Press Release: HMC150, "the new DVX"

[filmguy123]

So, to clarify... please tell me if I state anything here incorrectly:


The problem with a statement like "[FONT=verdana, arial, helvetica]Even the $650 Canon HV20 with its native 1920x1080 CMOS chip out-shoots the HVX200, and will certainly beat this lame has-been-out-of-the-box AG-HMC150." Is the following:

- CCDs are currently overall better than CMOS, although CMOS may have great future potential

- native 1920x1080 may yield better resolution on the HV20, however, this is countered by being forced to use greater compression to keep the data rate and recording size manageable. What this means is using long-GOP compression vs Intra-Frame, and as a result, the effective resolution in movement or in action may actually be worse than the HVX's I-Frame 1280x1080 resolution (which also uses superior CCDs). Long-GOP also makes editing more difficult and can cause problems since it is not Intra-Frame.

Further, the HMC150 will employ the superior AVC-HD codec, which is a better compression scheme than HDV that looks better at lower bitrates. This why it can shoot 13Mbps rather than 19 or 24 and STILL look better than HDV. Another advantage of this AVC-HD codec (but not necessarily the HMC150) is that this lower bit rate allows bandwidth to be "spent" on other things, such as higher resolution and better color sampling. Overall, AVC-HD will allow much higher quality HD recording due to its superior compression scheme.

[/FONT][FONT=verdana, arial, helvetica] The problem is, there have been no *GOOD* AVC-HD cameras out yet to show this. The technology in the codec is superior, but the footage captured has thus far been with cheap consumer cameras, which people compare to prosumer HDV cameras and mistakenly claim HDV to thus be superior.

[/FONT][FONT=verdana, arial, helvetica]Yet, AVC-HD it still is a long-GOP based compression, meaning that it will never out-shoot the DVCPRO-HD I-frame compression (except in still image tests or low motion footage, such as interviews). This is because any long-GOP compression scheme will suffer from motion (although long-GOP compression has improved quite a bit over time and can now deal with motion pretty well). However, it can still be problematic. However, when combing long-GOP with CMOS chips, you have additional problems that crop up in motion, strobing lights, etc. CCDs & I-Frame Compression are a better combination. Of course, with I-Frame compression, due to bandwidth limitations, you cannot currently achieve native 1920x1080 resolution, and so scaling must be used. However, scaling is an oft used and very acceptable and legitimate way to achieve HD resolution.

All this is about to change, however, thanks to AVC-Intra. The superior compression in AVC-Intra allows higher quality images to be captured at a lower bitrate, freeing up more bandwidth to be used on resolution or color sampling. AVC-Intra @ 50 Mbps is equivalent to DVCPRO-HD @ 100Mbps. This means, when you pump AVC-Intra up to 100 Mbps, you can now achieve both Intra-Frame recording *AND* native 1920x1080 resolution!





----- Is this all correct? Did I say anything wrong, and/or am I missing any important chunks here?

Thanks I finally feel like I get it!! Woohoo!

[/FONT]

 

[filmguy123]

AND - a few more questions:

1.) AVC-Intra is a more advanced compression technique than DVCPRO-HD. Does this mean that, like AVC-HD, it will require more processing power to edit? Why or why not?

2.) AVC-Intra will allow for native 1920x1080 images, meaning 1920x1080 CCDs. However, this also means more pixels crammed onto a sensor = smaller pixels = decrease in light sensitivity. So, are we likely to see WORSE low-light performance still when AVC-Intra100 1/3 CCDs cameras come out... UNLESS they are first able to make each CCD faster? If implemented into current 2/3rd CCDs, such as the HPX500, does this also mean that the HPX500 would see a decrease in sensitivity? If so, what advances are being made in CCDs to get them faster so that low-light will not suffer when full 1920x1080 native CCDs are introduced?

3.) Are native 1920x1080 sensors the reason that the Sony EX1 had to use 1/2 sensors, in order to maintain decent low light performance? Why did they choose CMOS over CCDs? Are CMOS sensors cheaper?

4.) We know that DVC-PRO HD records 100Mbps [Megabits/sec] (around 12MBps [Megabytes/sec]). However, what is the overall bandwidth capability of P2 (notated in both Megabits and in Megabytes)? I am assuming it is musch faster than SDHC, which is capable of 6MBps (megabytes/sec) sustained or 20MBps (megabytes/sec) maximum data transfer rate?

