Tech Note Why The Panasonic Had Poor Resolution

only 960x560

only 960x560

Berga said:
I think the major reason for the problems with the HVX200 is the small CCD. It is only 960x560. They use pixel shift to make a resolution of 720.

The result is a much light sensitive camera but also a camera with lower resolutions.

All sub $10000 cameras has their pro and cons. I start thinking that Sony PWD-F330 would be the best low cost alternative. But it is sub $20000.
Click to expand...


I've tried to come up with a model to calculate CCD resolution vs. image resolution. The model also uses the Airy disk radius for a given F Stop. However, there are still too many unknowns where I have to make too many rough guesses.

Rough Guess x Another Rough Guess x Yet Another Rough Guess = Useless Answer :laugh:
​
A native resolution of 960x560 seems a bit too low to work, even with the VERY rough model I have. Maybe 1024x582 might work at the low end, but do NOT take this number as fact, because my model is still VERY crude.


Bob Diaz
 
BobDiaz said:
Maybe 1024x582 might work at the low end, but do NOT take this number as fact, because my model is still VERY crude.


Bob Diaz
Click to expand...
If the CCD was 1024x582 than why is the HVX's horizontal messured resolution as low as it is?
 
If the CCD was 1024x582 (never heard of such a sensor) and used pixel shifting in both directions (as Jan said on a presentation) it should be able to do 1536x873 pixels which is 864x873lines.

Reel stream gets similar or better resolution and it uses 720x480 sensors. So it is either a bad camera, or a mistake in user settings (lens, focus?), or an implementation bug, or a low res sensor.

I see the "Jan statement" thread is locked and it had 1000 views. This is definetely a hot subject. Imagine if Jan actually responded and released some info!
 
SergejIvanovits said:
If the CCD was 1024x582 than why is the HVX's horizontal messured resolution as low as it is?
Click to expand...

The 550 measured lines are 16:9 Tv lines which is 550x16/9=978lines. 978x550 is really close to 1024x582.
 
pertierr said:
Just for the record, the DVX has three 770x492 pixel-shifted sensors.
Click to expand...

Are all pixels active? (they actually contain color on the output?) There should be dark pixels arround. 20 on each side or something like that.
 
Hmmm... yes... the HVX chart looks very "steppy" and "digital" compared to the others as well, which Barry says may be attributable to the Raylight codec... but it's very curious. How closely does Reel-Stream work with Panasonic, and has there been any discussion of this between them? I'd like to imagine that they talked at least a bit during the HVX design process...
 
770x492 is the active pixel count. You are correct that there are 'inactive' pixels around the edge used for various purposes such as black level adjustment, however these are not included in the 770x492 figure.

About the HVX charts, we now have 3 sets of results which confirm the same approximate numbers...i don't think it is a focus issue.

Cheers,
Juan
 
I don't know if this has been mentioned, or even relevant to this thread. But if you are on a Mac and have QTpro open any of the HVX clips that people have posted and under properties then video you can set the video to hight quality. For some reason all QT movies play at low quality out of the bag. If you do the above you will se a very slight boost in quality most of the time.

\v
 
Ullanta:

We didn't have any contact with Panasonic related to the HVX design process. Our work is completely independent from Panasonic, at least up to date.

Cheers,
Juan
 
ullanta said:
Hmmm... yes... the HVX chart looks very "steppy" and "digital" compared to the others as well, which Barry says may be attributable to the Raylight codec... but it's very curious. How closely does Reel-Stream work with Panasonic, and has there been any discussion of this between them? I'd like to imagine that they talked at least a bit during the HVX design process...
Click to expand...

The Combi-HVX200.jpg chart on the 6cam test does not look digital or steppy. The steppiness on other tests came from pushing the sharpness too far IMHO. If I had to choose I would choose lower resolution and better image. I believe all cameras should be tested at a point that has zero sharpening artifacts. That's what most people want to use. If we don't do that, it's just a competition over artifacts and sharpening algorithms, that does not does not change anything in the bottom line. Sharpening does not increase resolution. I just improves MTF slightly. More contrast near the limit. But also more contrast on the problems that are in that region:)
 
pertierr,

I know it is not your intention, but the HVX chart you have in your website is full of artifacts. It is in no case representative of what the HVX is capable of.
I've seen recorded footage and res charts. It looks so much better than this chart, really.

The HVX is capable of producing smooth images.
Not clean and defined as the Andromeda, but way way way cleaner that the chart you've posted. Maybe it's the scaling or the Raylight, dont know, but you should remove it from your website.

BTW, you should work on a HVX/Andromeda project.
 
That chart was shot by Barry Green, and i've made that clear all over the post.

The reason why I decided to post that chart was because:
A.) It shows more resolution than all the other HVX tests and
B.)It's the same standard EIA chart as the other camera tests(Z1, HD100, etc) so it makes for a better direct comparison.

You are right that the other charts are 'smoother' but they show considerably less resolution. If I posted those, then you would probably be telling me that the HVX was capable of producing better resolution. :)

Since this is a resolution comparison, not an artifact comparison, I chose the HVX chart that showed the best results so far.

I don't see how I can be more fair. It's not my fault the HVX produces the results it does.

Cheers,
Juan
 
I said when I posted it that it was stairsteppy and that it may be due to raylight or some other reasons. Now that I have the Mac working I'll try to re-import it and see if I get better results. If so, I trust that Juan would replace it with a more representative shot.
 
Absolutely. I made clear in the post that I picked the EIA charts shots for each camera that showed the best resolution results, if anything better arises i'll replace it.

As far as resolution goes, despite the artifacts the chart Barry shot does show considerably more resolution than any other HVX tests so far. I guess the question is, can we get these same resolution numbers without the artifacts somehow.

Cheers,
Juan
 
Also, given the Codec issue, we can assume that this is a RECORDED format chart (i.e., compressed); do we know if the ones posted by real-stream were the same?

If I'm not mistaken, this is the first RECORDED format chart from the HVX that we've seen...
 
I
f the CCD was 1024x582 than why is the HVX's horizontal messured resolution as low as it is?
Click to expand...

If the CCD was 1024x582 (never heard of such a sensor) and used pixel shifting in both directions (as Jan said on a presentation) it should be able to do 1536x873 pixels which is 864x873lines.
Click to expand...



First, the number I gave was only a wild guess, as I said...

I've tried to come up with a model to calculate CCD resolution vs. image resolution. The model also uses the Airy disk radius for a given F Stop. However, there are still too many unknowns where I have to make too many rough guesses.

Rough Guess x Another Rough Guess x Yet Another Rough Guess = Useless Answer

A native resolution of 960x560 seems a bit too low to work, even with the VERY rough model I have. Maybe 1024x582 might work at the low end, but do NOT take this number as fact, because my model is still VERY crude.
Click to expand...


The key words are marked in bold, like "Rough Guess", "unknowns", and "VERY crude". Do NOT take my number as FACT, as I slowly try to sort things out the numbers for the model changes. Add to this that there are several DIFFERENT charts with different numbers, so a solid number may not be possible.

I'm not going to go throug all the math, because I haven't a clear cut model.


"it should be able to do 1536x873 pixels which is 864x873lines"
Click to expand...

The increase of 1.5 for shifted pixels may not be correct. I've tried factors from 1.2 (a bit low) to 1.67 (a bit high). Right now around 1.4 looks reasonable, BUT this is still a work in progress.

IF the Kell Factor = 1.0, yes; BUT try x 0.80 or x 0.70 for the Kell Factor and the numbers are closer to what we saw in the charts.


Bob Diaz
 
You must log in or register to reply here.
Back
Top