HPX2700 or PMW350?

[Cris Daniels]

http://www.dvinfo.net/forum/panasoni...0-varicam.html




Well I'm sorry but these native 720p cameras clearly advertise their 1080p capabilites, simply check Panasonic's own website to see how these units are advertised. I do not consider a 720p front end, covered by a "full 1080p processing engine", a 1080 camera.

I know they are fine cameras and I wont argue that one bit, but to say there is no discernable difference between 720p and 1080p is simply untrue. What I would believe is a broadcast enginneer is looking at a native 720 camera that says it shoots 1080, and then says "the 1080 footage really isn't any better".

Basic logic would tell you that an image with real 921,600 pixels of information versus an image in the same space (both videos on a 42" Plasma for example) as real 2,073,600 pixels. Based on both the the math and my eyes, I feel like there is a clear visual difference.


I never record the EX's in interlaced, to me interlaced looks horrendous no matter the camera. I'm not saying a lot of people dont do it and need that capability but that is not what I would ever shoot.

[Jeff Regan]

Cris,

Yes, I agree with what you just posted, except the part where you can see a big difference at normal viewing distances. On your previous post, you mentioned a Panasonic camera, no model number and pixel shifting--this is not a native 720P camera, that was my only point.

I definitely agree with your view of 1080/60i or any interlace signal, and feel that XDCAM EX would be the last codec I'd want to shoot it with. I do have clients that want 1080/60i, typically HDCAM and DVCPRO HD.

David,

The MTF curves of a 2700 and 3700 are quite similar. My engineer friend was adamant that no 1080/24Psf or 30Psf is broadcast in the U.S.

There is a difference in temporal resolution between 1080/60i and 720/60P. The smoother,
more detailed motion of the latter is a big part of why ESPN chose that format for sports programming, ditto Fox.

Yes, I do believe that a higher pixel count pays dividends on resolving fine details in wide shots, I've gone to 65mm movie screenings and there is a big difference with wide shots, closeups and medium shots, not so much.

I certainly understand the on paper differences in pixel count between 720 and 1080, even though for years much of what has been seen from HD video sources has been sub-sampled 1080(and 720), so only in the last couple of years have full raster one-piece cameras become more common.

Going by comments that full sample 1080 is such a huge, glaringly improved image over full sample 720, one would think that every F900 series camera, every Varicam 27 series camera would have been retired on the spot. Reality is that DP's have a lot of other priorities when it comes to image making and camera feature sets--not just pixel counting. Same reason a Panavison Genesis and Sony F35 are 1080 cameras, not 4K.

Jeff Regan
Shooting Star Video

[Steve Connor]

Originally Posted by Jeff Regan

Cris,


I was not referencing an average viewer, but a seasoned broadcast television engineer who works in network affiliates, sat trucks, microwave trucks. I trust his eyes more than any average viewer.


Jeff Regan
Shooting Star Video

Yes, but we make programmes FOR the average viewer not "Seasoned" broadcast television engineers and many people actually CAN tell the difference.

But if he's a seasoned broadcast engineer he must be right and my seasoned online editor trained eyes must be wrong

[Jeff Regan]

Steve,

How close is your evaluation monitor to your sitting position?

Jeff Regan
Shooting Star Video

[Alister Chapman]

Originally Posted by Jeff Regan

The P2 720P cameras scale to 1080 in a 32 bit processing environment and are recorded onto a full sample AVC-Intra codec, not DVCPRO HD that sub-samples 720P to 960 horizontal rectangular pixels.

But it still isn't what I would regard as real 1080. Its 720 in a 1080 wrapper. IMHO this is worse than using pixel shift. You cannot gain resolution that is not there in the first place, certainly not with the kind of processing going on in-camera. No wonder you cant see a difference between 720 and 1080.... because in this case there isn't one.