 

[mikkowilson]

[FONT=verdana, arial, helvetica]----- Is this all correct? Did I say anything wrong, and/or am I missing any important chunks here?[/FONT]

Yup, starting to get the right idea.

AND - two more questions:

1.) AVC-Intra is a more advanced compression technique than DVCPRO-HD. Does this mean that, like AVC-HD, it will require more processing power to edit? Why or why not?

2.) We know that DVC-PRO HD records 100Mbps [Megabits/sec] (around 12MBps [Megabytes/sec]). However, what is the overall bandwidth capability of P2 (notated in both Megabits and in Megabytes)? I am assuming it is musch faster than SDHC, which is capable of 6MBps (megabytes/sec) sustained or 20MBps (megabytes/sec) maximum data transfer rate?

AVC-Intra probably does require more processing power, yes. But still no where near as much as an inter-frame format that requires images be processed in large groups at a time. With an intra-frame format, the processor can handle each frame on it's own, which is less work.


P2's maximum data-rate is 640Mbps. (80MBps). Certainly not a problem for dealing with 100Mbps DVCPRO-HD streams, even at multitudes of real time.

- Mikko

 

[Barry_Green]

So, to clarify... please tell me if I state anything here incorrectly:

Your summation is largely accurate. There are minor things I could point out, but for the most part, yes you've got it.

The problem with a statement like "[FONT=verdana, arial, helvetica]Even the $650 Canon HV20 with its native 1920x1080 CMOS chip out-shoots the HVX200, and will certainly beat this lame has-been-out-of-the-box AG-HMC150." Is the following:

[/FONT]
There's many problems with that guy's statement, not the least of which is: the HVX is at least two stops faster than the HV20. The HV20 isn't a native 1920x1080 sensor, it's actually a 3-million-pixel sensor with a Bayer pattern over it. As such, it has extraordinarily tiny pixels and delivers a final ISO rating of about 80, vs. about 320 for the HVX. The de-Bayered performance is about akin to 1920x1080 though. Secondly, the rolling shutter can render footage from the HV20 unusable in several circumstances. And third, there's no way to bypass the interlaced recording system (i.e., there's no "24pA" equivalent) so you're always going to have interlace issues that need to be dealt with, and that means you're getting less compression efficiency.

[FONT=verdana, arial, helvetica]- CCDs are currently overall better than CMOS, although CMOS may have great future potential

[/FONT]
In general, there is no such thing as "overall better", there's only "better for a specific purpose". CMOS has several advantages, including lower cost and lower power consumption; a camera like the Red or the EX1 could not have been done on CCD. But, both of them have rolling shutter issues because of their CMOS origins. Several of us (me, Dalsa, and Sony) all believe that CCD is the better solution for digital cinema applications. But that doesn't mean they're better "overall" than CMOS; there are areas where CMOS does things that CCDs don't.

[FONT=verdana, arial, helvetica]

Another advantage of this AVC-HD codec (but not necessarily the HMC150) is that this lower bit rate allows bandwidth to be "spent" on other things, such as higher resolution and better color sampling. Overall, AVC-HD will allow much higher quality HD recording due to its superior compression scheme.

[/FONT]
Yes, except for the bit about better color sampling; AVC-HD is limited by design to 4:2:0. There is no provision for better color sampling in the specification.

[FONT=verdana, arial, helvetica]

which people compare to prosumer HDV cameras and mistakenly claim HDV to thus be superior.

[/FONT]
Even Sony, the largest HDV maker on the globe, says AVC-HD is superior to HDV. How any respectable reviewer could claim otherwise is just not comprehendable.

[FONT=verdana, arial, helvetica]

This means, when you pump AVC-Intra up to 100 Mbps, you can now achieve both Intra-Frame recording *AND* native 1920x1080 resolution!

[/FONT]
And, Panasonic claims it's visually equivalent to HD-D5, the de-facto Hollywood studio mastering format.

 

[filmguy123]

Yup, starting to get the right idea.

Awesome! "starting to" or "got"? haha, as long as I didn't say anything WRONG or semi-wrong... right???

P2's maximum data-rate is 640Mbps. (80MBps). Certainly not a problem for dealing with 100Mbps DVCPRO-HD streams, even at multitudes of real time.

