RX10: Internal AVCHD vs External Recorder.

Wesley Byram said:
View attachment 80765

Ouch! Terrible macroblocking. So what happened? Well, in this case the sun hit the lens and caused a bit of flaring for a moment. I kind of liked the flare effect and J.J. Abrams would certainly approve, but the RX10's codec did not. To be honest, I was not expecting to see this much breakup in a shot like this, even if it only lasted a few frames.

So, I've now found an example of an otherwise ok shot that the codec couldn't handle.
Click to expand...

So sorry, but this is a terrible example. A backlit, poorly composed, lens flair is something J.J. Abrams might fire the camera operator for, not condemn the camera. Good luck finding any camera to handle this well. Expecting a camera or codec to turn poor camera technique into something useful is folly.

For those interested in a real world RX10 run & gun project, this may be useful:

 
dan.carter said:
So sorry, but this is a terrible example. A backlit, poorly composed, lens flair is something J.J. Abrams might fire the camera operator for, not condemn the camera. Good luck finding any camera to handle this well. Expecting a camera or codec to turn poor camera technique into something useful is folly.
Click to expand...

I'm hardly condemning the camera. I've already stated that if you use good technique, there usually doesn't seem to be a problem with the codec. In my previous posts, I mentioned that all of the breakup I observed was solely in shots with excessive camera movement. Here, I found one example of an exception, where the camera wasn't moving all that much, but instead it was the appearance of a lens flare that caused noticeable compression artifacts. I point it out for the sake of thoroughness. I picked a crop of the frame that showed the effect most strongly, and probably not coincidentally, it's also one of the darker parts of the frame.

As for J.J. Abrams firing a camera operator for such a shot, only if he would be willing to fire himself: http://www.theverge.com/2013/9/30/4788758/j-j-abrams-apologizes-for-his-overusing-lens-flares

In an interview with Crave Online, the man who will helm Star Wars: Episode VII confessed that there were so many lens flares in early cuts of Star Trek Into Darkness that he hired special effects house Industrial Light & Magic to erase some of his signature embellishments from the film.
Click to expand...
 
I have to agree with Dan here.
I don't understand how intentionally shooting poor footage at exaggerated settings gives any sense of what a camera can or can't do.

If you want to test a camera, as we do ours, extensively before production use, the logical way to do it is to test it under actual production conditions.

If you're shooting a narrative, propped and lit scene then that's how to test your cameras.
If you're shooting doc and real time, quick shooting (and you're any good at it) then you always find the "good light" and the best composition for every scene - not the worst.

The bad or borderline footage (a I believe you've stated Wesley) never makes it to edit, so why bother trying to reproduce it.
Looking at what you've shot and shown, one would think that this is some crappy little thing.

It can take some awesome footage, which is all I and my clients are interested in seeing from any camera.
Anyone can make footage look like s*%t, making a camera look good is what it's all about.
How good can you make the RX10 look, not how bad is the only creative imperative -and why I'm posting this.
 
jimagine said:
I have to agree with Dan here.
I don't understand how intentionally shooting poor footage at exaggerated settings gives any sense of what a camera can or can't do.

If you want to test a camera, as we do ours, extensively before production use, the logical way to do it is to test it under actual production conditions.

If you're shooting a narrative, propped and lit scene then that's how to test your cameras.
If you're shooting doc and real time, quick shooting (and you're any good at it) then you always find the "good light" and the best composition for every scene - not the worst.

The bad or borderline footage (a I believe you've stated Wesley) never makes it to edit, so why bother trying to reproduce it.
Looking at what you've shot and shown, one would think that this is some crappy little thing.

It can take some awesome footage, which is all I and my clients are interested in seeing from any camera.
Anyone can make footage look like s*%t, making a camera look good is what it's all about.
How good can you make the RX10 look, not how bad is the only creative imperative -and why I'm posting this.
Click to expand...

I don't disagree with any of this, but I feel I need to add some proper context. I'm not testing the camera. I'm testing the codec. I'm happy with the camera as a whole.

