HPX 171 best recording workflow

[Jeff Regan]

Sub-sampling is used to allow lower bit rates, and/or higher color space, and/or I-Frame structure. It is a way to reduce bandwidth. This is not necessary with newer codecs that are more advanced and efficient, such as h.264 based codecs. Remember that HDCAM is a late '90's codec, DVCPRO HD is an early '00's codec. HDCAM is also 3:1:1 color space, another way of lowering bit rate and allowing for less compression, just as so many very low bit rate codecs have 4:2:0 color space, or Long GOP frame structure--this allows for less compression at a given bit rate. One need only look at HDCAM SR to see what a huge bit rate is needed for 4:4:4, 10-bit, full sample, I-Frame HD codecs.

All of the above is why my favorite in-board, affordable codec is AVC-Intra 100. They are all compromises, all compressed, all sub-sampled in some way--just a matter choosing your poison!

Jeff Regan
Shooting Star Video

[Robert Lane]

Sanjin,

The advice from Dan and Jeff to move away from the deep techno-junkie stuff is spot-on. At the end of the day it's all about how good any image looks both in-camera and in post. You can spend literally days driving yourself nuts trying to decipher all the technical differences between camera systems but it's absolutely nonsense, trust me. It's not the perspective to think about when buying *any* camera, period.

My company was one of the first to fully adopt the P2 system when it was introduced in 2005 and we started with only the HVX200. Through my own testing I came up with camera settings that made the footage from the 200 really pop and many thought it came from 1/2" inch camera, not a 1/3" inch with SD chips being pixel-shifted. In fact several DVinfo users like Kevin Railsback have taken those settings as a jump-off point and made them even more perfect and are getting output from cameras like the HPX170 that would blow you away - and the EX cameras.

The bigger point to consider - which has nothing to do with counting pixels or codec sub-sampling - is that of color. None of the EX cameras shoot a color space superior to DVCPRO's 4:2:2, it's all 4:2:0 just like HDV. And as I always try to remind people, once the camera has thrown away color information there's no way to somehow magically regain it - once it's gone, it's gone, period. And color my friend, is far more important than overall resolution.

Think of it this way: AJA created the KiPro, a stand-alone tapeless recorder that takes HD-SDI from any source and homogenizes it into ProRes 422. It was designed in response primarily to the HDV and EX series cameras that shoot both the inferior color space but a long-GOP format, both of which have huge consequences in post that most seem to constantlly gloss over - until they have to deal with it directly. If 4:2:0 long-GOP wasn't such a problem there's no way AJA would have spent millions in development and deployment of such a device.

All P2 cameras already shoot 4:2:2 "i-frame" formats so there's no need for a device like the KiPro - or the Convergent design device you're considering.

Lastly, I'm as brand-agostic as they come. Every product on the market is nothing but a tool and I spend hundreds of hours each year testing various products to see how they stand up to real-world use and my own scrutiny. If the EX series or any other HDV-based camera actually produced better-looking and easier-to-handle footage than the HPX170 or HPX500 I'd be using them. But in point of fact, they don't. The only cameras that do? The HPX2000, 2700 and above.

Get the HPX170, learn how to properly set it up (you can find my baseline settings for the HVX200 on my site and use that as a jumping off point) and you won't be disappointed.

[Sanjin Svajger]

Yes Robert, you are right. All this thinking and calculating, it's driving me crazy.
But what happens when I want to broadcast the material from this cam over a HD TV station, who output full res...? And plus, this cam without it's 4:2:2 subsampling would have a to low chroma res for it to be useful with HD. So in my opinion Panasonic made good - they made a camera with small chips, saved money and added dvcpro hd codec with 4:2:2 subsampling (imagine it had 4:2:0 mpeg 2 - how usable would the material be). And the 4:2:2 subsampling is all that "everybody" see's and buys the camera - not realizing that the EX with it's full res chips and 4:2:0 subsampling delivers the same or a better chroma res.
So, my point being: don't rely just on your subjective view, read some specs before buying (but don't go all crazy) and get the camera that best suits your needs.

