Af 105a announced in japan

harddrive said:
Jan - I agree with your main premise (that you have to take price point into account) but on that basis a main comparison for the AF100 should be based on the Sony FS100, would you not agree? In the UK, they seem to be more or less the same street price, at least within about 10% or less.

You may quite truthfully say that the FS100 lacks the ND filters and SDI output of the AF100. But technically, the FS100 has a far better performance in terms of resolution, dynamic range, low light performance, noise level etc - and is super35 in addition. I'm sure you'll disagree, but I know many people who will say that *OVERALL* it offers more than the AF100 at the price point.
Click to expand...

This comparison has already been done, and the AF100 seems to be pretty feature rich compared to the FS100:

http://www.dvxuser.com/V6/showthread.php?260461-FS100-and-AF100-compared

I've cut a fair amount of AF100 and FS100 footage, and I have to say... you have to be a pretty die hard pixel peeper to tell the difference on real world images. I keep hearing about the IQ advantage that the FS100 has over the AF100, I just don't see it. We can debate if cameras costing twice as much or are yet to be released are somehow better, but I just find it perplexing that the AF100 has gotten such a bad rap, especially when it's capable of making such nice images.

Shawn
 
Editman Sweden said:
For the record it was Panasonic that stated 800 lines of resolution and I would consider the AF100 to be a 720p camera when it comes to resolved detail or a little bit below 720. I see about the same amount of detail from the 5D2 and 5D3 as from the AF100. The F3 resolves much more detail, how much more I don't know but it's about as much as you can get out of a 1080p camera I guess.
Click to expand...

Based on the price difference, the F3 should resolve 3-1/2 times as much (2520 lines). Good luck with that.
 
harddrive said:
My posts were in specific response to Editman Sweden giving the res figure for the AF100 to be 800 lines, and a later claim of 750 lines. Adam Wilts chart clearly shows it is about 650-680, no more.
Click to expand...
Not quite, and I think you're crossing over the line that separates technical measurements from subjective hyperbole. If you examine Wilts' charts closely, do you see the resolution drop off abruptly at 680 lines, with "no more" details present beyond that? No, what I see is that the genuine details beyond 680 lines begin to show artificial "enhancement" from spurious aliasing details. It is only at much finer resolution that the aliases become strong enough to completely dominate. The saving grace, as you point out, is that the aliases are restricted to the luma channel and are free of unsightly chroma moire.

What this shows is that the AF100's maximum resolution is compromised, but not ruined, by its aliasing artifacts. When you examine the test charts, the only reason you can reliably identify the point where aliasing starts to contaminate the genuine details is because in this case, you know in advance what the test chart is supposed to look like. If it was a natural image with irregular details, you'd have no way to distinguish genuine from aliased details, and your evaluation would necessarily hinge on how subjectively "sharp" you perceived the image to be. (Assuming, of course, that telltale moire patterns did not emerge with object movement.)

Ideally, we'd all prefer to shoot with a camera that delivers true 1080p resolution with no significant contamination from aliasing artifacts. But at this point in digital camera evolution, that level of performance in a videocam with professional features costs far more than the AF100. As it stands, the only thing that could reduce aliasing on the AF100 would be a stronger OLPF filter that would blur its genuine details along with its aliases. Panasonic's solution to this dilemma was a technical compromise that delivers a subjectively sharp image free of discolored moire, at a very affordable price. The ultimate question for us is not how many lines of resolution the AF100 rates, but whether we can work with that compromise to produce results that please both ourselves and our clients. If it falls short of your requirements, you'll have to consider how much you're willing to spend on a different set of compromises.
 
To me the only thing that would keep the AF105 competitive in this market would be if it were to have internal 10bit recording on a dedicated capture codec - not on a display codec as is with the present camera config.

I would also change it's name for PR reasons alone. As has been said for valid and not so valid reasons the "AF" has gotten a terrible rep.
 
lpowell said:
If you examine Wilts' charts closely, do you see the resolution drop off abruptly at 680 lines, with "no more" details present beyond that? No, what I see is that the genuine details beyond 680 lines begin to show artificial "enhancement" from spurious aliasing details.
Click to expand...
It's important to think what aliasing actually is. It's a "bending round" of the frequency around a critical limit. If we gradully make the lines closer and closer together, (and assume the critical limit is 680) then the image will match the input - up to a point. Look at 400 lines, and you see 400 on the image. Look at 500 and you see 500. Same with 600. But by 700 something strange has happened - you now see only 660! Make the lines finer still (say 800) and the image now consists of only 560!! When you reach the resolution limit of the system, the ideal response should be to just become grey. Which is what happens with film.

