[konquerian]

yes.

and it is for the first time i noticed a message needs to be 5 characters.. -

Thanks so much grispie! That was definitely fast and painless.

[Mutters]

Thanks so much grispie! That was definitely fast and painless.

Plus, you can also adjust shutter speed and iso, dependent on shooting mode.

AB

[kenyee]

AVCHD 2.0 expands to include 1080 50p and 60p: Digital Photography Review
AVCHD INFORMATION WEB SITE

If they have that much bandwidth, they really should consider a 1536p mode so you can use Warp Stabilizer in post to do software SR to get back to 1080p...

[kadajawi]

If they have that much bandwidth, they really should consider a 1536p mode so you can use Warp Stabilizer in post to do software SR to get back to 1080p...

Do they? Sony is doing 25p at 24 Mbps... aaaaand 50p at 28 Mbps. And I thought 25p needed more than 24 Mbps... Increasing the bitrate by 4 Mbps when there are twice as many frames... (I don't think you need to double the frame rate, but still...)

Otherwise I agree, though I'd rsther keep the mechanical SR and use the bandwidth for higher bitrates (especially proper VBR using CRF... Let me chose the CRF, and the camera uses as many bits as it needs, only cutting off when it exceeds memory card speeds (but using the SD cards in RAID 0 (?) mode could give enough speed). I'd also - if the readout speed from the sensor is high enough - want pixel binning. If the full 24 MP is too much, use every second pixel and downsample that. Or every third. Fourth. Anything would be better.

---

Last edited by kadajawi; 12-28-2013 at 12:52 AM.

[jogiba]

If they have that much bandwidth, they really should consider a 1536p mode so you can use Warp Stabilizer in post to do software SR to get back to 1080p...

4K 2160p with IS lenses will be the way to go in 2014 IMHO.

[kenyee]

Do they? Sony is doing 25p at 24 Mbps... aaaaand 50p at 28 Mbps.

Makes you wonder what the quality is at 50p, doesn't it?

2K would be too quick for Pentax in 2014, sadly. No consumer level devices yet at the $1300 price point they're happy in that do this (aka, the chips aren't fast enough and cheap enough to do this currently). I'd love to see the internal hardware of a RED...they probably are running a crapload of parts in parallel to get the speeds they need...

Looking forward to seeing the GH5 (they'll skip the gh4 moniker probably) w/ 4K support next year. They can't downsample as much so it should be interesting to see how much worse their DR will be...

[PiDicus Rex]

I'd love to see the internal hardware of a RED...they probably are running a crapload of parts in parallel to get the speeds they need...

So would a lot of us,... but from what I understand so far, it's more a matter of choosing a much higher bandwidth recording media, and having a lot more buffer memory.
The Wavelet Compression used in RedCode Raw can be easier on the processor too, as it's compressing a lot less then AVCHD.

[kadajawi]

I wonder if we're going to see added CPUs... Canon is just using several DIGIC CPUs depending on how much performance they need. The problem is that Pentax seems to be buying off the shelf parts... and unless their supplier gives them a suitable CPU...

I wonder what BlackMagic is using. Or if BlackMagic and Pentax could join forces. The K-3 Black edition. With the added hardware to make it a serious video camera. And of course sensor based SR.

As for storage speed... The K-3 has USB 3.0. They could just give it NTFS support (or exFAT for example) and write to that. Plug in a SSD and you're good to go. Those who want to shoot 4K RAW will probably be shooting with some rig anyway, which should give enough space for the drive.

I could also, for the next generation imagine a port at the bottom that gives direct access to the sensor. So that they can offer a grip that offers all that is neccessary for video shooting, like storage, CPU, encoder chip, ... . Still shooters wouldn't have to pay for it. Some Canons have a video version... I think the 1Dc...?

Isn't Pentax already doing 2K? Almost at least...

http://www.wirelessgoodness.com/2012/01/11/reds-35k-epic-x-camera-gets-torn-down-in-incredible-detail/

This link is better:
Http://www.brightsideofnews.com/news/2011/5/5/revolutionary-red-epic-camera-taken-apart-by-fcc.aspx

[PiDicus Rex]

Epic Link,....
(see what I did there,...)

"Buying off the shelf parts,." or parts to order?
It's not always right to say 'what the supplier gives them'. When an order is big enough, the supplier makes the parts to the Customers specifications. That's exactly what BMD does.
Right now, I wouldn't want Pentax and BMD paired up, not with the number of issues with the Pocket Cine Cam regarding Black Dots / White Orbs in the vision.

Speaking of, and back to the media bandwidth limitations, the K3 has a UHS1 SD Controller, with matched cards, that should be more then fast enough to record high data rate HD. I have been looking, but I can't find any information of the type of controller that is fitted to the BMD Pocket, but it is capable of recording ProRes and Compressed Cinema DNG to an SD card.
It's my opinion that the hardware inside the K3 should be capable of recording to these same formats, as they are LESS compressed then the AVCHD currently used.
So long as the UHS1 controller can sustain the data throughput, the K3 would be doing less CPU intensive operations by recording to higher bandwidth files.

