10 bit 422 HDMI with TC?

[Piotr Wozniacki]

First of all, let me sincerely and wholeheartedly congratulate Convergent Design on the new Gemini 4:4:4 recorder - a true piece of Art&Technology; pity I cannot afford it...

I only hope - and I mean it - that it doesn't hamper development of even more reliable firmware for the current nanoFlash (the way Flash XDR became sort of obsolete once the nano hit the market). Please remember there are still some promises waiting to be kept, like mixing the embedded and analogue audio. Also, the current firmware just doesn't work for cranking in the MXF format.

OK so now to my question:

With the Sony NEX-FS100 camera, the HDMI-out is supposed to output 10bit 4:2:2 with embedded TC.

Is it technically viable that the nanoFlash handle that?

Piotr

[Dan Keaton]

Dear Piotr,

The nanoFlash is an 8-Bit recorder, just as MPEG-2 is always 8-Bit.

(Thus, it can record over 16 million colors (8-bits x 8-bits x 8 bits = 16.7 million colors).
Many people equate 8-Bit with 256 colors, which is wrong, it is approxiately 256 shades per color.)

The nanoFlash can not record 10-bit.

[Piotr Wozniacki]

Thanks Dan.

Please note, however, that when I asked whether the nanoFlash could handle the NEX-FS100's HDMI output, I didn't mean the current firmware. Just the hardware potential to implement an appropriate firmware upgrade, perhaps a paid one.

Also, please note I emphasized the word "supposed" in my assumption on the camera's HDMI actually outputting 10bit with embedded TC. So - even if the 10bit part is out of question (please say whether or not it is/is not definitely) - you haven''t answered the other part of my question: TC embedded in HDMI; is an a firmware upgrade viable to handle this?

Thanks a lot

Piotr

[Dan Keaton]

Dear Piortr,

Yes, we will be providing a free firmware upgrade for the nanoFlash so that it can accept and use the timecode embedded in the HDMI output of the Sony FS-100.

The nanoFlash can not record 10-bit.

Our Gemini 4:4:4 is designed as a 8-Bit or 10-Bit, full uncompressed, HD-SDI, 4:2:2 and 4:4:4 recorder.

But, the Gemini 4:4:4 does not have HDMI in.

I am not in a position to speak with authority if the Sony FS-100 has 8-Bit or 10-Bit output over HDMI.

[Piotr Wozniacki]

Dear Dan,

Things are evolving quickly....While we will know everything for sure after NAB, it's rumored that Juan Martinez of Sony stated the HDMI on the NEX-FS100 is only 8 bit, but capable of 4:4:4!

So we already know that the nanoFlash will not be able to record 10 bits even with future firmware releases - now, how about 8 bit 4:4:4?

I for one would happily trade the highest bitrates possible now (especially anything over 100 Mbps in L-GoP) for the ability to record 8-bit 4:4:4 over HDMI from the FS100!

Yeah, I'm dreaming - but 100 Mbps MPEG-2 would be so much easier to edit than the 24 Mbps AVCHD this camera is recording at 1080/25p :)

Piotr

[Dan Keaton]

Dear Piortr,

While I personally do not thinkg 4:4:4 is possible with the nanoFlash, I will check with our engineers.

[Dan Keaton]

Dear Piotr,

I have checked with our engineers.

The nanoFlash is not capable of 8-Bit 4:4:4.

First MPEG-2 does not support 4:4:4.

Second the Sony Codec Module does not support 4:4:4.

[Rafael Amador]

Piotr,
The NANO has an MPEG_2 processor and can't record but MPEG-2.
You should know MPEG-2 doesn't supports 10b (dropped few years ago: MPEG-2 - Wikipedia, the free encyclopedia), neither any 444 scheme (RGB or YUV).

Originally Posted by Piotr Wozniacki

Dear Dan,

Things are evolving quickly....While we will know everything for sure after NAB, it's rumored that Juan Martinez of Sony stated the HDMI on the NEX-FS100 is only 8 bit, but capable of 4:4:4!