A number of reviewers pointed out the codec as something that they found as a particular weakness of the RX10. Reading elsewhere would give the impression that the RX10's codec is super robust, as good as the GH3's even. They couldn't both be right, so I was curious to settle this for myself. I have an upcoming project that involves a couple of steadycam style shots in a wooded area where a UFO will be composited in later. Being small and with a very sharp image, I want to see if the RX10 could do that job by itself.

I meant to accomplish two things: see if the codec is as bad as some people say and verify that the RX10's HDMI port really is outputting a clean, 4:2:2 image as Sony claims. On the second point, the camera passes with flying colors.

On the first point, it's still a little iffy for me. With the flycam test, I did not set out to deliberately break the image, nor to produce great footage either, just to walk around slowly with the camera on a flycam against a detailed background to see how the RX10 copes. It was mostly fine, but there was this one moment where the codec glitched, and I thought that it was worth mentioning as it contradicted what I had previously thought: that there would be no practical encoding issues as long as camera movement was controlled.

I'm not trying to make the RX10 look good or bad, just trying to find out where the boundaries of its performance are, at least w.r.t. the codec. It's easy enough to see how good it can look by just checking out Vimeo. (Case in point, Dan's beautiful Grand Canyon video really helped sell me on what this little cam can do.) I want to see where it breaks too, so I know what to avoid.
 
Last edited:
Not sure Wes. I've been flying the RX10 on the defy gimbal and the images are holding up fairly well. But then again I'm not gonna assume the superior codec of a 10 bit 444 on a $90,000 camera. It is what it is and the images are holding up well enough for web display. There is room to enhance the image in grading but one should get a proper exposure out of the camera. You're not going to get Hollywood drastic grades out of most consumer level cameras.

Knowing this, the camera is sharp and makes a nice image in a small but well featured device (nd, zebra, peaking, audio levels, constant 2.8, and 1080/60p with 24p and uncompressed out). It would be nice to hand us the better Sony codec and maybe they will. Oh yeah, it takes amazing stills too.

If we keep wanting or expecting 422 10 bit 200 Mbps codecs in $1200 cameras we're always going to be disappointed. And I give a rats a** what the bmd cameras give. They are loaded with moire and other quirks. They have their place just as Sony consumer camera have a place. But look at what Sony is doing. Pushing out competitive camera with 4k capabilities at reachable prices. It's just a good thing for those who like making images.
 
Thing about those reviewers that talked about the codec weaknesses - they weren't putting out anything but very average or below average images in their quick reviews.
They do this stuff to get web hits and make a living. Many of them aren't very good artists or DPs or photographers.

The codec is similar to the C100 which is good enough for just about anything in skilled hands.
With many of these cameras, we're at the point of micro comparative differences in regard to what the viewers of our work consider important.
An RX10 is close in IQ to a C100 which looks indistinguishable, in most circumstances, to a C300, which could be used for a TV show vs an Alexa if the DP wants or budget dictates.

Give a good film maker of any type, a C300 a GH3, let them have the lenses they need to compose their shots, and, all other things being equal, I'll bet 99% of the viewing public wouldn't have a preference or be able to distinguish.

That's why I was drawn to the RX10 (and the BMPCC in the same purchase) two little, cheap, cameras that could solve certain problems.
What the RX10 has going on that only the C100 does in the price for feature game is everything you need to just go out and get great footage easily.
As I get busier all the time, I find that the most compelling argument for any camera now or on the horizon.

FWIW Christopher, I just finished a weeks shooting the 60 sec opening for a commercial short that's just being delivered tomorrow.
We shot all the scenes on location through the summer on a pair of GH3s for convenience (awesome images) and I used the new BMPCC to shoot the set scenes for the opening staged montage (think the opening credits for "Babe" which was my inspiration.)
Tons of table top sets with lots of small fresnels and mirrors and detailed objects with the camera on a slider or jib for each shot and not a single shot with moire and just the littlest bit of aliasing in a few places.
I wasn't expecting that but there was truly no moire in 8 very detailed, moving scenes.

Oh yeah...and I really think the GH4 may be a tool I shoot for a while to come.
.
 