[Robert Lane]

The mantra is always, "buy what fits your needs" however your logic about subsampling is actually backwards. It's been proven time and again - starting way back with Adam Wilt and his testing of the various DV codecs - that in fact 4:2:2 is superior to 4:2:0. Logically it has to be, what do you think happens to that "zero" in the EX codec? It doesn't magically reappear later somewhere, it's gone! Adding more pixels doesn't make up for a loss of chroma, it just makes that loss *sharper*. Think about it.

And with respect to the 200 or 170 not being able to pull off broadcast-quality finals that's also a myth; several national ad campaigns have been shot on the HVX200 and the HPX170. Did you see the Doritos Superbowl ad from 2007?

HVX200 shoots Doritos' Super Bowl ad - Online Video

Do you think Frito-Lay would have used a camera on a $2.5M spot that couldn't have pulled off commercial-grade?

If you're stuck on the higher pixel-count of the EX series then Sony's glossy marketing campaign has won you over. But if you were to actually compare the output between the 170 and the EX's you'd most likely be surprised - especially when you know how to setup the camera.

[Jeff Regan]

My gaffer recently worked on two weeks of Ford national spots using HPX170's!

Jeff Regan
Shooting Star Video

[Sanjin Svajger]

Originally Posted by Robert Lane

4:2:2 is superior to 4:2:0. Logically it has to be, what do you think happens to that "zero" in the EX codec? It doesn't magically reappear later somewhere, it's gone! Adding more pixels doesn't make up for a loss of chroma, it just makes that loss *sharper*. Think about it.

The HPX chips have 960 x 540 pixels. The EX1-3 have 1920 x 1080 pixels. HPX uses pixel shift which increases luma res but doesn't increase chroma res. This is what I'm referring to.

So, what brings a better chroma res:

4:2:0 from 1920 x 1080 or
4:2:2 from 960 x 540
?

This is how I understand the whole matter anyway... And I'm not saying the HPX 171 doesn't deliver good pictures. I like the look it gives but am wondering that it isn't gonna hold up for HDTV 1080.

[David Heath]

Originally Posted by Robert Lane

The advice from Dan and Jeff to move away from the deep techno-junkie stuff is spot-on.

Robert - how you can say that and then come out with :

The bigger point to consider - which has nothing to do with counting pixels or codec sub-sampling - is that of color. None of the EX cameras shoot a color space superior to DVCPRO's 4:2:2, it's all 4:2:0 just like HDV.

and

AJA created the KiPro, a stand-alone tapeless recorder that takes HD-SDI from any source and homogenizes it into ProRes 422. It was designed in response primarily to the HDV and EX series cameras that shoot both the inferior color space but a long-GOP format, both of which have huge consequences in post that most seem to constantlly gloss over - until they have to deal with it directly. If 4:2:0 long-GOP wasn't such a problem there's no way AJA would have spent millions in development and deployment of such a device.

All P2 cameras already shoot 4:2:2 "i-frame" formats so there's no need for a device like the KiPro - or the Convergent design device you're considering.

leaves me wondering. Let's all either ignore technical details totally - and there's a lot to be said for all of us doing just that - or give the full picture. Don't just quote the cherry-picked stats that Panasonic promote.

Having said that, it's not right to not justify why I don't feel you're giving an accurate picture with the details you quote. It's all been said before, but for Sanjin's benefit, here goes.

You make great play on the difference between 4:2:0 and 4:2:2. You don't mention the numbers are only ratios - not absolute values. Hence, if system A has twice the overall number of pixels of system B, then if it has 4:2:0 sampling, it will have exactly the same number of chrominance pixels as system B, if that has 4:2:2 sampling.

