Announcement

**Erik B** · 4 January 2000, 17:07

50 fps, I need a better slow motion

Need your helicopter perspective on this one. I will film a fairly fast moving liquid in a glass tube, transfer the film to PC for analyze picture by picture. For me the time resolution is important and I have read somewhere that it is possible to look at separate fields giving 50 fps. However I want to look at every field (left,rightleft...) in a sequence. I have a Sony trv900 camera (minDV) and is thinking about buying a DVraptor card (which comes with Premiere). Is this a good choice for my needs?

Comments, help, tip...
Erik Björk

**Stewman** · 4 January 2000, 17:27

That's one thing I'd like to see: A dv cam that can capture video at HIGHER frame-rates than NTSC. This is something that film cameras have been able to do for more than 40 years - I used to have an 8mm movie camera that could film at almost 70fps.

Regards,
Stew

------------------

**Erik B** · 5 January 2000, 06:10

Chris,

Perhaps I am wrong on this but is it not possible to build a picture (frame) out of a field (interpolation)? I am aware that the geometrical, vertical resolution will be the half but if time resolution is a priority?? You wrote:
"While it is true that PAL uses 50 fields per second, these are alternating "semi frames" which cover either alternate odd or even traced lines."
That means first field even lines, second field odd lines and so on right?
I am not sure if 25 fps is good enough for me. I was looking at a high speed system of 8000 fps but it was to expencive so I decided to use standard equipment and squeze the most out of it.

Erik

**This_Idiot** · 5 January 2000, 12:03

Hi Erik,

Yes, you picked up my gist over the fields correctly.

But I'm not aware of any software that can create single "pictures" from each field (rather than frame). Needless to say that doesn't mean there isn't such software.

I'm not sure how accurate you'd find it if you upped your shutter speed on the camera and then captured at normal speed. My understanding is that this should display in slow motion, which would stretch your timescale to some degree.

**fluggo99** · 5 January 2000, 13:54

According to Alexei's assertions in <a href="http://forums.murc.ws/ubb/Forum2/HTML/002591.html">Slow motion quality using MSP</a>, when going at 50% speed on a clip with the field options properly set, each field is turned into a separate frame.

Unfortunately, the reason for the thread was that Patrick was noticing poor quality on the output. I think the problem there was that one field represents the parts of the picture missing from the other field. That is, given a picture like so:

Code:

-----********------- (field A)
---*****--******---- (field B)
--****-------****--- (field A)
-***----**-----***-- (field B)

...MSP will use line-doubling to produce the following two frames:

Field A:

Code:

-----********-------
-----********-------
--****-------****---
--****-------****---

Field B:

Code:

---*****--******----
---*****--******----
-***----**-----***--
-***----**-----***--

...causing flickering on pictures with a plentitude of horizontalish stripes.

So yes, MSP would do the fields for you, but the results might not be great in some situations.

**Brian Ellis** · 5 January 2000, 17:45

Stew

I've little doubt that it would be technically possible to generate a video signal at, say, a multiple of 2, 4 or 8 times PAL or NTSC frame rates, which could be slowed down to the normal rates BUT you are going to need to double, quadruple... etc. the bandwidth. This means that your signal/noise ratio will go down (I think by 6 db for each doubling, all other things being equal). The really big thing is that tape would be OUT, so you would have to record on, say hard disk. Now, there ain't many hard disks with a data transfer rate for writing at, say, 40 Mb/s, not to mention that you'ld need a supercomputer to compress the raw data to usable proportions. If you were recording an ephemeral phenomenon lasting just a second or two, then you could use SDRAM as your medium until it could be transferred to disk but if you want to record an athlete swimming the 100 m freestyle or Martina Hingis knocking hell out of a ball during a long volley with Venus Williams,you would need a LOT of RAM. Bearing in mind such equipment would be hellishly expensive, the market would not be likely to develop to tempt any but the most specialised videoshooters from spending umpteen kilobucks, especially as chemical film is better for this kind of application (each frame is integral in time, whereas video distorts the shape of rapidly-moving objects because there is finite time between one line and the next, especially if interlaced). I think you will have to wait a long time before you see your heart's desire.