Originally Posted by Jeff Regan

Live action, same scene at 720 and 1080, not visible to me in normal conditions, not visible to TV display reveiwers, not visible to experienced broadcast engineer, not visible to DLP projection engineer, never mind small screens, I'm talking large screens.

There you go again "In normal conditions" but not all conditions are normal. What about all the people that have posted to this thread that say they
see a clear difference? Are you suggesting they are wrong? Maybe not everyone thinks there is a difference but clearly many people can see a difference. If you can't accept that, then that's great for me and all those that can see the difference because we won't have to worry about any competition from you in the future. TV will move on and you will still be adamant that 720P is good enough.
I am a seasoned engineer with 25+ years of broadcast experience. I have a degree in electronic engineering. One of my first jobs was developing cameras for the BBC to use in F1 racing and the world rally championship and I'm not talking about camera systems, I'm talking about board level design. I can fault find to component level (a skill that is becoming rare theses days) and I absolutely do know what I am talking about.

Originally Posted by Jeff Regan

Given two CCD's of the same size, one with full raster 1080, the other with full raster 720, the latter will have better light sensitivity or less noise for the same amount of sensitivity.

Oh really? Explain 1920x1080, F13 @ 2000 lux and 59db noise figure. A full stop more sensitive than the 1280x720 HPX2700 and 5 db less noise. Sensor performance depends on the design of the sensor, pixel level noise reduction, microlenses and much much more. That statement that a 720 sensor will be more sensitive than a 1080 sensor is just far to general. Yes if you had identically designed pixels, microlenses etc and then made one chip with bigger microlenses and pixels it may be true, but sensor design is rarely that straight forward . Even if your argument was correct, just how much sensitivity do you need? I struggle with the PDW-700 in bright light, it's simply too sensitive at times.

Originally Posted by Jeff Regan

XDCAM EX in 1080/60i is very poor, so that limits any camera using that codec.

Your opinion and one not shared by a lot of well regarded individuals and organizations.

Originally Posted by Jeff Regan

EX1 and EX3 don't use a low pass filter in front of the sensors and are known to produce lots of IR contamination. Only circular polarizer filters should be used, not linear. I hope the 350 has a low pass filter in it, every other 2/3" camera I know of does.

They do have a low pass filter, if they did not then you would have all kinds of IR issues. Pointing an IR device at an EX will clearly demonstrate that they do have an IR cut filter. The problem is that there is a trade off between red reproduction and far red cut-off. I agree it would be nice if the cameras didn't exhibit the hue shift on certain man made fabrics under certain lighting conditions, but anyone that's used the Tiffen T1 on an EX1 will also tell you that a lot of the cameras "richness" is lost. The EX is not alone with this problem. Adam Wilts test of the 350 show that the issue is drastically reduced, possibly to the point where any hue is actually comparable to what people actually see with the naked eye.

[Steve Connor]

Originally Posted by Jeff Regan

Steve,

How close is your evaluation monitor to your sitting position?

Jeff Regan
Shooting Star Video

I have a broadcast HD monitor about 4 feet away and a client HD monitor about 8 feet away, I can spot the difference on both.

This is becoming a ridiculous argument, you are trying to tell people that the differences they can see aren't actually there. Lots of people have said it on this thread are you saying they are all wrong?

[Steve Phillipps]

Originally Posted by Cris Daniels

The difference in quality is so big that I almost cringe kicking down to 720p to overcrank, I already know that in post I will see the resolution drop when I intercut that footage into my native 1080p files.

Is that with an EX? I'm sure you'd find it very different with a Varicam.
Don't know if you've had the new "Life" series in the US yet, but over here I've heard nothing but good comments about the quality of the pictures. I only did a small amount of it but know that everyone else shot the same as me - tape Varicam.
Steve

[Steve Phillipps]

Originally Posted by Cris Daniels

http://www.dvinfo.net/forum/panasoni...0-varicam.html




Well I'm sorry but these native 720p cameras clearly advertise their 1080p capabilites, simply check Panasonic's own website to see how these units are advertised. I do not consider a 720p front end, covered by a "full 1080p processing engine", a 1080 camera.