I've used the analogy before, but think of my offering you a quarter of the money in my left pocket, or half that in my right pocket. Which would you prefer? If I had $20 in my left, $10 in my right, you'd get $5 in each case. So the gut reaction of "half must be better than a quarter!!!" needn't be right - so why be taken in with blind statements such as "4:2:0 must be better than 4:2:2!" Quoting colour space figures whilst ignoring the wider picture is meaningless.

And colour sampling only applies to recording formats anyway. The comparison doesn't say anything about the camera front ends. And these are the stats Panasonic is less keen to have quoted about most of their range - especially the HVX200 and the HPX500. They are only capable of producing vertical colour detail of 540 lines max - even though the recording system would be capable of recording 1080 lines. (At least in 1080p mode.) Pour a litre into a 2 litre pot and you only have 1 litre still.

Robert, you say:

Lastly, I'm as brand-agostic as they come. .....

then you go on to say:

My company was one of the first to fully adopt the P2 system when it was introduced in 2005 and we started with only the HVX200.

Do you not think the last 5 years may have given you a certain bias?

If the EX series or any other HDV-based camera actually produced better-looking and easier-to-handle footage than the HPX170 or HPX500 I'd be using them. But in point of fact, they don't. The only cameras that do? The HPX2000, 2700 and above.

Firstly, the EX series cameras are not HDV - the HDV spec specifically lays down 25Mbs. Secondly, I've given details in another thread of very formal tests done by the British Society of Cinematographers in which (and alongside many others, including 35mm film, and digital cinema cameras) the difference between the HVX200 and the EX3 was chalk and cheese - in favour of the EX3. Thirdly, if you really are brand agnostic, why only list Panasonic cameras as models that you think give superior footage to the HPX170 or HPX500?

The bigger point to consider - which has nothing to do with counting pixels or codec sub-sampling - is that of color. None of the EX cameras shoot a color space superior to DVCPRO's 4:2:2, it's all 4:2:0 just like HDV. And as I always try to remind people, once the camera has thrown away color information there's no way to somehow magically regain it - once it's gone, it's gone, period.

Why don't you remind people about what happens in the horizontal direction, Robert? About how DVCProHD is only capable of recording a horizontal luminance resolution of 1280, whilst the EX codec does 1920? So how DVCProHD with it's 4:2:2 then only manages 640 chroma samples against the 960 of the EX codec? Or do you think throwing away horizontal colour somehow doesn't matter, as long as it's preserved in the vertical direction?

To say nothing of cameras like the HVX200 not being able to resolve enough colour information to do a 1080 4:2:2 system justice - if the camera doesn't make enough color information originally, there's no way to somehow magically regain it - once it's not been produced, it's not been produced, period.

The chips on the 200, 171 and 500 are 960x540. Chips on standard PAL cameras are 720x576. That's 518,400 versus 414,720, or only about 25% more!!! Just compare that to a camera with 1920x1080 chips - 2,073,600 in total, or no less than **400%** more!

And color my friend, is far more important than overall resolution.

Just what evidence do you have for that statement? Because it goes against almost every single bit of relevant physical theory I've ever read. All the scientific evidence indicates that the human eye is far more sensitive to luminance detail than colour detail. Your statement is contrary to the basic principles behind every single broadcast colour TV system so far engineered. In NTSC the chrominance bandwidth is far less than luminance, and PAL takes it a stage further by averaging chroma line by line in the vertical direction. Reason? Overall resolution is far more important than chrominance resolution.

[David Heath]

Originally Posted by Sanjin Svajger

So, what brings a better chroma res:

4:2:0 from 1920 x 1080 or
4:2:2 from 960 x 540
?

You have to define the systems you're working with, most especially the system luminance resolution. 4:2:2 then means half horiz, equal vert. 4:2:0 means half in each direction. You then have to take into account what the chips are capable of generating. (The result is only as strong as the weakest link.)

So, assuming we are talking about an EX and an HPX171.

For the EX, expect a chroma res of 960x540. (Limited in each case by the subsampling.)