------------------
Brian (the terrible)

**Erik B** · 6 January 2000, 07:14

I think I found a solution of the problem made by Grigory when I looked deep in this archive. He wrote:

"-If you want to convert ordinary 25 fps interlaced movie to 50 fps non-interlaced, read my paper at <A HREF="http://www.geocities.com/ResearchTriangle/Forum/1022/papers/50FPS.html"" TARGET=_blank>http://www.geocities.com/ResearchTriangle/Forum/1022/papers/50FPS.html"</A>

I will comment this method on this thread when I got my Raptorcard.

Erik

**fluggo99** · 6 January 2000, 12:32

Funny, that's exactly the process I described.

**Erik B** · 6 January 2000, 13:57

Fluggo,

Sorry. I forgot to share my thoughts. This MSP program I can´t find at my shop. The Raptorcard, however comes with Premiere and Grigorys solution was with that software and Bingo! I thought. Your explanation was very clear and I undestand the flickering problem. However mostly I will look at picture by picture not the movie. I think it would be an easy task for the software to make an interpolation between the lines and not simply double them.

Erik

[This message has been edited by Erik B (edited 06 January 2000).]

**fluggo99** · 6 January 2000, 21:39

Searching through the internet for info on the Filter Factory, I came to the conclusion that it can't do the proper between-frame interpolation. (that is, interpolating between field A's and field B's on consecutive frames) This may be a limitation of the filter architecture (MSP doesn't appear to have the ability either; I know DirectX Media can't).

However, should you want to do interpolation within a frame, I've come up with the following modification of Grigory's formulas:

(1-t%2)*(r*(y%2)+(1-(y%2))*mix(src(x,y-1,2),src(x,y+1,2),1,2))+ ((t%2))*(mix(src(x,y-1,2),src(x,y+1,2),1,2)*(y%2)+r*(1-y%2)) for R
(1-t%2)*(g*(y%2)+(1-(y%2))*mix(src(x,y-1,1),src(x,y+1,1),1,2))+ ((t%2))*(mix(src(x,y-1,1),src(x,y+1,1),1,2)*(y%2)+g*(1-y%2)) for G
(1-t%2)*(b*(y%2)+(1-(y%2))*mix(src(x,y-1,0),src(x,y+1,0),1,2))+ ((t%2))*(mix(src(x,y-1,0),src(x,y+1,0),1,2)*(y%2)+b*(1-y%2)) for B

I couldn't test these (I don't use Premiere), but try them out. Note that you may still have to do the (1-t%2)|(t%2) swap.

**Patrick** · 6 January 2000, 22:12

Fluggo, is there anything you can add (in plain English please! ) that would lead me to believe that there is some way of improving the quality of slow motion while using MSP? (Over and above what was discussed in my thread that you've referred to.)

**jeff b** · 7 January 2000, 04:56

I had the use of a high speed slo-mo system at the last place I worked. It was a very expensive S-VHS all integrated on a rolling rack, which we used to troubleshoot our machinery. In retrospect, I would have to say that setting up to do the same kind of thing more cheaply would be mostly a matter of what was used for a camcorder. A mini-DV camcorder would allow one to pause playback, and then move forward frame by frame without breakup, the way that analog camcorders usually do. It seems to me that that would be the long and short of it: hook the mini-DV camcorder up to a big TV and look at the playback frame by frame. The only thing you'd need the computer for would be if you wanted to capture stills and make a report, I should think...

**fluggo99** · 7 January 2000, 15:17

Pat: Just a diagnosis of the problem is all I can provide. (Perhaps I should write a slow-motion program to cure this madness...)

(ahem) In plain American English: Using all frame-based settings like you said is probably the next-to-best if not the best way to do high quality slow motion in MSP. What results is complete pictures simply repeated. Placing the proper field order preserves smoother motion, but in scenes with high motion or horazontalish stripes, your way is probably better.

But it's not the best way to do it, MSP or not. The problem is, filters are written to work on only single frames at a time. My theory for a really good slow motion system involves working with two frames at the same time, and there's no way I know of to write a plug-in to do that job. Either Premiere or MSP has to provide it (just like they provide a speed control) or a standalone program has to be written to accomodate.

There you have it.

(I think I'll just try writing such a program...)

**Patrick** · 7 January 2000, 15:50

Thanks Fluggo, it's nice to know that I stumbled upon a better way to do slow motion with MSP all by myself.

By the way Fluggo, after reading numerous posts of yours, I find it difficult to believe that you're really a high school student. You're much too well spoken! Are you sure you're not a forty year old professor? If you're not actually pulling our legs, there's hope for the youth of today after all.