Both the K3 and the BMD PC are recent cameras. If Magic Lantern can get RAW in Cinema DNG files out of models that were launched 2 and 3 years ago, there's no reason to assume the same isn't possible form a brand new Pentax.

[kenyee]

As for storage speed... The K-3 has USB 3.0. They could just give it NTFS support (or exFAT for example) and write to that. Plug in a SSD and you're good to go.

Someone tested various speed SD cards and reading the cards via USB 3.0. Turns out it's running at USB 2.0 speeds

[kadajawi]

Epic Link,....
(see what I did there,...)

"Buying off the shelf parts,." or parts to order?
It's not always right to say 'what the supplier gives them'. When an order is big enough, the supplier makes the parts to the Customers specifications. That's exactly what BMD does.
Right now, I wouldn't want Pentax and BMD paired up, not with the number of issues with the Pocket Cine Cam regarding Black Dots / White Orbs in the vision.

Speaking of, and back to the media bandwidth limitations, the K3 has a UHS1 SD Controller, with matched cards, that should be more then fast enough to record high data rate HD. I have been looking, but I can't find any information of the type of controller that is fitted to the BMD Pocket, but it is capable of recording ProRes and Compressed Cinema DNG to an SD card.
It's my opinion that the hardware inside the K3 should be capable of recording to these same formats, as they are LESS compressed then the AVCHD currently used.
So long as the UHS1 controller can sustain the data throughput, the K3 would be doing less CPU intensive operations by recording to higher bandwidth files.

Both the K3 and the BMD PC are recent cameras. If Magic Lantern can get RAW in Cinema DNG files out of models that were launched 2 and 3 years ago, there's no reason to assume the same isn't possible form a brand new Pentax.

Hahaha.


I doubt that BMD is selling more cameras than Pentax... if they do that clearly means Pentax didn't take the video market serious enough. I think BMD will be paying more for the chips than Pentax, so that they can get custom hardware. Encoding to ProRes or Cinema DNG is less CPU intensive than h264 perhaps, but I doubt that the h264 is encoded in software on a generic CPU. They will have a dedicated hardware encoder in there, which might also limit their options in bitrates and quality settings. Most ARM CPUs found in modern phones struggle at decoding 1080p h264 video... let alone encode it. They need to have a dedicated hardware decoder. Sadly I can't find a teardown of any BlackMagic camera... that would be quite interesting to see how much power they need, what they have inside the camera. Keep in mind those cameras were made for this sort of stuff, Pentaxes weren't. So even though they may be older designs... If the K-3 would cost more to add the necessary processing power those "oh I only want stills, remove the video functions!" guys here will be whining again.


Canon's have been hacked to record CineDNG, but not at full resolution, despite those cameras packing some serious processing punch. Maybe if the DNG isn't compressed, but then bitrates would be high. A 12 bit DNG file is 3.1 MB (uncompressed?). So at 24p that would already be 74.6 MB/s. Megabyte. Sounds doable. 93.3 MB/s at 30p.


I wonder why the EPIC needs such epic cooling. Certainly not feasible for a small and light DSLR like the K-3. Could the sensor heat up that much because it needs to read out every pixel? Then again, how can Panasonic do it with the GH3 then?


@kenyee: So... they took a USB 2.0 port and painted it blue?!

[pTom]

Could someone hook up a HDMI monitor while in video record mode, and let us know if you can turn off all the on-screen display items, and if the video image fills the entire HDMI screen ?

Would like to know if v1.01 fixes this, so the camera has Clean Feed Out.

Thanks in advance.

I've just updated K-3 firmware 1.01 and checked with BlackMagic Intensity Pro card , I'm afraid that the K-3 doesn't have Clean Feed Out . BM Media Express monitor shows nothing.


About K-3 USB 3.0 , i may say that it's neither USB 2.0 nor USB 3.0 . It's USB 2.2 I tested with palimpsest in Linux and Sandisk Extreme 45MB/s . The average speed on K-3 is 23.6 , while Lexar CF SD reader shows 44 .

[PiDicus Rex]

Someone tested various speed SD cards and reading the cards via USB 3.0. Turns out it's running at USB 2.0 speeds

Well, that was always doomed to fall in to the circular file marked "Fail".

Class 10 SD cards only need to read and write at a sustained 10 Mbytes/sec
USB2 supports a maximum data rate of 480mBits/sec, which is 60 Mbytes/sec
USB3 supports a maximum data rate of 5gBits/sec, which is 625 Mbytes/sec

When the SD cards are running slower then the speed of the USB, you cannot use it to evaluate the USB.
Even a UHS-1 SD card only has to run at 50Mbytes/sec, still slower then USB2.

Unless you can hook up an external SSD or HDD and record directly to it, and are able to run Burst Mode still with no slow down after the usual number of frames, you will have no accurate data as to whether the USB3 port is running at the full USB3 speed.

but I doubt that the h264 is encoded in software on a generic CPU.