But in a system prone to aliasing, then above a critical number (it's called the Nyquist limit) fine detail gets turned into coarse detail. In a still image, it may not be too serious, in video there are complications.

The point is that above the Nyquist limit there is no such thing as "genuine" detail. You're getting a false pattern.

You may put lines closer than 680lpph into the system - you'll never get line spacing finer than 680 out. Look at lines at about 1000lpph and you'll see a coarse image pattern of just 360 lpph if the OLPF is weak - which is exactly what we do see on Adams chart. ( http://provideocoalition.com/index.php/awilt/story/ag-af100_and_pmw-f3_on_the_charts/P1/ ) Look between the 1000 and 1250 lpph rings. You should see very fine lines, but you actually see very coarse lines.
What this shows is that the AF100's maximum resolution is compromised, but not ruined, by its aliasing artifacts. When you examine the test charts, the only reason you can reliably identify the point where aliasing starts to contaminate the genuine details is because in this case, you know in advance what the test chart is supposed to look like. If it was a natural image with irregular details, you'd have no way to distinguish genuine from aliased details, and your evaluation would necessarily hinge on how subjectively "sharp" you perceived the image to be. (Assuming, of course, that telltale moire patterns did not emerge with object movement.)
Click to expand...
No. The maximum resolution is DEFINED by the system Nyquist limit. Ideally, any input above it should give a grey output. If you do see ANYTHING, it's not true resolution - it's aliasing. You make an interesting point with ........a natural image with irregular details, you'd have no way to distinguish genuine from aliased details. In practice you'd find the image looked optimal if detail above Nyquist just gave grey. If it gave an alias, it would lead to a degradation.

That is why, when you know what the critical resolution limit is, it's so wrong to treat the aliases as part of the response. The image would be better without them!

That's for stills. For video it gets worse. We've seen how detail above the real resolution limit gives the wrong number of lines. Move the camera and the lines go the wrong way! Imagine tilting the camera used for Adam Wilts test up and down. Obviously, the pattern will move in a predictable way - all except the aliases, which will appear to ripple the wrong way.

And on real life pictures that can not only degrade the picture in it's own right, but lead to inefficiencies in coders. And that's why broadcasters in particular get so bothered by aliasing.
Ideally, we'd all prefer to shoot with a camera that delivers true 1080p resolution with no significant contamination from aliasing artifacts. But at this point in digital camera evolution, that level of performance in a videocam with professional features costs far more than the AF100. As it stands, the only thing that could reduce aliasing on the AF100 would be a stronger OLPF filter that would blur its genuine details along with its aliases. Panasonic's solution to this dilemma was a technical compromise that delivers a subjectively sharp image free of discolored moire, at a very affordable price. The ultimate question for us is not how many lines of resolution the AF100 rates, but whether we can work with that compromise to produce results that please both ourselves and our clients. If it falls short of your requirements, you'll have to consider how much you're willing to spend on a different set of compromises.
Click to expand...
Well, the unfortunate fact facing Panasonic is that many of their competitors now have optimal "designed for video" chips which do give close to true 1080p with far, far less aliasing. Which is the whole point of why many of us find this upgrade such a disappointment. We'd like to see such chips in Panasonic cameras as well.

In forums, then so often everything gets presented as totally good, or totally bad. In practice, the real world is not like that. And let's try to separate the technicalities from the actuality. Way back (post no 100) I said This does not mean it's a bad camera - but it does mean that reports of it being soft relative to other cameras are correct - not FUD. Even the 680lpph figure is well below what the 720p system is capable of, so those who want true 1080p resolution in an updated camera have a valid point. If you want 720p, none of this is that relevant.

On the good side, it's clear from the chart that the aliasing is not coloured, as is the case for earlier vDSLRs. On the bad side, there is a considerable amount of aliasing above the 680lpph mark.

In other words, it's not a bad camera ( especially when considered a 720p camera) - but it could be improved.

But technically, it is completely wrong to claim anything more than around 680lpph as resolution for this camera.
 
harddrive said:
But in a system prone to aliasing, then above a critical number (it's called the Nyquist limit) fine detail gets turned into coarse detail....