That means nothing.
Every HDMI 1.3 device is "4.4.4 capable".
Ask Juan Martinez in which conditions the NEX puts out 444 (guess you are talking about 8b RGB).
rafael
PS: Don't miss this: http://www.netmode.ntua.gr/courses/p...les_MPEG-2.pdf

[Piotr Wozniacki]

Dear Dan,

Considering all the breath-taking development in external recorders, announced just before and during NAB (like the Sound Devices PIX, or BlacMagic HyperDeck) - I guess CD needs to reconsider your somewhat "stiff" attitude towards what is possible or not with both the current (nanoFlash), and the future (Gemini) products...

What I mean is:

Originally Posted by Dan Keaton

The nanoFlash is not capable of 8-Bit 4:4:4.

First MPEG-2 does not support 4:4:4.

So what? The MPEG-2 standard has been constantly evolving, and even CD's own, current implementation on the nanoFlash is not fully compliant with the original specs. Why not extend it even further?

Originally Posted by Dan Keaton

Second the Sony Codec Module does not support 4:4:4.

Well, this one is a bit more difficult, as it deals with hardware - but I'm sure there are ways to:

- either squeeze more of the (2) current encoder chips,

- or offer a (paid, of course) hardware upgrade, so that CD's loyal customers can keep using our old good and reliable nanoFlashes, while being able to take advantage of the low-noise 10 bit signal that the S35 sensor now offers...

PS Or at the very least, please consider properly dithering the 10bit input into the 8 bit encoded output, so that the color banding is eliminated! This could be implemented along with TC over HDMI support upgrade of the firmware.

Just my thoughts; Regards

Piotr

[Rafael Amador]

Dear Piotr,
Sorry but I can not understand your repetitive questions about 10b or 444 recording on a NANO.

What you want could be possible. You just need to modify the MPEG-2 standard (private property) and build a new processor supporting the new implementations. That's all.
With the present NANO, is physically impossible.
CD could rewrite the codec, but the NANO would need a new processor to run it.

What I don't understand is why you try to achieve a format that was dismissed years ago for inefficient.
And of course i understand that CD don't want even to discuss this because they know this path leads nowhere.

Piotr, if you are a developer and you say that in 2011 you are investing money in MPEG-2 technology, your fellows manufacturers will thing that you went nuts.
rafael
PS: MPEG-2 have a limited life. After XDCAM will be no more.

[Piotr Wozniacki]

Dear Dan,

I understand that not answering to my considerations above has been due to your being busy at NAB; now that it's over, I'd hope you will address them.

Here is some more ideas; I'm sure you understand perfectly that we all want to preserve the value of our nanoFlashes. Let's do some maths on the bandwidth required for uncompressed 422, using the example of my format of choice - 1080/25p, in both 8 bits (as from the new FS100, or AF100) and 10 bits (as from the current EX1 and many other cameras, including the F3 in basic configuration):

8 bit @ 1920 x 1080 @ 25fps = 99 MB per/sec
10 bit @ 1920 x 1080 @ 25fps = 132 MB per/sec

I've been thinking:

- while most of the time some mild compression is our friend (like the 220-280 I-Fo the nanoFlash encoders deliver), sometimes - with the highest quality in mind - we would like to capture the highest quality possible, and only decide at post what to do with it. This very approach is now possible using the new recording devices, including the <$400 BlacMagic HyperDeck; why not the nanoFlash?

Well, what if - as a menu option - CD allows to by-pass the Sony encoder chip(s), and record the uncompressed signal straight to the 2 fast enough CF cards in RAID 0 mode?

If my math above is not wrong, two striped CF cards would be plenty enough fast for that! Of course, there is also the space consideration - but with the upcoming 128GB cards, 2 of them would give us quite a long recording time (with even greater capacities to come in the future).

This could be the right answer from CD to the current market challenges! Please let us know what you think;

Piotr

[Andy Mangrum]

Hello Piotr,

There is a lot to address in the last several post, I will start, and I am sure when Dan has time he will elaborate much further.

