reduce noise. This confirms your finding.

Later I'll burn my hd dvd and see what happens when it feeds the signal to the hdtv.

[Douglas Spotted Eagle]

Originally Posted by Bob Grant

Thanks for that.

In theory in I a camera is supposed to employ line averaging to avoid twitter but from what Steve was saying some time ago, this camera doesn't do that. I'm not so certain this is correct. The noise levels are lower in I and the twitter is missing from the display.

Bob, as mentioned in my post a while back (twitter being different than...etc)
it was mentioned that the V1 does indeed have a set level average to avoid twitter. Sony confirmed the feature a while back.

[Steve Mullen]

Originally Posted by Douglas Spotted Eagle

Bob, as mentioned in my post a while back (twitter being different than...etc)
it was mentioned that the V1 does indeed have a set level average to avoid twitter. Sony confirmed the feature a while back.

I wonder why Sony hasn't adjusted the settings for the E/P. Turning down Sharpness is a brute force solution as it kills both H. rez. (bad, and totally unneeded) as well as V. rez. (ok, but not ideal).

A flicker filter is a much more precise way of handling flicker.

[Bob Grant]

Originally Posted by Douglas Spotted Eagle

Bob, as mentioned in my post a while back (twitter being different than...etc)
it was mentioned that the V1 does indeed have a set level average to avoid twitter. Sony confirmed the feature a while back.

Must admit I'd missed that one, thanks.