The point is that above the Nyquist limit there is no such thing as "genuine" detail. You're getting a false pattern... When you reach the resolution limit of the system, the ideal response should be to just become grey. Which is what happens with film.
Click to expand...
Your detailed analysis would be correct only for a brick wall OLPF that passed 100% of all spatial frequencies below the Nyquist limit and blocked 100% of all spatial frequencies above that limit. In reality, the OLPF is a physical device with a mild slope that allows the aliasing artifacts to increase gradually beyond the point where they first start to visibly affect the sensor resolution. Contrary to what you claim, the region past that point is not an entirely "false pattern", it is a mixture of genuine image detail combined with aliased artifacts that gradually increase at higher resolutions. Your "Nyquist limit" analysis is purely theoretical and fails to accurately describe what the real-life camera produces.

The reason that a film resolution chart becomes gradually gray as details become finer is because film is not digitally sampled. To reproduce that ideal in a digital motion picture camera would require an ultra-high-resolution sensor whose photocells were read out at an extremely fast rate without line-skipping or pixel-binning. Until sensor technology evolves to that point, we will have to make do with technical compromises between maximum sensor resolution and freedom from aliasing artifacts. Each camera design must balance the trade-offs of this complex engineering challenge with the manufacturing costs and it is up to us to evaluate the perceptual results based on our own esthetic tastes.

But technically, it is completely wrong to claim anything more than around 680lpph as resolution for this camera.
Click to expand...
That is what I mean by crossing the line into hyperbole. The number you obtain depends on how you choose to define the term "resolution".
 
Last edited:
lpowell said:
Contrary to what you claim, the region above that point is not an entirely "false pattern", it is a mixture of genuine image detail combined with aliased artifacts that gradually increase at higher resolutions. Your "Nyquist limit" analysis is purely theoretical and fails to accurately describe what the real-life camera produces.
Click to expand...
Look at it this way. If the Nyquist limit is 680, then if you steadily increase the fineness of the detail, this is what you can measure:
Chart lines ...... No of lines measured on image
670...................... 670
675...................... 675
680...................... 680
685...................... 675
690...................... 670
695...................... 665
etc

So whilst you are correct that you can't construct an OLPF with a "brick wall" response (no matter how desirable it would be in theory :) ), the way in which the sensor itself responds does change abruptly. Up until the Nyquist limit, the output accurately reflects the input - immediately beyond it , it doesn't. And this is exactly what happens in a camera. It's measurable, and it's exactly what Adams charts show graphically. You simply don't get a situation where above the Nyquist limit you get a mixture of "genuine image detail combined with aliased artifacts". Whatever you point the camera at - 685, 690, 1000, 1100, whatever - you will NEVER see a higher spatial frequency than 680 on the output. Point it at 700, and you wan't see "genuine 700lpph image detail combined with aliased artifacts ...." You'll see a false image of 660lpph.

Practically, because the OLPF does indeed not have a brick wall response, a design choice has to be made between allowing some aliasing and excessive softening of the image. That doesn't affect the form of what happens when any detail higher than Nyquist gets through to the sensor - just the amount.
The reason that a film resolution chart becomes gradually gray as details become finer is because film is not digitally sampled. To reproduce that ideal in a digital motion picture camera would require a high-resolution sensor whose photocells were read out at an extremely fast rate with no line-skipping or pixel-binning. Until sensor technology evolves to that point, we will have to make do with technical compromises between maximum sensor resolution and freedom from aliasing artifacts. Each camera design must negotiate the trade-offs of this complex engineering challenge and it is up to us to evaluate the perceptual results based on our own esthetic tastes.
Click to expand...
I don't disagree with that at all. The AF100 was a quite clever way at the time of using an existing still camera sensor beyond what it was designed for, with cost and development time savings. But it can be bettered. And the easy way to do it is to make the dimensions twice the video system dimensions.
That is what I mean by crossing the line into hyperbole.
Click to expand...
Look at the audio equivalent. Assume digital sampling of analogue audio at 20kHz - 20,000 samples per second. You need at least two samples per audio cycle to be able to ressurrect the sound, yes? So the highest frequency that can be accurately coded is 10kHz (in theory). What happens when you input a frequency of more than 10KHz? It gets folded back, so that for a 15kHz input, the coded and replayed sound would be 5kHz. Input 19kHz, and the recorded/replayed sound will be 1kHz.