OK in the meantime, I read here (http://www.broadcastnewsroom.com/articles/viewarticle.jsp?id=268586) that "P2 cards handle a data transfer speed of 1.2gbps per second to SxS 2.5gbps" - 1.2gbps = 1228MBps or 153MBps, about twice as much as you mentioned. Is that 1.2gbps misinformation, or where is that drawn from? If so, does that mean SxS has a much greater bandwidth capability? Isn't this a benefit of SxS?

 

[Barry_Green]

And yes, AVC-Intra definitely requires a lot more horsepower than DVCPRO-HD did. On EDIUS you can get one stream of AVC-Intra in realtime on a dual-core system; with DVCPRO-HD I've had four streams playing simultaneously on the same laptop.

nVidia and ATI graphics cards have onboard AVC hardware acceleration; I would expect that sooner or later someone'll come out with a hardware hook that'll let somebody's NLE tap into that power and when they do, AVC-HD and AVC-Intra editing will become a lot faster.

 

[filmguy123]

2.) AVC-Intra will allow for native 1920x1080 images, meaning 1920x1080 CCDs. However, this also means more pixels crammed onto a sensor = smaller pixels = decrease in light sensitivity. So, are we likely to see WORSE low-light performance still when AVC-Intra100 1/3 CCDs cameras come out... UNLESS they are first able to make each CCD faster? If implemented into current 2/3rd CCDs, such as the HPX500, does this also mean that the HPX500 would see a decrease in sensitivity? If so, what advances are being made in CCDs to get them faster so that low-light will not suffer when full 1920x1080 native CCDs are introduced?

3.) Are native 1920x1080 sensors the reason that the Sony EX1 had to use 1/2 sensors, in order to maintain decent low light performance? Why did they choose CMOS over CCDs? Are CMOS sensors cheaper?

Wow, thanks guys you are on the ball! haha. While you were writing I actually modified my post with 2 more questions that I think I modified while you were responding!

 

[Barry_Green]

"P2 cards handle a data transfer speed of 1.2gbps per second to SxS 2.5gbps" - 1.2gbps = 1228MBps or 153MBps, about twice as much as you mentioned. Is that 1.2gbps misinformation, or where is that drawn from? If so, does that mean SxS has a much greater bandwidth capability? Isn't this a benefit of SxS?

Yet another lie shrouded in truth.

The INTERFACE that P2 uses has a 1.2gbps data transfer rate, and the INTERFACE that SxS uses has 2.5gbps max. However, the underlying memory technology is nowhere close to exploiting that; neither a P2 card nor an SxS card comes anywhere close to that, and in fact an SxS card doesn't come close to maxxing out what a P2 card's interface can handle!

P2 uses Cardbus, with a potential maximum speed of 1.2gbps. But the P2 card system is currently rated at 640mbps maximum transfer rate (or, about half of what the total bus speed can handle). SxS, on the other hand, uses the ExpressCard bus, with a potential maximum speed of 2.5gbps. But SxS cards aren't designed to even nearly approach that, they are spec'd to max out at a claimed 800mbps.

It's like if you took a moped (a motorized bicycle with a maximum allowable legal limit of 27mph) and put it on the freeway. Just because there's a higher speed limit, that doesn't mean the moped can go any faster! It'll go 27mph, even though the speed limit may be 70mph.

So, we've got fanboys out there claiming that SxS is "faster" than P2, even claiming that it's "twice as fast" -- um, no. If you put an 8GB SxS card in a laptop and copy its contents to a hard disk, it takes a little over 4 minutes. Put an 8GB P2 card in that same laptop, copy its contents to that same hard disk, and it'll take a little over 4 minutes.

Here's the hard cold fact: SxS and P2 are exactly the same price, and basically exactly the same speed. And P2 does have room to go faster, if they RAIDed more chips together they could go about 50% faster. SxS may or may not have room to increase, depending on whether they're using a single block of memory or if they're raided.

 

[Barry_Green]

2.) AVC-Intra will allow for native 1920x1080 images, meaning 1920x1080 CCDs.

You don't have to use 1920x1080 CCDs to get 1920x1080 images. The HPX2100 delivers fantastic images in AVC-Intra and it uses 1280x720 chips. You'd have to use spatial offset to take advantage of this though.

However, this also means more pixels crammed onto a sensor = smaller pixels = decrease in light sensitivity.

Correct.