I agree entirely, there is no way they can be claimed to be 1080 - it's very misleading and several producers I've spoken to have been under the impression that they are 1080 cameras - which of course they sort of are, but as you say, not really.
Steve

[Jeff Regan]

Originally Posted by Alister Chapman

But it still isn't what I would regard as real 1080. Its 720 in a 1080 wrapper. IMHO this is worse than using pixel shift. You cannot gain resolution that is not there in the first place, certainly not with the kind of processing going on in-camera. No wonder you cant see a difference between 720 and 1080.... because in this case there isn't one.

edit
Oh really? Explain 1920x1080, F13 @ 2000 lux and 59db noise figure. A full stop more sensitive than the 1280x720 HPX2700 and 5 db less noise. Sensor performance depends on the design of the sensor, pixel level noise reduction, microlenses and much much more. That statement that a 720 sensor will be more sensitive than a 1080 sensor is just far to general.
edit
Your opinion and one not shared by a lot of well regarded individuals and organizations.
edit
They do have a low pass filter, if they did not then you would have all kinds of IR issues. Pointing an IR device at an EX will clearly demonstrate that they do have an IR cut filter. The problem is that there is a trade off between red reproduction and far red cut-off. I agree it would be nice if the cameras didn't exhibit the hue shift on certain man made fabrics under certain lighting conditions, but anyone that's used the Tiffen T1 on an EX1 will also tell you that a lot of the cameras "richness" is lost.

Alister,

In-camera upscaling is done in a 32 bit processing environment, doing the same in an 8-bit environment, once the image is recorded is not as good, but I agree that if content is not resolved in the first place, it can't be manufactured later.

Please remember that I own an EX1, I have done lots of viewing of the same scene shot in both resolutions with that camera. If I thought the EX1 was all the camera any of my clients would ever need, I wouldn't have bought an HDX900 or HPX2700. I do not see myself spending three times as much as an EX1 for another XDCAM EX camera, despite its
2/3" sensors.

The difference in sensitivity between 720 and 1080 CCD's is what I cited, the 350 uses CMOS. If you took a native 720 CMOS sensor of the same size as a 1080 sensor(not sure any are available in a camera), all else being equal, the former would have a sensitivity advantage, if CMOS responds similarly to CCD's. The 2/3" CMOS sensors in the 350 look to be very fast and quiet, but we also know that they suffer from the same CMOS artifacts as the 1/2" EX1 CMOS sensors.

Adam Wilt has told me personally that 1080/60i shows the weaknesses of XDCAM EX, it's not just my observations.

You're correct about the low pass filtering, most of the problems are far red, not IR. Clearly Sony decided that filtering out more near red would affect the color rendering characteristics of the camera too much. Again, another CMOS artifact that I don't worry about with CCD's. If this has been improved on the 350, that's good, even if image skew and flash band artifacts haven't.

Again, all else being equal between a 1080 native camera and a 720 native camera, I would choose the former, and I do own an EX1. However, my P2 Varicam suits the needs of many high end DP's better than an EX1. Many DP's are resistant to CMOS sensors, prefer high bit rate, 10-bit, I-frame, 4:2:2 codecs, love the frame rate flexibility and look of the Varicam and Panasonic cameras in general. Some will opt for the 3700 instead.

At the end of the day, it's my own eyes and the needs of my clients that count. Doesn't mean anybody on this thread is wrong. Alister, I do not doubt your credentials in any way.

Jeff Regan
Shooting Star Video

[Alister Chapman]

Jeff, this discussion isn't about EX1's it's about PMW-350's, a totally different camera, not just different sensors but different processing, capabilities, form factor etc.