And around the 10kHz mark, then as you go even slightly above, the coded signal will go slightly below. That's aliasing starting abruptly. Input 10.001kHz, and it codes to 9.999kHz. 10.1kHz codes to 9.9kHz - NOT "10.1kHz with aliased artifacts"

You can never get a frequency higher than 10kHz out of the system - no matter what you put in.

So would you agree that you could never claim a frequency response any higher than 10kHz for this system in the limit? Even if it may show false response to frequencies higher than 10kHz?

Substitute spatial factors for audio waveforms and sampling, and that is why it is not hyperbole to say it's completely wrong to claim anything more than around 680lpph as resolution for this camera.
 
Last edited:
As others have said, I would probably pay for an upgrade to my existing hardware, but probably not purchase the new camera.

When will we know the details for the U.S. program?
 
harddrive said:
Look at the audio equivalent. Assume digital sampling of analogue audio at 20kHz - 20,000 samples per second. You need at least two samples per audio cycle to be able to ressurrect the sound, yes? So the highest frequency that can be accurately coded is 10kHz (in theory). What happens when you input a frequency of more than 10KHz? It gets folded back, so that for a 15kHz input, the coded and replayed sound would be 5kHz. Input 19kHz, and the recorded/replayed sound will be 1kHz.

You can never get a frequency higher than 10kHz out of the system - no matter what you put in.
Click to expand...
Sorry, but your audio analogy only demonstrates the faulty assumptions in your misinformed video claims. That analogy does not support your claim that the AF100 is constrained by the Nyquist limit to no more than 680 lines of resolution. That would be the case only if the sensor's native resolution was no more than 1360 photocells. In reality, the AF100 has a 12Mpixel sensor with more than twice that photocell resolution. This puts the AF100's Nyquist limit well beyond its 1080p video resolution.

Your description of how ultra-fine spatial frequencies are folded down into the visible range to produce aliasing artifacts is correct. However, in the AF100 this process does not begin at 680 lines, but at a spatial frequency beyond the camera's 1080p video resolution. The reason the artifacts are visible below 1080p is not because the sensor's Nyquist limit is too low, it's because the OLPF has too mild a slope to completely block the ultra-fine details that become aliases at 680 lines and above. An OLPF that was capable of blocking all visible aliases would degrade the camera's fine-detail sharpness as well.

In short, your claim that the AF100's theoretical resolution is limted to 680 lines is incorrect. Its Nyquist limit is significantly higher than that, and the aliases you see starting at 680 lines are the result of the limited effectiveness of its OLPF.
 
BTW, people keep asking when the AF105A will be sold in the states. It won't.

What will be sold in the U.S. is the AF100A.

AF105 is the model number -- including region code -- for those distributed in Japan, just as the AF102 is distributed in Australia (or is it just New Zealand?), the AF101 in the EU, and others. If it's the same as previous cameras, the AF105 has its menu in both Japanese and English and the kit is furnished with a power supply plug suitable for that region. The AF102 (for Australia) has the menu upside down (JOKE).

This is the same coding Panasonic has used for previous cameras such as the HVX200, etc.

So, unless Panasonic changes how they do things, the new version will be distributed by region model number, followed by the "A" designation, just as the DVX100A and the HVX200A were.

FWIW
 
Last edited:
Name me one camera that has all the features the AF100 and now 100a that lists under $5000?

Large chip
ND
XLR
HDMI
SDI now with 422 10 bit
1080/60
waveform, vector scope
peaking
And can easily take any lens you throw at it.

Even so. It still won't sell. Unless what the product we care about improves, that being the image.

Unless one could get a paid upgrade. I'll send it in and you replace the hardware for say.....$1200???? Naaahh.. Make it $850
 
That's right, rational people buy BENEFITS not features. Features are only valuable if their implementation creates benefits for the user. The problem with the AF100 is that its central features don't create the benefits in image quality or work flow that people are willing to pay for. m43rds has a lot of promise despite its non-standard sensor size because it has so many outstanding small light-weight high-quality lenses. But to realize that promise, Panasonic needs to deliver a video camera with superior image quality. More than 680 lines of measured resolution. More dynamic range. Internal recording with more than 8-bit color depth. More than 28Mbps. More than 4:2:0. The AF100a specs are so humdrum you can get better image quality out of the Nokia PureView camera phone!