So, are we likely to see WORSE low-light performance still when AVC-Intra100 1/3 CCDs cameras come out... UNLESS they are first able to make each CCD faster?

Of course.

If implemented into current 2/3rd CCDs, such as the HPX500, does this also mean that the HPX500 would see a decrease in sensitivity?

Of course.

If so, what advances are being made in CCDs to get them faster so that low-light will not suffer when full 1920x1080 native CCDs are introduced?

That's a question for chip engineers. There are several techniques that have been employed; for example, Canon and Sony used interlaced chips in their first-generation HDV cameras because interlaced scanning gives you twice the sensitivity and lower noise; that allowed them to pack more pixels on the chip and get comparable sensitivity. At the expense of losing resolution when going into "faux progressive" mode. Sony's EX1 is employing all sorts of noise reduction wizardry to get decent sensitivity out of its chips; it lets them "gain up" the chips internally without noise showing up because it's been electronically compensated for at the chip level. As technology improves, solutions such as this will present themselves.

In fact, I would bet that the newer Panasonic HMC150, if it indeed uses a "new chip block" as Engadget said, should benefit from some of those advances in technology.

3.) Are native 1920x1080 sensors the reason that the Sony EX1 had to use 1/2 sensors, in order to maintain decent low light performance?

Yes.

Why did they choose CMOS over CCDs? Are CMOS sensors cheaper?

Yes, much cheaper, but also more power-efficient. CCDs draw a lot more power than CMOS does, and power consumption also = heat. Had they used CCDs they would have very short battery life and excessive heat troubles when trying to put CCD chips in a body that small. Going CMOS let them use small lithium-ion batteries and actually get the heat performance down to a manageable level. At the expense of introducing rolling shutter issues. Sony uses CCDs in all their digital cinema cameras, and Juan Martinez of Sony said that CCD is the superior technology for digital cinema applications in big-chip cameras. CMOS was appropriate for a handheld unit specifically because of the size and power issues; he said that if they'd put three 1920x1080 1/2" CCDs in a form factor as small as the EX1, it would have run so hot it would have melted.

However, what is the overall bandwidth capability of P2 (notated in both Megabits and in Megabytes)?

P2 cards are rated at 640mbps, or 80 megabytes per second. There is no published spec for actual read and write speed, only the 80mBps figure is published.

 

[filmguy123]

I see. Wow, ok. The learning never stops, haha!

Barry, have you thought of combining all of the information even in the last few pages of this thread into an article of some sort that would help sort out all of this stuff? It has been incredibly useful, and is excellent resource that I think many would find helpful. Best of all, most of it is already written!

I've been reading the AVC-Intra FAQ and after reading what you said about processing power needs, I had a thought:

- If P2 is capable of 640Mbps and DVCPRO-HD is only using 100Mbps, why not create the option of DVCPRO-HD200? Basically, allowing for 1920x1080 native images with 4:2:2 and Intra-Frame, without needing the more processor intensive AVC-Intra codec? Sure, it would take more space, but P2 card sizes will be up to 64gb by years end, which would be 30 minutes or DVCPRO-HD200 recording per card. Yes it would take a lot of HDD space, but for some people that may be more desirable than needing more processor power. Why not give the option? Do you think we might see something like this?

- P2 cards - so, we might being seeing a P2-Ultra card someday w/a higher bandwidth? Although, really, what is the point. Even DVCPRO-HD200 would utilize less than 1/3 of the P2 available bandwidth, correct? Really, there's no need for P2 to be any faster, is there?

 

[filmguy123]

You don't have to use 1920x1080 CCDs to get 1920x1080 images. The HPX2100 delivers fantastic images in AVC-Intra and it uses 1280x720 chips.

I don't understand; how does this work? Or, where can I read up about it?

If that's the case, that you can get TRUE 1920x1080 images using 1280x720 chips, isn't that more desireable, because then the low light sensitivity is better?

 

[Barry_Green]

- If P2 is capable of 640Mbps and DVCPRO-HD is only using 100Mbps, why not create the option of DVCPRO-HD200? Basically, allowing for 1920x1080 native images with 4:2:2 and Intra-Frame, without needing the more processor intensive AVC-Intra codec?

Because that would involve creating a new standard, new SMPTE certification, and take up twice as much space, cut your recording time in half, and double your archiving needs.