For the 171, expect a chroma res of 640x540. (Limited horizontally by the subsampling, vertically by the chip resolution.)

In practice, you have to consider lens resolutions and a lot of other things, and progressive/interlace will have a big effect.

Don't get too bothered about chroma resolution anyway. Either of them should be fine in practice, and both are less than ideal for such as chroma key. (Though a nanoFlash largely sorts that for an EX.) Luminance resolution will have a far higher order effect - and it doesn't just mean sharper pictures, it means you can turn the detail correction down and get less edgy pictures which still look sharp.

[Dan Brockett]

Sanjin:

I just shot a PBS doc with the HPX170. In the States, the PBS Red Book specifications are widely regarded as the most technically stringent and difficult to pass of any broadcaster. I shot 1080 60i because most the archival material in the show had been telecined from film masters to 1080 60i. The footage from the HPX170 passed with flying colors, no pun intended. The footage looks superb when broadcast, the clients were happy and the show is running on PBS affiliates all over the country during their pledge drives. The director called me last week, saying that he has another PBS project that he wants me to DP, he loved the look that we obtained on the last project and wants to do more.

I also had a producer call me the other day. We worked together last year to shoot green screen interviews for a very popular FOX animated series. This producer, then went on to work at a different production company late last year and worked with other DPs to shoot green screen interviews. She called me this week and said that FOX wants her to work with me again, SPECIFICALLY BECAUSE OF HOW BEAUTIFUL, FLAWLESS AND CLEAN the green screen footage I shot for them was (shot also with my HPX170). Let me re-iterate, a television network awarded her work for three series that I will be shooting specifically because of the look I was able to obtain with my HPX170! The comment that the FOX client made to her was that some of the other green screen she had shot with a different DP, using a much higher end camera "looked like a used car commercial".

I think I have made my point. We can talk numbers and specs all day, or we can talk about tangible results that I have experienced and the money that the camera has made me. This camera has already paid for itself three times in the past 18 months and it continues to pay me dividends, no mean feat in this time of such a horrible economy. My point is not that the HPX170 is a better camera than the EX1/EX3, it is not. FWIW, I have also shot green screen with the EX1 and it too keyed fine. I just wanted to clarify, in practical terms, that many of your concerns about the DVCPRO HD codec, 960x540 chipsets and chroma luminance are, at best, a very minor thing unless you get paid for measuring and debating specs. As for me, my 170 continues to make me money (along with a few skills I have ;-)

Dan

[Sanjin Svajger]

Originally Posted by David Heath

In practice, you have to consider lens resolutions and a lot of other things, and progressive/interlace will have a big effect.

Yes I know but don't quite understand "lens resolution"... What about the A/D converter, that's an important part of the chain to. And wouldn't the vertical res of the HPX be 960/2 = 480 ?

Originally Posted by Dan Brockett

Sanjin:
The comment that the FOX client made to her was that some of the other green screen she had shot with a different DP, using a much higher end camera "looked like a used car commercial".

I think that the other person, who used a "much higher end camera" a) didn't know what he or she was doing or b) had the wrong camera for the job maybe or c) just did something wrong in post.

Anyway, congrats for the FOX gig! Here in our little country everything is a bit unstable (due to recession) and getting such a solid gig is very nice.

I just have a one last question. Would it be better for a camera to use pixel shift even if it doesn't need it to achieve the desired resolution? Example: EX1,3 using pixel shift to get 1920 x 1080.

[David Heath]

All that pixel counting can tell you is *the best possible case*. It can predict a max performance above which it's theoretically impossible for the camera performance to be. But put a poor lens on, and it can be worse - it's another case of only being as strong as the weakest link. (In practice, what gets measured tends to agree quite closely with predictions, though resolution tends to fade gently away - don't expect any camera to resolve 540 lines perfectly, but output mushy grey when presented with 541!)