Here are some things the AF100A could have offered that would have made the AF101a a huge hit. More spatial resolution - at least 3k measured lines of resolution in a 4k 24p 8-bit 4:2:0 format. More temporal resolution - 1080p120 8-bit 4:2:0. More color information - internal recording with 1080p24 10-bit 4:4:4 in a CineformRAW codec or AVC Ultra. Hell even internal recording with a full 1080 measured lines of resolution in 50+Mbps 4:2:2 MPEG2 would have made many of us take notice. But the same old tired AVCHD 2.0 that we can get anywhere else cheaper? No sale.

The world is moving fast, Panasonic, but you are just watching it happen. A Red One MX with 4.5kp30 RAW internal recording for $4k. A tiny GoPro Hero3 with 4kp15 for $400! A BMCC with a large sensor and CiinemaDNG RAW for $3k. Are you reading this, Panasonic? You have the ingredients. Good sensors. Great codecs. The best contrast-detect autofocus. Reasonable menus and ergonomics. If you can put those things together in the right combination, you can beat them all. But if you are trying to protect your existing markets by stifling your own innovation, you will loose. In today's world, that's just a license for your competitors to cannibalize your business.

dustylense said:
Name me one camera that has all the features the AF100 and now 100a that lists under $5000?

Large chip
ND
XLR
HDMI
SDI now with 422 10 bit
1080/60
waveform, vector scope
peaking
And can easily take any lens you throw at it.

Even so. It still won't sell. Unless what the product we care about improves, that being the image.

Unless one could get a paid upgrade. I'll send it in and you replace the hardware for say.....$1200???? Naaahh.. Make it $850
Click to expand...
 
dcloud said:
Why the original didnthave the focus assist magnify eludes me. It wouldve been awesome.
Click to expand...

Didn't the same happen with the AC130 and AC160? Is it a coincidence? Maybe. But I see a pattern here. If they give you all in the beginning, how else will they get you to buy that A "upgrade" 2 years later?
 
noirist said:
That's right, rational people buy BENEFITS not features. Features are only valuable if their implementation creates benefits for the user. The problem with the AF100 is that its central features don't create the benefits in image quality or work flow that people are willing to pay for. m43rds has a lot of promise despite its non-standard sensor size because it has so many outstanding small light-weight high-quality lenses.
Click to expand...

I agree with that, and would add that there's an "irrational" rational element too, which is the feeling the images coming from the camera invoke.

The 5D2 sold and sold and sold because of the wow factor of the footage. Us rational types know what a bastard that camera is to shoot with, how many compromises it has on its image quality from line skipping etc.. But at first glance, the footage is awesome. I put that down to two things: large sensor, and Canon colour science.

For me, the ultimate failing of the AF100 was that the footage just didn't have awesomeness out of the camera. I'm quite prepared to believe it is capable of awesome results, but it never seemed to generate them naturally in my hands. I'd light a scene the way I like, that I thought looked beautiful. I'd get back footage I hated. Banding, lack of detail, mush, muddy mid tones if I protected my highlights, clipped highlights if I went for full mid tones, thin colour if I turned saturation down, ghastly stuff happening to roll-off on practical sources in frame if I didn't. I never did get a default setting i liked despite months of playing around, and worst of all I was never happy with the skin tones. And it wasn't much fun to grade either.

I still have my AF100, but in all honesty I prefer the image characteristics from my HVX200 and 7D/5D2.

It is a pleasure to shoot with the AF100 from a technical standpoint (except for focussing!) but not a pleasure to look at the footage and be disappointed. (* I know an external recorder might have improved things, but I must admit I hated the look of Raid: Redemption too, whilst loving the action. I thought it had ghastly magenta colour casts and horrible banding).

I know I've banged on about loving the Scarlet I moved up to, and I know it is a hell of a lot more expensive. I just wanted to add parenthetically that all these issues were gone the moment we started shooting with it. Now I can light the way I like, and get footage I'm gobsmacked by from the camera.

So in the end the AF100, despite all its rational advantages, failed to deliver for me on an emotional level, the level of loving the look and feel of the footage.

Cheers, Hywel.
 
Last edited:
You must log in or register to reply here.
Back
Top