Yeah again, I basically agree with you guys, so let me turn it around and come at it from another angle: that of praise via faint condemnation. The fact that it takes dodgy conditions to introduce visible codec issues should be reassuring. In fact, I'm pretty sure I've satisfied my curiosity. The common element for those few glitches that I did spot has always been a sudden shift in the scene. It can be a quick camera move or the sudden appearance of a lens flare. I now know to always avoid that when recording internally with this camera. Besides, that's exactly the sort of thing that will cause a web stream to degrade anyway, so if the target medium is the web, then I can find no fault of consequence.

I will be shooting those UFO scenes with the Hyperdeck though, just in case.

To add a non-scientific point of comparison, I spotted similar breakup on my old rusty-trusty AVCHD shooting HMC40 exactly twice in the 3-odd years I've used it. Once it was when I was intentionally trying to see if would produce artifacts if I shook the camera around like a madman (it did, but only a little). The other time was on a greenscreen shoot when an actor's arm got pixelated for a moment.

The RX10 does produce a much better, more detailed image than that old thing, though.
 
jimagine said:
I don't understand how intentionally shooting poor footage at exaggerated settings gives any sense of what a camera can or can't do.
Click to expand...

Just to clarify, the only time I deliberately over stressed the codec and posted it here was at the beginning of this thread. The purpose of that was to demonstrate clearly that the HDMI port was behaving as it should, with no compression artifacts. I needed a clear difference between internal and uncompressed to make it obvious. (and with the re-encode test, I wanted to demonstrate how another h.264 encoder handles exactly the same shot.) The second set of shots was to show that the output is really 4:2:2. The third was as a curiosity, something that struck me as a little off, which in retrospect, I probably should have just mentioned in passing, but not actually posted images from.
 
Last edited:
Wesley Byram said:
Yeah again, I basically agree with you guys, so let me turn it around and come at it from another angle: that of praise via faint condemnation. The fact that it takes dodgy conditions to introduce visible codec issues should be reassuring. In fact, I'm pretty sure I've satisfied my curiosity. The common element for those few glitches that I did spot has always been a sudden shift in the scene. It can be a quick camera move or the sudden appearance of a lens flare. I now know to always avoid that when recording internally with this camera. Besides, that's exactly the sort of thing that will cause a web stream to degrade anyway, so if the target medium is the web, then I can find no fault of consequence.

I will be shooting those UFO scenes with the Hyperdeck though, just in case.

To add a non-scientific point of comparison, I spotted similar breakup on my old rusty-trusty AVCHD shooting HMC40 exactly twice in the 3-odd years I've used it. Once it was when I was intentionally trying to see if would produce artifacts if I shook the camera around like a madman (it did, but only a little). The other time was on a greenscreen shoot when an actor's arm got pixelated for a moment.

The RX10 does produce a much better, more detailed image than that old thing, though.
Click to expand...

I like it! And it doesn't give the "Wanna shoot fine cinema with absolutely no effort." crowd ammunition to discredit a camera they've never touched.

Was reviewing some of my HMC40 projects a few weeks ago. Things have come a long way in a short time.

Curse jimagine for bringing up the GH4! I've been hoping to get through a full year without another camera purchase. And 4K is going to cost us all a lot more than just the camera...
 
jimagine said:
I have to agree with Dan here.
I don't understand how intentionally shooting poor footage at exaggerated settings gives any sense of what a camera can or can't do.
Click to expand...

Sorry but to me that statement makes no sense. To find out the limit of any piece of equipment, you obviously DO have to push it PAST it's limit, even if it takes some extreme measures to do so. (Then, once you know where the limit is, you can determine whether or not that's a point you'll ever be likely to reach in your normal usage.)
 
I've never felt a need to do that with any of my studios or location gear, music or film making, or cars, or microwaves for that matter. So sure, it makes perfect sense to me.
But if it makes no sense to you then by all means flail away at your equipment with abandon, point your cameras at trash cans under the full moon, do a whip pan with a whip...and a pan.
 
jimagine said:
I've never felt a need to do that with any of my studios or location gear, music or film making, or cars, or microwaves for that matter. So sure, it makes perfect sense to me.
But if it makes no sense to you then by all means flail away at your equipment with abandon, point your cameras at trash cans under the full moon, do a whip pan with a whip...and a pan.
Click to expand...