As I understand it, Panasonic went with the new format because they were able to give a substantial boost in quality without taking up more space. Bandwidth limits are a very real issue to broadcasters, and archiving is always a question that gets raised. That's why AVC-Intra50 exists at all; comparable quality at half the storage space. Since they can deliver D5-HD quality in 100 megabits, I don't think there's much appetite for doubling the file sizes.

The main advantage to the high-speed card is offloading and editing; you can edit up to six streams of DVCPRO-HD simultaneously, in real time, straight off the card. That's something you can't do from hard disks. The P2 system's origins were in instant-turnaround news applications, and that was a major design criteria.

- P2 cards - so, we might being seeing a P2-Ultra card someday w/a higher bandwidth?

I hope so. They've been RAIDing four SD cards together and getting 640mbps; I'd like to see them RAID eight cheaper cards together and get faster performance at lower price per gig. But, just because it's logical and obvious, that doesn't mean it'll happen!

Although, really, what is the point.

Offload speed.

Really, there's no need for P2 to be any faster, is there?

Need? No. Desire? Obviously. Who doesn't want faster?

 

[Barry_Green]

I don't understand; how does this work?

Very few people understand; they think that the pixels on the chip = the resolution of the system, and it just ain't so.

Or, where can I read up about it?

It's called Spatial Offset (sometimes referred to as "pixel shift"). The only halfway decent description of the technique is a white paper on Panasonic's HVX200 page, go to "support & resources" and look for "Tech Paper - AG-HVX200's Advanced Progressive CCDs"

If that's the case, that you can get TRUE 1920x1080 images using 1280x720 chips, isn't that more desireable, because then the low light sensitivity is better?

Yes, but as with everything there's a tradeoff. There is no free lunch. Spatial offset works to create up to a 50% increase in resolution (the BBC says 33% is a more realistic figure), at the expense of introducing some chroma noise and high-frequency aliasing.

 

[filmguy123]

Oh cool! I didn't realize you could edit directly off P2 at such a high data rate. So, you'd need a HDD capable of 80MBps to match that, correct?

However, if editing P2 over firewire, you are limited to Firewire's 480mbps transfer, right? So what do you use to take full advantage of the 640mbps? A cardbus slot directly to your laptop or PC?

 

[Barry_Green]

Yep.

 

[filmguy123]

I read the CMOS article. So if sounds like if CMOS would just use a global shutter, it would be superior to CCD since it would exhibit none of those four artifacts, while being cheaper to produce and more energy efficient. So why don't they just use a global shutter? Why doesn't the RED use a global shutter? Isn't that the "Ultimate" solution?

 

[filmguy123]

And also, isn't a rolling shutter the same as what an actual physical shutter of film camera uses? I read in the HVX book that a film camera using a rotating disc; isn't this the same concept as a rolling shutter? So shouldn't a rolling shutter give more of a film look...? I'm confused.

 

[Barry_Green]

I read the CMOS article. So if sounds like if CMOS would just use a global shutter, it would be superior to CCD since it would exhibit none of those four artifacts, while being cheaper to produce and more energy efficient. So why don't they just use a global shutter? Why doesn't the RED use a global shutter? Isn't that the "Ultimate" solution?

There is no free lunch. Adding a global shutter would mean a massive price increase, and the additional transistors take up space on the chip so that'll mean less sensitivity (less room for light-sensitive pixels). If you want a global shutter, a CCD makes a lot more sense.

 

[Barry_Green]

And also, isn't a rolling shutter the same as what an actual physical shutter of film camera uses? I read in the HVX book that a film camera using a rotating disc; isn't this the same concept as a rolling shutter? So shouldn't a rolling shutter give more of a film look...? I'm confused.

Lots of people are. The rolling shutter of a film camera exposes a large wide swath that means the entire frame is exposed simultaneously; the "rolling" effect only happens for a smaller percentage of the time. On a CMOS camera only a thin band is exposed at once.

If you exposed a film camera with a 20-degree shutter, then maybe you'd get similar performance to a rolling-shutter CMOS.

Several people are reporting that film suffers the same artifacts that rolling-shutter CMOS does; I think that's nonsense, we've never seen "jello" or discontinuous lighting causing rolling bands through film footage. I keep meaning to test this; I should put my Konvas up against my Red and shoot some rolling-shutter-inducing stuff to see just how similar or dissimilar they are.

 

[filmguy123]

So in the HVX when using film cam mode, is it still a global shutter, just notated in film shutter degrees?