HPX 171 4:2:2 or EX3?
 

Hi all! As my first post Im just gonna go out and ask a question!

Im in the process of buying a new camera. I shoot mostly music videos and now moving on to commercials and short films, maybe documentaries... I'm thinking of buying either a Panasonic HPX 171 or a Sony EX3. I know this are two differently oriented cameras but the price difference (that was offered to me) between the EX1R and EX3 is negligible (700€).

I can get a used but practically a new HPX for about 3500 € from a friend and a new EX3 for 7200 €. I've already shoot with the HPX and got satisfactory result but haven't tried the EX yet (but have seen loads of material shoot with it). The HPX is a great camera when shooting 720p but comparing it's 1080p to the EX 1080p is a no brainer to me (even if shoot on SxS - mpge2 35 Mbit).

The problem here is: in a year - year and a half most of our TV stations are going HD. I'm inclined to buy the HPX but am worried about the broadcast quality of this camera. How much future proof is it - I can't really upgrade it, like I would the EX3 with buying the NanoFlash?

I've thought about this and I think for me it comes down to this: the HPX shoots 4:2:2 in dvcpro hd and the EX 4:2:0 mpeg long gop or 4:2:2 iframe or long gop if using the Convergent designs NanoFlash. What I don't understand here is what is the actual chroma resolution of the HPX in 1080p or 720p due to the pixel shift it uses? In my calculations the HPX should have approximately the same chroma bit depth when shooting in 1080p comparing to the 1080p of the EX cams shoot with mpeg 2 35 Mbit long gop? Is this somewhat correct or am I way of here? By the logic I applied to get this calculation the HPX should be better for shooting 720p or SD. Or just shooting 720p or 1080p and then converting it to SD... Yes?

If I decide to buy the EX I'll get the NanoFlash along with it. Maybe not right away but sooner or later. I just don't know if the extra price is worth it due to the questions I need answered... :)

The EX3 + NanoFlash adds up to approximately 10500€.

Oh yes I forgot - I plan on buying a DOF adapter to. Is it worth spending extra cash on a relay lens when using a camera that has a removable lens - in my case the EX3 or am I gonna get acceptable results with just using the HPX and its lens??

Please help, I just want to start creating again! All this technical reading and thinking is getting me crazy! :)

Tank you!

In my opinion there is no substitute for a larger sensor. All things being equal (or at least similar) a larger sensor gives lower noise, better sensitivity and higher dynamic range. This is why the EX cameras do so well compared to the competition. The EX1R and EX3 have full raster 1920x1080 sensors and the HDSDi output is 10 bit uncompressed.

In progressive the difference between 4:2:2 color space and 4:2:0 is extremely difficult to see. In interlace the difference is greater. The EX records marginally more pixels of color information at 1920x1080 4:2:0 compared to the HPX's 1280x720 4:2:2, but thats just numbers.

In Europe many broadcasters are looking at following the guidelines for HD set by the EBU which are for a minimum of 1/2" sensors recording at minimum 50Mb/s. This can be achieved with an EX and nanoflash but not with the HPX.

If you are worried about future proofing for HD broadcast then the EX may be the better camera. I would suggest if you shop around you should be able to get an EX1R with NanoFlash for less than 8000 euros plus VAT. Many dealers have stocks of the original EX1 for even less money.

Rather than a DoF adapter I would look at a DSLR. The new Canon 550D is only around 800 Euros, about the price of a DoF adapter.

Hi Alister! Yes, I've used a HPX - 7D combo just a week ago. I don't know... 2 cameras, to codecs, to diferent images... 7D was nice to use on the Glidecam 4000 though - wery light combo -> easy to use. I don't know - it's something to think about.

About the chroma subsampling: would you consider HPXs 1080p material broadcast quality?

I think you're on the right lines, but wrong wording. "Bit depth" means how many bits get used to code each sample - normally 8 or 10 for our purposes. To all intents and purposes, the EX is capable of resolving 1080 samples resolution vertically for both luminance and chrominance - but then throws away half the vertical chrominance resolution because of the 4:2:0 recording.

The HPX is capable of *recording* 1080 samples of both luminance and chrominance vertically - but the chips can't deliver that level. Pixel shift improves LUMINANCE resolution - but can't do the same for chrominance. Hence you can expect (and get) about 540 lines vertically for chroma - same as the EX.

[For the EX it's a bit like starting with 2 litres of water and trying to pour it into a 1 litre pot. (Half will overflow and be lost.) For the 171, you only start with 1 litre, then pour it in a 2 litre pot. You still only have 1 litre in both cases.]