Old habits die hard, I guess. When I was in school for Computer Science, a lesson they drilled into us early and often was: once your new system is ready to test, immediately try as hard as you can to break it. Don't just stick to safe "real world" scenarios, but throw everything at it. Deliberately give it bad input and be creative with it. If you're testing the accuracy of a sales tax computation subroutine, don't just give it realistic figures, also give it the first line of the Declaration of Independence written backwards in Klingon and see how it handles the error. Find every possible point of failure. If you can't find a fault anywhere, then you're not looking hard enough.

For several years I worked for credit card processing companies, and one of the jobs I had for a time was running repeated, tedious tests on new software for terminals as part of the process of certifying the software. It was a similar thing there. I would have to do things like load the software into the terminals over and over while timing the procedure down to the second, then report back with the numbers and also mention anything that seemed a little off, no matter how inconsequential.

Again, old habits die hard. :D
 
Last edited:
Thanks to this interesting discussion I bought my Sony RX10 without the recorder. Sounds like most of the time it provides excellent results if it is exposed correctly.
 
szimmer said:
Thanks to this interesting discussion I bought my Sony RX10 without the recorder. Sounds like most of the time it provides excellent results if it is exposed correctly.
Click to expand...

Enjoy your new RX10. I think you'll like it. Treat it right and it can be shockingly good, all things considered.
 
Wesley Byram said:
Enjoy your new RX10. I think you'll like it. Treat it right and it can be shockingly good, all things considered.
Click to expand...

Hi Wesley,

Yes, I am very excited. I am sure I will have some questions here to make the best use of it. My last camera was a HMC-150. This is a different beast but I am excited to have one piece of equipment to take great photos and good video.

Have a good weekend.

Simon
 
jimagine said:
I have to agree with Dan here.
I don't understand how intentionally shooting poor footage at exaggerated settings gives any sense of what a camera can or can't do.

If you want to test a camera, as we do ours, extensively before production use, the logical way to do it is to test it under actual production conditions.

If you're shooting a narrative, propped and lit scene then that's how to test your cameras.
If you're shooting doc and real time, quick shooting (and you're any good at it) then you always find the "good light" and the best composition for every scene - not the worst.

The bad or borderline footage (a I believe you've stated Wesley) never makes it to edit, so why bother trying to reproduce it.
Looking at what you've shot and shown, one would think that this is some crappy little thing.

It can take some awesome footage, which is all I and my clients are interested in seeing from any camera.
Anyone can make footage look like s*%t, making a camera look good is what it's all about.
How good can you make the RX10 look, not how bad is the only creative imperative -and why I'm posting this.
Click to expand...
Amen brother. I see more posts about what any camera not just the RX10, but especially the RX10, won't do in a supposed situation rather than what it can do when used properly.

I'm not a professional so at most I'll buy a Ninja2 for the highest quality output. However I think the camera's internal spec does a fine job on the footage I've seen. I just like gadgets and high quality and the Ninja2 is a useful one capable of very high quality.

Sure the internal codec has some moiré, a little aliasing and mushyness sometimes, but like jimagine stated we're trying to utilize the best scenes the camera can produce, not the worst and on that front the RX10 is a wimmer. Hands down.
 
HionHiFi said:
...
Sure the internal codec has some moiré, a little aliasing and mushyness sometimes, but like jimagine stated we're trying to utilize the best scenes the camera can produce, not the worst and on that front the RX10 is a wimmer. Hands down.
Click to expand...

Actually, the RX10's internal codec doesn't have any more aliasing/moire than uncompressed. The aliasing is coming from how the data from the sensor is downscaled to fit into a 1080p frame. It appears that what the RX10 does is read the entire sensor without binning or line-skipping, then uses a nearest-neighbour algorithm to scale it down to HD. The result is that there is a little bit of luma aliasing, but none of the nasty false color aliasing that tends to show up in on the Canon DSLRs (save the 5DMKIII) and the Blackmagics.


Regarding the codec, after spending way more time than is healthy comparing the RX10's AVCHD to it's own HDMI output, here's how it shakes out under different scenarios. Hopefully this will disambiguate the issue:

-Mostly static scene with slow, smooth camera movement: Internal and uncompressed are almost indistinguishable. Internal holds detail very well, even in the shadows. If you don't need 4:2:2, then there's not much reason to record externally.