So what about the PDW-700 sensitivity and noise, almost exactly the same as the 2700. Your argument still doesn't hold water. On both CCD and CMOS sensors the light sensitive bit, the pixel or MOS capacitor is in many cases exactly the same technology. Sensitivity and dynamic range are limited by the MOS capacitors, it's the way the pixels are connected and the amount of electronics on the chip that's different between CMOS and CCD. Again design variations in individual sensors are what makes the difference.

Yes CMOS has artifacts. Artifacts that are only present in certain types of shot. Skew is rarely noticeable, let alone a problem. Agreed, Flash and strobe lighting handling is different to CCD and as I said at the very outset, if you shoot under strobe lighting day in day out, consider a different camera. IMHO shooting 720P compromises every shot, not just a few very specific circumstances. Sure there will be circumstances where you have no choice such as overcrank, but for everything else why compromise?

If your have clients that will make your camera pay then you have not made a bad choice. Perhaps for you specific needs you have made the best choice. But right now I personally just would not want to invest in yesterdays technology.

[Jeff Regan]

Alister,

I bring up the EX1 because you made it sound like I don't have access to a full raster camera. I also brought it up because it features the same low end version of XDCAM that the 350 has.

Are you saying that if the CCD's in a 700 were native 720P, with larger photo sites per pixel, there wouldn't be a sensitivity advantage? Or were you saying that the 700 and 2700 are the same sensitivity wise? Even in 720P mode?

Regarding image skew from CMOS, it is a problem. I have DP's that refuse an EX1 on that basis.

Just as you believe a native 720P sensor compromises every shot, I believe the same about low bit rate, 8-bit, 4:2:0 codecs. That is yesterday's technology to me. Of course one can always strap on a nanoFlash for higher bit rate and 4:2:2, but it's still 8-bit.

Anyway, I look forward to seeing the 350 at a trade show this month in SF. If I get clients asking for it, I will give strong consideration to buying it.

Jeff Regan
Shooting Star Video

[David Heath]

Originally Posted by Jeff Regan

The MTF curves of a 2700 and 3700 are quite similar.

I just don't see how you can say that, when one has 2 megapixel chips, the other 1 megapixel. The 3700 will deliver quite a high percentage mtf around 8-900 lines, the 2700 will only deliver aliasing at that spatial frequency! How can they have "quite similar" mtfs!!

Originally Posted by Jeff Regan

My engineer friend was adamant that no 1080/24Psf or 30Psf is broadcast in the U.S.

I certainly understand what he described ("60 fields interlace, 3:2 cadence from film source or 23.98 frame video") to be psf. If he disagrees, I'm very interested to know exactly what he does think psf means? It may be a matter of differing nomenclature? That doesn't change the fundamental point - you can still get full 24/25fps progressive resolution through a 1080i system, whatever you call it. Over here, it's definitely psf.

Originally Posted by Jeff Regan

Of course one can always strap on a nanoFlash for higher bit rate and 4:2:2, but it's still 8-bit.

Meaning? So have DVCProHD, HDCAM etc been, but how much has it mattered in the past?

OK, everyone wants more and more, and 10 bit may well be more than 8. But at the moment, the whole business is one of compromise, and it's necessary to ask which factors are significant and which aren't. And if 10 bit means higher compression within the frame, it's necessary to ask if the precious bits may not be better spent in other ways. Have you ever looked at anything in the past, on HDCAM or DVCProHD, and said "that looks bad, if only it was 10 bit....!"

I am in absolutely no doubt that given a choice between 10 bit and 1920x1080 chips, one or the other, I'd take the latter any day. If you have an EX, and have done side by side comparisons with it in 720p and 1080p modes, I just can't believe you can't see a big difference, and it seems that's the feeling of a lot of others here. Obviously, I'd like both 10 bit AND full raster chips if possible, and far higher bitrates than either XDCAM 422 or AVC-Intra, but aren't we now into HDCAM-SR territory and prices? I also don't deny the 50Mbs XDCAM codec would be desirable, but as you yourself say - there's always the nanoFlash option.