A-D convertors are a separate issue, they influence other matters than resolution. Noise, gamma etc yes - but not resolution.

And wouldn't the vertical res of the HPX be 960/2 = 480 ?

No, I think you're confusing the vertical with the horizontal.

The HPX has 540 pixels vertically, 540 each of red, green and blue. That limits it's chroma resolution to no more than 540 - which the codec can easily preserve.

For luminance you have to take the pixel shifting into account. The overall benefit is generally accepted to be an improvement factor of about 1.5x overall. Here, the benefit is shared between hor and vert, so expect about sq rt 1.5x on each axis (about 1.2x). This is where the (approx) prediction for 1200x650 luminance resolution comes from. All I'll say is it agrees quite well with measurements.

I just have a one last question. Would it be better for a camera to use pixel shift even if it doesn't need it to achieve the desired resolution? Example: EX1,3 using pixel shift to get 1920 x 1080.

I think the answer has to be "no". It gets highly complex, but I can think of two main reasons.

Firstly, to get up to around 1920x1080 output resolution, then you'd need to divide those figures by 1.2 to get to approx the number of photosites needed on the chips. So somewhere around 1600x900. But you then have to process a 1600x900 matrix to get a 1920x1080 output. It may be possible, but far more difficult than processing 960x540 in a 1920x1080 signal processor. So the question becomes whether it's worth it?

Secondly, there's the subject of aliasing. Normally, there is a filter to stop detail finer than the basic resolution of the chip from getting through - because that's what causes aliasing. For pixel shift to work, you have to allow higher frequencies through, and the nett result is the inevitability of coloured aliases. It may not be the end of the world, but why bother?

The only benefit to pixel shifted 1600x900 chips in the situation you describe is that the individual photosites are larger than is the case with 1920x1080 for the same overall chip size, so you may expect better sensitivity, or lower noise. In practice, the additional processing means you wouldn't get all that may be expected anyway. It's just easier to use 1920x1080 - and the real beauty of what Sony have done with the EX is to increase the OVERALL sensor size to 1/2". So bigger individual photosites AND 1920x1080 compared to 1/3" designs.

But nothing stands still, and it'll be interesting to see what Canon come up with. The codec news is encouraging - full broadcast approved XDCAM 422 50Mbs. The big question is what the front end will be. If it's 1/3" chips, my own feeling will be "so near and yet so far", but who knows? The new Canon may be the EX beater.

[Robert Lane]

David, you and Sanjin are two peas from the same pod; you're both overly concerned and focused on theory, scientific data and numbers more than real-world results. If shopping for a camera is all about the science rather than end-of-day results then there'll never be any camera at any price that will satisfy your need for technical perfection.

In point of fact, the reason I chose and continue to use Panasonic cameras isn't because of some brand-blinding loyalty or commissions (of which there are none, thank you) it's because I have spent my time and money physically testing *all* the various systems offered on the market. I've had them all in-hand with production face-time, not staring down numbers on a sheet of paper or listening to opinions from others who've never actually held the camera. And the reason I settled on and continue to use Panny cams is because they produce the best-looking imagery, have the easiest and most logical workflow and provide a set of features that fit *my company's* needs.

Truly, if Sony, JVC, Canon or even RED had produced a system that *on-the-whole* was a better system all around then I'd be using it. In fact, when RED was first announced I fully expected that I'd drop using the Panny's in favor of the RED system but after real-world use it turns out it's not well suited to *my companys* work at all.

Are the 170/500 or any other P2 camera perfect? Well gee, is anything produced by humans? Of course not, but when you step away from the infinitesimally finite and narrow-focused view of tech specs and actually work with these various systems and put them through their paces it simply turns out that P2 - and the Panny system as a whole - stand out clearly above the competitors in all the mission-critical areas required for successful commercial production. Period.