That's only a part of the story. Pixel shifting won't give equivalent luminance res to 1920x1080, it works out to about 1200x650 for 960x540 chips. Hence less vertical luminance resolution for the 171.

But it's in the horizontal direction that the EX is the clear winner. The 171 and DVCProHD subsample the luminance to 1280 horizontally, and the 4:2:2 means 640 chroma samples horizontally. With the EX, it's 1920 for luminance, and 960 for chroma - and the difference really shows.
Vertically, the 171 should have a slight advantage in chroma res vertically - it will still be limited by the chips, but still at 540. The EX will be half of 720, or 360.

It's a different story horizontally. In 720 mode DVCProHD subsamples luminance to 960, and therefore the chroma is. With the EX, no subsampling, so 1280 for luminance and 640 for chrominance.
Theory aside, it's as Alister says. Most stations are keen to keep to the EBU guidelines, which basically state 1/2" or larger sensors, 1920x1080 sensors and recording, and either the XDCAM 422 50Mbs codec or AVC-Intra 100. The 171 doesn't meet any of the criteria, the EX meets all but the codec - which can be fixed with something like the nanoFlash.

Draw your own conclusions.

Ups, missed that one. Chroma res not bits!

Half? Wouldn't it be 270 vertical?

But for the down-conversion from HD to SD I just remembered: all the cameras shoot in one format and then convert the signal accordingly to the desired format selected, right? I don't know in which format EX shoots, but if it shoots in 1080p and then down-converts the signal to SD internally the chroma res of SD would be what? 720 x 576? That can't be right... I know that the HPX shoots in 1080 and then down-converts the signal to the selected format...


I see that for future HD work the HPX probably isn't gonna cut it. At least not in 1080 res... But when HD gets more firmly established, cameras like the HPX or almost all 1/3" camers that use pixel shift wont be acceptable for HD broadcast? And coming from that statement, there should be a considerable price drop in such cams, don't you think so?

You can pick up good used EX1's from dealers for less than 3000 euros. Add a NanoFlash for 2,500 euros and your all set. For SD set the EX1 to output SD over the HDSDi connector and record that using the NanoFlash set to IMX50. You will have full SD 4:2:2 colour and be recording using the very good, broadcast quality IMX50 codec.
Make a few optimizations in the picture profiles for SD (using HD detail settings doesn't work well) and you'll have a pretty respectable SD system that IMHO will be better than most small DV or DVCAM cameras. You will also have the advantage of being able to record HD in the camera and SD on the NanoFlash at the same time. When you want to move up to HD simply switch off the HDSDi downconvert and record on the NanoFlash at 50Mb/s or higher and your all set to make broadcast programmes.

A quick reality check here though: While the EX1/nanoflash will meet with the EBU regulations, TV channels or commissioners may want a full size camera with expensive lenses for some projects, don't expect to just go out and make high end drama with an EX1. While it may well produce beautiful pictures you may be expected to use a high end camera on high end productions.

The EBU guidelines are have been around for a year or so and so far have made no difference to the price of non EBU spec gear. There is, don't forget a huge market for HD in non broadcast applications.

No. For a vertical column the EX has 1080 photosites for each of Red, Green and Blue. Hence (subject to lens etc) it can resolve up to 1080 lines resolution in the vertical direction, and that's irrelevant what the colour of the image is, so it's reasonable to say vert res of 1080 (max) for luminance and chrominance.

Along comes vertical subsampling, and 4:2:0, and that means only half as many chroma samples as luminance. So still 1080 of luminance, but now down to 540 of chrominance. In practice it's not quite as simple as that, especially for interlace working.

As far as the wider question goes, then I see the lowest res chips for HD (960x540, or 1/2 megapixel) not being sold much longer as new models are introduced. At that point, it's likely that such cameras will be seen as less attractive on the second hand market, and their price drop.

Full 1920x1080 chips are increasingly seen as necessary if you're interested in going for broadcast, but I'd expect to see 1 megapixel chips around in new cheaper cameras for quite a while. They offer quite a good compromise in 1/3" cameras between resolution and sensitivity.

Some things can be predicted more accurately than others from specs, and specs don't really give much idea as to how good downconversion will be. All the HD cameras have enough basic resolution to give a good SD in theory, doing downconversion well is not as easy as it may seem. It's important to filter all the fine detail off first, otherwise it just gives aliasing in the downconvert. (Too fine detail folds back as coarse artifacts.)

If SD performance is important to you, (at least at the moment) I'd be inclined to look at the EX1R, rather than the EX1, which records it in camera. I have heard reports that the SD downconversion of the EX1 wasn't all that good - it was really intended more for monitoring than for programme use, it wasn't able to be recorded within the camera. The EX1R should be much better. Or shoot HD and downconvert to SD in software.