The nanoFlash Architecture is built around the Sony Module, which enables us to do the various formats, and Codec's, but it is only a 8 bit device, Building an Architecture which would have two sony modules would be much more complicated and most likely more expensive then Gemini, due to the Software behind it, and would require a completely new Architecture design.

As for the bit rate you have listed that looks fairly accurate, But there are issues involved with using Compact Flash Cards, Compared, Simply they are to slow, even in a raid Configuration.

So below is the needed bit rate for uncompressed.

"8 bit @ 1920 x 1080 @ 25fps = 99 MB per/sec
10 bit @ 1920 x 1080 @ 25fps = 132 MB per/sec"

With our internal Testing we have found that the fastest Card's actually only perform, at best around 48 MBps read, and the write is always slower. also, (Keep in mind to record 280Mb+8 Channels of Audio is around 30MB's (Factor of 8 + Overhead requires 300 Mbps Write) so even in a raid 0 Configuration, It is not exactly a 2x Read / Write,since there is usually 10% loss per drive. that is only around 90 MBps so you would not be able to achieve 132 MBps before any overhead is considered.

Also note you can test your own CF cards with a Hard Drive speed test, but keep in mind you may be limited by your Median, I would suggest a Firewire 800 (80MB(s)) or Express Card reader, so you know that is not the bottleneck. as well as a copy test will always verify the results,Since to maintain the data rate it would have to but a sustained write. but the Hard drive being copied to must be faster as well, with out test I copied to a WD Black rated around 110-120MB(s) as well as a raid 0 rated around 400MB(s).

Granted this is a very great concept, It does not seem as practical, thus Convergent Design engineered the Gemini to meet the needs, while keeping everything affordable,

As Dan has pointed out 2 San Disk Extreme Pro 128GB cards list for $1,500x2 is $3000 for 256GB of record time which is about 20-30 Min's (10 bit to 8 Bit)

Where SSD's are much more affordable, and are capable of much larger capacities and data rates. let alone transfer rates since SSD's can be Sata 6GB, so your slowest point will either be the drive itself or the drive being copied to. but the potential bandwidth is around 580MB(s), compared to a Compact flash Card current best at 48 MB(s).

Hope this Helps and I am sure Dan will fill in the blanks in due time.

Best Regards

[Piotr Wozniacki]

Thanks Andy for your realistic assessment of my concept - even though it's a pity it can't be done, I'm not going to blame the messenger :)

Piotr

[Dan Keaton]

Dear Piotr,

I am sorry that I have not been able to reply earlier.

Yes, I have been at NAB this week and we were very busy.

Yesterday we traveled all day, getting back home about 12:30 am.

Today I tried to catch up on my sleep.

We offer two solutions.

The nanoFlash, a very successful compressed recorder. It is 8-bit and it conforms to the MPEG-2 standard. With two 128 GB CompactFlash cards, it can record broadcast quality over 10 hours uninterrupted. And it can record up to 280 Mbps with the proper cards.

The nanoFlash is a very viable solution for many, and remains a fully supported product.

It is especially nice for use with the very low noise Sony F3 camera as well as many others.

Of course, if one has a Sony F3 and will be adding the first firmware upgrade to obtain Dual-Link HD-SDI, 3G and S-Log, our new product, the Gemini 4:4:4 is an even better solution.

I am sorry, but we are not in a position to re-engineer the Sony Module, nor modify the MPEG-2 standard to support 10-Bit.

Recording to full uncompressed is quite a technical challenge. While it is technically possible to do this in the nanoFlash, it would require an extensive engineering effort. We have already seriously studied this concept.

A far better solution, in our opinion, would be to use the far more capable Gemini 4:4;4, which has features that we would never be able to retrofit into existing nanoFlashes.

[Michael Carmine]

Dan,

I am #16 to receive a Gemini in late July, will there be a Prores 422 HQ format. And if not, what will be the workflow to send the footage back to Burbank. They will deal with .mov files but we would love not to transcode.

Thanks for looking into this.