[Sanjin Svajger]

Originally Posted by David Heath

But nothing stands still, and it'll be interesting to see what Canon come up with. The codec news is encouraging - full broadcast approved XDCAM 422 50Mbs. The big question is what the front end will be. If it's 1/3" chips, my own feeling will be "so near and yet so far", but who knows? The new Canon may be the EX beater.

I'm eager to see what canon brings to the tables to. EX1 and EX3 have had and have the throne for the last 2 years now. It would be about time for the competition to pick up their pace.

Hi Robert! I myself don't rely only on subjective viewing. There's a distinct way of how I operate and function in general. For me, it's important to understand the "mechanics" and the basics of matter. It's from that knowledge that I build up my understanding. Relying just on subjective viewing in my opinion is negligent and not professional. That's just like a video editor not knowing what a jump cut or continuous editing is and in his work relying only on his felling or "gut", or not understanding the basics of narrative form, etc. That's wrong. And I know a lot of those kind. These are generally the people who refuse to read books and to study and are usually the most unprofessional people to work with.

For good understanding of matter one needs knowledge and experience. Not just subjective knowledge shaped from experience. Your probably coming from: that the likes of me and david "can't see the forest anymore because of the trees". This sometimes happens, but generally it's all about being clever.

Sorry if my articulation isn't at it's best - I'm rely tired...

[David Heath]

Originally Posted by Robert Lane

David, you and Sanjin are two peas from the same pod; you're both overly concerned and focused on theory, scientific data and numbers more than real-world results. If shopping for a camera is all about the science rather than end-of-day results then there'll never be any camera at any price that will satisfy your need for technical perfection.

Robert, you can rest assured that it's end-of-day results that are my prime interest. But you seem to make it sound that "theory, scientific data and numbers" and "real-world results" are totally unconnected. I'll be the first to admit that the theory doesn't - can't - tell the whole story, but it can give you a very good idea. As long as you understand it properly. Most importantly, theory can often accurately predict when a claim can't be true - but only when the numbers are accurate. Either get the science right - or ignore it.

And what I see (from more than one manufacturer) is cherry picking of statistics. Make a big deal of the numbers that suit them, downplay the ones that don't. In Panasonics case "4:2:2 v 4:2:0" is maybe the best example. If all else is equal, the statement is true enough - but in what we've been talking about, all else is far from equal - hence the comparison becomes meaningless.

........when you step away from the infinitesimally finite and narrow-focused view of tech specs and actually work with these various systems and put them through their paces it simply turns out that P2 - and the Panny system as a whole - stand out clearly above the competitors in all the mission-critical areas required for successful commercial production. Period.

Well, that's pretty tendencious stuff. You make it sound so cut and dried - but I certainly know many many people and organisations who feel they have gone into the whole subject thoroughly and decided that the Panny system does not stand head and shoulders above the rest. P2 worked well for a few people from day one - many many others felt the "download and format" workflow (especially in the field) was just too prone to human error. (And either stayed with tape or went to disc.)

There's no doubt that the viability of solid state is rapidly improving (and I'd expect it to become dominant before long) - but now P2 has it's competitors from others who have let the technology mature before bringing product to market.

[David Heath]

Originally Posted by Sanjin Svajger

Hi Robert! I myself don't rely only on subjective viewing. There's a distinct way of how I operate and function in general. For me, it's important to understand the "mechanics" and the basics of matter. It's from that knowledge that I build up my understanding....

Yes. It's a bit like driving a car. You don't need any real knowledge of mechanics or how engines work to get a licence, and you could be a reasonably good driver without such knowledge.

But in my experience the BEST drivers have at the very least a rough idea of what's going on under the bonnet. It gives an empathy with the car which can only help, quite apart from being able to perform routine maintenance. It can also help them judge whether an issue is important ("stop NOW!"), or something which needs looking at - but can wait until they get home.

It can also help make a more informed decision when it comes to buying a new one. Otherwise you buy it on the basis of "oohh, what a nice colour", or (worse) some half truth a salesman tells you.