Af 105a announced in japan

Ok so if we can agree that a 10 bit or higher signal is being seen by the sensor and that via the HD-SDI a 10 bit 422 signal can be recorded then suly the knowledge of double the colour resolution (4:2:0 to 4:2:2) and a quadruple in colour information (250 odd to 1000 odd shades of colour) should get few people to hope for an improvement in image quality, height light roll off and grading potential?

this really isn't my strong suit so I say the above as more to learn what's right not to spread the gospel of the AF100A followers :0P
 
XSVpro said:
So your saying the sensor is only capable of "seeing" a 8 bit 422 image and will only pass a 10 bit 422 wrapped version via the HD-SDI? I'm sure that's wrong, the sensor is video optimised gh1 sensor and the gh1 was capable of 12 bit Raw photographs.
Click to expand...
According to DxOMark, the GH1 has a 12-bit image sensor with a RAW dynamic range of 11-stops at ISO 200. The AF100 uses a sensor based on similar technology:

http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Panasonic/Lumix-DMC-GH1

Where 10-bit and higher precision makes a big difference is in capturing fine-grained shadow detail and gradients. The sensor's output is resolved on a linear scale, and at very low illumination levels, an 8-bit scale uses only 2-bits to encode near-black colors. With a 10-bit range, those details would be encoded much more finely at 4-bit precision. There is no reason to doubt the AF100A's internal image processing circuits will be capable of maintaining 10-bit output precision.
 
  • There should definitely be a difference with 10 bit 4:2:2 that would benefit potential buyers
albeit too little too late :(
 
XSVpro said:
So surely at the sensor level a lot more information is seen than just 8 bit 422 and what we have on the current AF is a downscaling of 12 bit 422 to 8 bit 420 and now 10 bit 422 via the HD-SDI?
Click to expand...
It's normal for the signal straight from the sensor to be digitised to a very high precision - yes, 12 bit or more. But this is the raw data, pre any sort of processing, so any sorts of comparisons with the bit depth of the final signal are meaningless.

The processing will involve gamma correction, gain setting, colour balancing etc etc which may be likened to multiplying the values of the original. It's also possible that the actual processing may be done at an even greater bit depth (say 14 bit) to allow for rounding during the maths. But the point is that for the output signal, the least significant binary bits will not consist of anything significant - either noise or maybe just zeroes. As such they can be lost without losing anything important.
.......I would argue this is false advertising for Panasonic to claim 10 bit 422 if its just a wrapped 8 bit 422 and I can't see Panasonic doing that.
Click to expand...
No, not false advertising. The question is not whether or not it's GENUINE (I'm sure it is), the question is whether or not it's WORTHWHILE.

You will get 14 bit binary numbers out of the DSP, but the least significant bits won't have any real significance in image terms and it's sensible to round them up. Think of the whole thing like doing a set of maths to work out some sum such as how many grams of flour are needed in a recipe, if you alter the quantities of other ingredients. Put the numbers in and the calculator may come out with some figure like 63.43867 grams. In that context, the numbers after the decimal point are effectively meaningless and "63 grams" is a far more sensible answer, probably even just 60 grams would be good enough. You'd never be able to measure those last 0.00007 grams, and they'd be meaningless because an approximate assumption may have been made about the weight of an egg anyway! The video equivalent is random noise.

In the case we're talking about, the DSP will give out it's values in binary form, the question is how many of the most significant digits to keep, how many of the least significant to discard. Engineering wise, the answer should be related to the noise level of the device - throw away those that are just noise, keep the ones which reflect real signal.

Bitdepth importance varies dependent on what type of video signal we are dealing with. For one intended as ready to be displayed, 8 bit should be more than sufficient - it can specify more shades of grey than the human eye can distinguish. The real advantage of 10 bit is when significant post work is to be done on the signal - but it really needs to be done in conjunction with something like S-log, and has little point unless the input is very noise free. Finally, if you record the RAW data off the sensor, "RAW" means that no gain, gamma etc processing has yet been done and you need far higher than 10 bit depth for that.

Sorry - it's complicated, technical, but I'm afraid that's the way it is. If such bores you, then it can be summed up as 10 bit working will only bring significant real benefits when used with a high quality front end camera.
 
lpowell said:
The sensor's output is resolved on a linear scale, and at very low illumination levels, an 8-bit scale uses only 2-bits to encode near-black colors. With a 10-bit range, those details would be encoded much more finely at 4-bit precision. There is no reason to doubt the AF100A's internal image processing circuits will be capable of maintaining 10-bit output precision.
Click to expand...
But that takes no account of the DSP and the image processing it does, gamma correction maybe being the most relevant here. That will typically add a lot of gain to the low level parts (the blacks) and likely reduce the gain in the highlights.

In such cases, gain means multiplying the digitised signal by a certain amount - let's say the figure is 8x (so 1000 in binary). In this case, if a low level 12 bit sample was (say) 000000011011, multiplying by 8 would give binary 000011011000. It follows that if the input increased by a single bit, the output would increase by 1000 (8 in decimal), so in this range the least three significant bits will always be zero.