-Talking heads, such as in an interview: Same as above. Internal codec is pretty close to perfect. External recorder is unnecessary.

-Dense detail covering most of the frame combined with complex motion, such as trees blowing in the wind while the camera is moving at normal walking speed on a stabilizer: Uncompressed starts to pull ahead of internal codec. Subtle signs of compression become visible in AVCHD, such as mosquito noise around areas of high contrast and blocking artifacts, especially in the shadows. Avoid adding post-sharpening as this will make the artifacts stand out more. A casual viewer might not notice, and Youtube will introduce as much or more degradation to this kind of footage. There's some gain to be had recording externally, though.

-Above scenario plus a sudden change affecting the whole frame, like a quick camera move or lens flare: Blocking artifacts can appear over the entire image for a few frames in AVCHD.

-Normal walking speed while hand-holding the camera, (think Blair Witch Project and others in the "found footage" genre): This is pretty much nothing but sudden changes from frame to frame. Significant differences between internal codec and uncompressed. Uncompressed is clearly superior. If you need to do this for some reason, record externally.

-Running while hand-holding the camera: Please...just don't do this. ;-)

There's natually going to be some variation in practice. If half the frame is clear blue sky, for example, then the RX10's AVCHD will fare better. If you use a high shutter speed, then it will do somewhat worse.

Hopefully the above information will be more useful than saying "the codec is great" or "the codec stinks."
 
Last edited:
Wesley Byram said:
Actually, the RX10's internal codec doesn't have any more aliasing/moire than uncompressed. The aliasing is coming from how the data from the sensor is down scaled to fit into a 1080p frame. It appears that what the RX10 does is read the entire sensor without binning or line-skipping, then uses a nearest-neighbor algorithm to scale it down to HD. The result is that there is a little bit of luma aliasing, but none of the nasty false color aliasing that tends to show up in on the Canon DSLRs (save the 5DMKIII) and the Blackmagics.


Regarding the codec, after spending way more time than is healthy comparing the RX10's AVCHD to it's own HDMI output, here's how it shakes out under different scenarios. Hopefully this will disambiguate the issue:

-Mostly static scene with slow, smooth camera movement: Internal and uncompressed are almost indistinguishable. Internal holds detail very well, even in the shadows. If you don't need 4:2:2, then there's not much reason to record externally.

-Talking heads, such as in an interview: Same as above. Internal codec is pretty close to perfect. External recorder is unnecessary.

-Dense detail covering most of the frame combined with complex motion, such as trees blowing in the wind while the camera is moving at normal walking speed on a stabilizer: Uncompressed starts to pull ahead of internal codec. Subtle signs of compression become visible in AVCHD, such as mosquito noise around areas of high contrast and blocking artifacts, especially in the shadows. Avoid adding post-sharpening as this will make the artifacts stand out more. A casual viewer might not notice, and YouTube will introduce as much or more degradation to this kind of footage. There's some gain to be had recording externally, though.

-Above scenario plus a sudden change affecting the whole frame, like a quick camera move or lens flare: Blocking artifacts can appear over the entire image for a few frames in AVCHD.

-Normal walking speed while hand-holding the camera, (think Blair Witch Project and others in the "found footage" genre): This is pretty much nothing but sudden changes from frame to frame. Significant differences between internal codec and uncompressed. Uncompressed is clearly superior. If you need to do this for some reason, record externally.

-Running while hand-holding the camera: Please...just don't do this. ;-)

There's naturally going to be some variation in practice. If half the frame is clear blue sky, for example, then the RX10's AVCHD will fare better. If you use a high shutter speed, then it will do somewhat worse.

Hopefully the above information will be more useful than saying "the codec is great" or "the codec stinks."
Click to expand...
Great breakdown Wesley. Thanks for that.
 
I've got an rx10 which I like very much so I picked up a ninja 2. I can only get the ninja to record when I select 1080i on the rx10. It doesn't recognize the signal of 1080p. Is this the best output I can get from the rx10? I'm guessing the ninja is taking the 60i and converting it to 30p??

thanks,
T
 
You must log in or register to reply here.
Back
Top