Still disagree? The EBU tests came out in favour of full 1920x1080 chips and recording, at least 4:2:2 XDCAM or AVC-Intra 100 recording codec, and chips of 1/2" or above as the most significant factors for next generation cameras, and specifically downplayed the advantage of 10 bit for all but the very top end systems. I'm more inclined to believe them than any manufacturer speaking with vested interests.

[Jeff Regan]

David,

I was told by Panasonic at Varicamp that the MTF curves of the 2700 and 3700 are similar, I have it in my notes.

My engineer friend said no 24 or 30Psf in the U.S.--not part of the ATSC digital standard.

I find it interesting for all that think full raster 1080 is such a big deal that for years we were all shooting with sub-sampled HDCAM and DVCPRO HD, many still are--yet full raster is mandatory for a good looking image all of a sudden? Absurd. Have I seen banding from 8-bit codecs in post? Yes. Noise? Yes. Compression artifacts? Yes. Intra 100 has less noise and more bit depth than any 8-bit codec I've seen. I have not seen compression artifacts thus far. I certainly have with DVCPRO HD.

I have a high end DIT friend in Hollywood and he dismisses 8-bit codecs out of hand. His favorite camera is the F35, although he ends up using the RED a lot.

He never used Panasonic P2 cameras or took them seriously until AVC-Intra was released. Now they shoot AVC-Intra 100 and convert the files to DPX for the DI.

Jeff Regan
Shooting Star Video

[Jeff Regan]

Simon,

They do seem to be a bit of a game changer, assuming things develop as planned. Awesome latitude with super low noise. But do we know much about what they're recording onto? The form factor is very lunch boxesque, but it seems to work for RED One--if not the operators who have to hand hold it.

Jeff Regan
Shooting Star Video

[David Heath]

Originally Posted by Jeff Regan

I was told by Panasonic at Varicamp that the MTF curves of the 2700 and 3700 are similar, I have it in my notes.

Jeff - just think about it. Look at the results and charts for those cameras with your own eyes. The 3700 has a high mtf % at places where the 2700 has none. How on earth can they therefore have similar mtf curves?

Either the person from Panasonic doesn't have a clue what they are talking about (Can you name who it was?), your notes are in error, or you're having a laugh at all of us. Which is it?

Originally Posted by Jeff Regan

...... yet full raster is mandatory for a good looking image all of a sudden?

"Mandatory" is your word, nobody else's. The EBU recommendations are just that - recommendations. Albeit independent of vested interest, and after a lot of science. See http://tech.ebu.ch/docs/r/r124.pdf if you don't believe me.

You are making a lot of the supposed huge significance of 10 bit v 8 bit, but the EBU tests simply don't bear that out. From the same document:

For acquisition of mainstream HD material, it is recommended that

• 8-bit bit-depth is sufficient for mainstream programme (10-bit bit-depth is preferred for
high-end acquisition).

and by "high-end acquisition", we are not talking about cameras in the price range of the 2700 or 350. It's quite clear that they regard full raster chips as far more significant than 10 bit. And there's more. Even for broadcast PRODUCTION (let alone acquisition) 10 bit doesn't seem to have a high importance compared to other factors:

• For normal moving pictures, an 8-bit bit-depth in production will not significantly degrade the HD picture quality at the consumer’s premises.

I don't know what your connection with Panasonic is, but their assertions of the "huge significance" of 10 bit are simply not borne out independently. It's one factor, yes, but relatively minor compared to others, even for mainstream HD broadcast production. But marketing departments love big numbers, and science is for geeks anyway.......

I'll let others make up their own minds as to who is right. Which is most important, 10 bit or full raster camera chips.

My personal feeling is that it's the more down to earth features of the 350 that really make it so appealing anyway. Power consumption - media costs - media versatility - weight - ............