By such logic, then whilst a single least significant bit of the input to the DSP may mean a lot, the same is not true of the output. Hence why the bitdepth of the processed signal is truncated.

Maybe it's time to write a book on the subject. I shall call it "8 bit video: 256 shades of greyscale". With the right cover and placing in the bookshop I anticipate huge sales - and a lot of disappointed readers when they realise it's not a sequel to a trilogy which sounds a bit similar. :happy:
 
Last edited:
Do it!

ill wait for some footage to come out, maybe bloom will get a few raw files to share to grade? I'm hoping its a least worth buying, if not then ill just stay where aim and re look at the BMC for 10 bit capture
 
harddrive said:
Sorry - it's complicated, technical, but I'm afraid that's the way it is. If such bores you, then it can be summed up as 10 bit working will only bring significant real benefits when used with a high quality front end camera.
Click to expand...
No, it's boring because you've wandered far off-topic into a monologue that has nothing to do with the upcoming AF100A.
 
harddrive said:
... But the point is that for the output signal, the least significant binary bits will not consist of anything significant - either noise or maybe just zeroes. As such they can be lost without losing anything important.

...
Click to expand...

This where I believe your logic breaks down. "Noise" is siginficant.

In a lot of digital devices, noise is intentionally added to the signal for smoothing purposes. So noise can be a real benefit.

Look at some of the threads about posterization by the Nikon D800. Shooting at ISO's below 400 apparently creates too clean a signal for the 8-bit codec, which is causing posterizaton. But keeping the ISO at 400 or more adds enough noise to the signal to eliminate or signicantly reduce posterization.

So I really find it difficult to believe that a low noise 8-bit image is going to look as good as a dithered 10-bit image.

BTW, I think the better argument against 10-bits is that those bits have come from somewhere. So unless you increase the codec's bandwidth, you are increasing the levels of shading at the expense of high frequency detail. Is it a smart trade-off? Honestly, I would probably go to 10-bit only if the codec were around 100Mbps or higher for I frame footage.
 
Last edited:
This is what s making the 10 bit output sound worthwhile. Before it was just 8 bit but at 422 instead f 420, despite that being twice the colour resolution the main benefit was the added I frame codec and the 24psf mode but now there should be a distinct advantage in that you are getting tis but also shooting at 10 bit colour space.

Any news anywhere of the camera release or better yet some footage?
 
Have we got any confirmation that the AF100A would be marketed here at all? I keep thinking about that sticky about "3 things." If I had a buck to bet, that I didn't already spend on gear, I'd bet that Panasonic does not market the AF100A here but introduces other features on a new AF100 that has AVC-Intra or Ultra and uses the Micro P2 card. They could include other refinements to make that bandwidth work, OLPF, etc. If they can keep it near the same price range I think it would sell better than an AF100A. Just a guess... or wishful thinking.
 
I wish Jan could just chime in and give us some indication. She's like the AF100 agony aunt anyway ;0)
 
I haven't seen anything other than the one at the top of this thread, which I've just re-read. Is "enlarged focus assist" 1:1 pixel zoom?
 
J Michael said:
Here is a translation from a product info page http://translate.google.com/translate?sl=ja&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&layout=2&eotf=1&u=http%3A%2F%2Fpanasonic.biz%2Fsav%2Favccam%2Fag-af105a%2Findex.html
Click to expand...

That page was mentioned earlier in this thread. And if my memory serves me, it looks exactly like the product page for the AF100, with very minor changes, mostly changing the name to AF105A. The new features, expanded focus and 10-bit 4:2:2 output, are buried near the bottom of the page.

I found the camera listed at a store in Singapore for 4,499 USD, but I'm not sure how reputable they are: http://www.expandore.biz/product_detail.asp?productid=4324

The lack of news about this camera is disturbing. Are they really planning to launch it this month?
 
Sumfun said:
Are they really planning to launch it this month?
Click to expand...
In Japan, perhaps, but I doubt it'll be launched here in the USA that quickly. I mean, remember, Panasonic's headquarters in the USA was pretty much right in the middle of "superstorm" Sandy. I would imagine that things have slowed down there a bit.

I have no doubt that when the USA version launches, there will be adequate info released.
 
Also remember that when the original AF100 was released, several people were allowed early access to the camera. As soon as that happened, there was a mad clamor in the forum for these people to put some preliminary footage online. As an accommodation footage was uploaded which resulted in a feeding frenzy that consisted primarily of self-appointed experts picking the images apart like vultures.

Look at this thread as an example, there are people on here already criticizing a new version that no one has even seen yet.

If I were Panasonic, I wouldn't be in a hurry to again release advance information or footage. I'd wait until I was ready to deliver the camera into the waiting hands of professionals who know how to use it to full benefit.
 
You must log in or register to reply here.
Back
Top