I've looked into this a fair bit and while the EX1 downconvert is not the best in the business it's not all bad. A big part of the problem is that the detail correction is optimised for HD with thin edges that twitter and buzz when down converted to SD. If you go in to the Picture Profiles and set the Detail Frequency all the way down to -99 the downconverted SD looks so much better. Some of this same issue is apparent on the EX1R and although this has separate detail settings for HD and SD the SD settings are not well optimised. I was asked to set up an EX1R to match some existing SD cameras for NRK (Norwegian state broadcaster) and I really had to thicken up the detail edges to get a satisfactory picture, or at least one that looked like a DSR570.

What are the differences in shooting SD or down converting in post? I don't see that as a problem - I've always been doing it this way... If shooting for chroma key for example, HD is better and then down converting it to SD. I don't know, I've never gotten myself into this topic much. But thinking about it now makes me wonder: does the camera use only the pixels it needs when shooting SD, because if it does, I don't see why one would shoot in SD and not in HD and the down converting (in a situation where this process isn't in some way a logistical problem of course). You get much more information this way... Don't you?

About DOF adaptors: does anybody have experience using a relay lens with a dof adapter? Is there really such a difference that justifies it's price (for low to mid end work)?

Hmmm, that sounds suspiciously like aliasing to me ........ And yes, turning the detail fully down should reduce the problem - but will mean that you can't do your original plan of simultaneous HD/SD recording via an external recorder. (Plus will mean going into menus everytime there's a change from HD/SD or back.)
In terms of information/quality I doubt there'd be much of a difference in theory, but it's worth bearing in mind that it can often be very difficult to do a process in real time with limited processing power (such as in a camera), relatively much easier in non-real time with the power of a desktop computer. So that's the argument for shooting HD, downconverting later.

I think you answered yourself "why, then, ever bother shooting SD if you can do HD?" - and the answer must be to save time. News is the most obvious example, and a lot of channels ae likely to remain SD for a few years to come. The last thing they want is to have to spend time on downconversion, they need the cameras to shoot SD, period. (But may well want to buy new cameras now for reasons of solid state, and with the thought that they may be used HD in the future.)
There's a whole lecture around this subject, but it boils down to aliasing. That's normally thought of in the sense of time - such as the stagecoach wagon wheels going backwards in westerns. If the 24fps "sampling" of the movement can't show what's really happening, it would be better to show a blur than backwards movement. The danger about what you suggest is that it can give the spatial equivalent to that. The answer would be to put a different soft focus filter in front of the chip for HD and SD modes - practically, that's not going to happen.

Good downconversion should filter out the fine detail from the HD image before the downconversion process. In other words, electronically soften the HD signal to no better than SD quality BEFORE the actual downconversion. From what Alister says, it doesn't sound as if the EX1 real time downconversion does that. (Not normally a problem - just downconvert in software later.)

Yes it is aliasing, I was just trying to use simple terms that everyone understands. My suggestion is to make the detail correction edges thicker (using the frequency adjustment) so they don't alias when converted to SD. Simply turning detail off is one option, but this leads to soft looking pictures, especially in SD. Optimising the detail settings for SD makes a dramatic difference to the perceived sharpness of any SD downconverts. The EX1/EX3 does do some interpolation during downconversion which reduces aliasing and produces a better downconvert than a non-antialiased conversion, but better results can be obtained with careful (slow?) downconversion in the edit. I did put up some examples of what happens if you don't anti-alias on my blog: XDCAM-USER.com Getting SD from HD and the problems of oversampling.

Mind you, Alister, do we need to worry about all the aliasing given how accepting people are of footage from the DSLR's?! ;-)

Well, very often you can get away with it, but as soon as you shoot something with a fine repeating pattern or texture it all goes rather nasty. Just point a vDSLR at a resolution chart and you get is a mess of colors and patterns that make it impossible to actually make a meaningful resolution measurement.

It's not just what the pictures look like during editing. Aliasing can survive relatively unnoticed through the production chain, only to really screw up the final transmission coding, likely as it is to be at a less than really desirable bitrate. So a production team are pretty happy with their material all the way through production, settle down to watch the material at home for transmission, and ..... oh my....!

At which point they blame somebody else. Aliasing's a bit like a virus - can pass a long way unnoticed, then pop up to cause a lot of harm.

Maybe this is why there's a charge if you want the SD option on the PMW350? "Good enough" may be acceptable on a camera the price of the EX1 in a way it isn't on a full broadcast camera.

Nice web site Alister!

I've heard something about a new EX cam coming out. Has anybody else heard something similar?