As I understand the process (and I'm not 100% on it, only done light reading on it), CPU isn't the right term,.. The firmware is flashed in to a NAND based chip, and builds the processor in the NAND, rather then it being a CPU that is etched in to Silicon wafers on a production line.

So when you upload fresh firmware, you're rearranging the NAND to effectively be a new processor with different wiring.

So the processing of commands in done in a hardware layer, not just by new software.

Now, if it was an Atom or i7 or the like, then there is hardware in the CPU that does the compression work.

Regardless of which type of Chip is used, it should still be fast enough to write out full dynamic range Cinema DNG files.


pTom: what does it show on a HDMI screen?
Given my personal experience with an Intensity Shuttle, I'm not that confident that the Pro PCIe card would work any better.

And was that test done with the Class10 '45Mb' Extreme, or the UHS1 version?
I've got the class10 sort and the speeds I get from USB2 external card reader, vs the one built in to my laptop, vary by as much as 20Mb/s

[kenyee]

Class 10 SD cards only need to read and write at a sustained 10 Mbytes/sec
USB2 supports a maximum data rate of 480mBits/sec, which is 60 Mbytes/sec

FWIW, these max data rates don't really seem to ever get hit w/ USB...e.g., I used to run backups via the firewire port on my PC instead because I found that it's speed was nearly 3x as fast as USB 2.0 and that was backing up only to an external hard drive. USB 3.0 is similar speed now.

And I do have a PNY SD card that will write at a sustained 62MB/sec (rated at 60MB write) on my PC's USB 3.0 port, so some cards do really go that fast. The Extreme Pro's go up to 90MB/sec sustained writes. Both will read in the 90MB/sec range...

If you look at the wikipedia page for USB 3.0, there is a mode where you can design it for the connector/signal, but it's really USB 2.0 speeds. I'd guess they're more limited by the hardware feeding it though because it'd be stupid to design a chip for this crippled mode...

[kadajawi]

Well, that was always doomed to fall in to the circular file marked "Fail".

Class 10 SD cards only need to read and write at a sustained 10 Mbytes/sec
USB2 supports a maximum data rate of 480mBits/sec, which is 60 Mbytes/sec
USB3 supports a maximum data rate of 5gBits/sec, which is 625 Mbytes/sec

When the SD cards are running slower then the speed of the USB, you cannot use it to evaluate the USB.
Even a UHS-1 SD card only has to run at 50Mbytes/sec, still slower then USB2.

Unless you can hook up an external SSD or HDD and record directly to it, and are able to run Burst Mode still with no slow down after the usual number of frames, you will have no accurate data as to whether the USB3 port is running at the full USB3 speed.




As I understand the process (and I'm not 100% on it, only done light reading on it), CPU isn't the right term,.. The firmware is flashed in to a NAND based chip, and builds the processor in the NAND, rather then it being a CPU that is etched in to Silicon wafers on a production line.

So when you upload fresh firmware, you're rearranging the NAND to effectively be a new processor with different wiring.

So the processing of commands in done in a hardware layer, not just by new software.

Now, if it was an Atom or i7 or the like, then there is hardware in the CPU that does the compression work.

Regardless of which type of Chip is used, it should still be fast enough to write out full dynamic range Cinema DNG files.


pTom: what does it show on a HDMI screen?
Given my personal experience with an Intensity Shuttle, I'm not that confident that the Pro PCIe card would work any better.

And was that test done with the Class10 '45Mb' Extreme, or the UHS1 version?
I've got the class10 sort and the speeds I get from USB2 external card reader, vs the one built in to my laptop, vary by as much as 20Mb/s

If the speeds coming from the K-3 is lower than that coming from a USB 3.0 card reader that is a bad sign. Either the port is too slow, or the card reader inside the camera is too slow. In the latter case the problem would be with trying to write huge amounts of data very fast onto the card (i.e. Cinema DNG), but at least SSDs would be an option, with serious rewriting of the firmware. Which, knowing Pentax a bit, is not going to happen.


Does the Atom have encoder hardware? The Core i series does have QuickSync, but the output quality seems to be a bit meh.


Regular ARM smartphone chips should have h264 encoding in hardware, otherwise there is no way they could be recording video to h264 in real time, not at 1080p.


Wasn't NAND memory only? Do you mean FPGA? That's probably what RED did with their 1242. However from what I've read once you need more than, I don't know, not that many chips, you'll be cheaper with a proper ASIC chip, and those are faster, more power efficient, smaller, ...


Anyway, the K-3 can shoot somewhere around 8 fps at 24 MP... I wonder if that is the speed it can convert to DNG, and it is just limited by the card speed, or if the processor is slower than that. Doing 25 fps at 2 MP may not be that much of a challenge.


The K-3 should be using this processor, by the way. Milbeaut Seems like it can process 24 MP photos at 12 fps... if that is the processing speed at which it can encode to RAW we shouldn't have any issues with Full HD content.