Anyone having card speed issues, please try again without native 24p

[PappasArts]

Yep, those work.

Thanks for checking Barry- I'm getting good detail out of these settings so far. I'm gonna up them a little bit at a time.

Here's a frame from an anamorphic test. Though this level of anamorphic glass is a detail thief; this
added boost in quality to the codec is showing less of that, and giving it as much as it can back.

 

[Mike@AF]

That's interesting idea! A constant bit rate- that would great if so.

If I were to test it, assuming it actually records, how would I know if it's constant bitrate or not? I'm on a Mac and use FCP.

 

[Lpowell]

My Settings are:
54.000.000
58.000.000
68.000.000
GOP: 15 - Deselected as it's 15 anyways.
24P NATIVE

I just tried these settings in AVCHD 720p 30fps and 1080p 24fps, using a Lumix 45-200mm auto-focus zoom. Panning across a field of grass, 720p recorded properly, but 1080p failed with an SD card writing speed error. I tested a Class 6 Transend 16GB SD card and a Class 10 Sandisk 16GB - both failed in the same manner.

Under the same test conditions, I have found the following AVCHD settings to work properly without any failures so far:

Video bitrate Simplified 50000000
Overall bitrate adjustment 52000000
Limiting bitrate adjustment 6000000
Native 24p
720p 59.94fps -> 29.97fps
720p GOP length 15
1080p GOP length 12

 

[Park Edwards]

No, if you read back I said I can't use the Native 24p patch with ANY bitrate settings, INCLUDING the stock settings. How many times do I need to repeat that? That translated says: "i've ALREADY tested ALL STOCK and A/B/C settings and cannot get any of them to work with the Native 24p settings IF the video is of a highly detailed STATIC picture". IF there is any movement in the frame, I can record as long as I want on just about any bitrate up to the limit of the class of card I am recording to. I can also record as long as I want to with just about any bitrate I want to without the Native 24p setting enabled, up to the point where the bitrate is too much for the card, then I get a legitimate card speed error.

Car3o, c'mon man, I'm trying to help out here. Don't try to negate what I'm seeing and trying to report on, that's not helpful to Vitaliy or the other testers.

I just posted my results for the "B" Ptool settings back a few posts.

stock settings are normally conveyed as the settings that come in camera. you're getting way too emotional guy. if you can't use the 24p patch with any settings it's YOUR card or camera. how many times must that be repeated? but nevertheless, no point in going back and forth over this. hope you get it all sorted out.

and it could help to mention if shooters mention their shutter speed and their preset, whether it's standard, nostalgia, etc...it might have to do with why yours isn't working. one tiny setting may be wrong. also, if you haven't already, format the card in camera, or format the card on the computer and see if that changes.

 

[Svart]

stock settings are normally conveyed as the settings that come in camera. you're getting way too emotional guy. if you can't use the 24p patch with any settings it's YOUR card or camera. how many times must that be repeated? but nevertheless, no point in going back and forth over this. hope you get it all sorted out.

and it could help to mention if shooters mention their shutter speed and their preset, whether it's standard, nostalgia, etc...it might have to do with why yours isn't working. one tiny setting may be wrong.

You are missing the statement I made about movement. How many times does that have to be repeated?

Just about EVERY clip I've seen has folks wildly waving their cameras around like crazy, which I can do just fine too with Native 24p enabled. It's when the scene I am shooting is perfectly still. That's when I get the card speed error.

I'll repeat, It's ONLY when the high detail scene is perfectly still. No wind, no birds, no people, no cars, nothing moving. If there is so much as a blade of grass moving around or a bug crawling on the ground in the shot, I don't get the errors.

How can it be my card or camera when I only get failures like this ONLY with Native 24p enabled?

I have also reset my camera many times, with and without custom in-camera settings, it doesn't matter. this doesn't work right no matter what I do.

All I'm saying is someone needs to look at this.

And Car3o, I'm not taking this personally and I don't mean to sound angry or anything, I just get frustrated when folks just brush comments aside because they don't see the same issue. I'm an EE and I deal with problems that other people don't see for a living so I kinda tune in on issues that have patterns, and this one definitely does.

I literally have 42 unique firmware files sitting on my desktop right now from trying all of these combinations. That's not to mention the 20 or so I deleted already. I've been doing one change at a time and trying it.

Once we have a pair of files from someone to compare and analyze I think we might shed some light on the issue.

 

[PappasArts]

I just tried these settings in AVCHD 720p 30fps and 1080p 24fps, using a Lumix 45-200mm auto-focus zoom. Panning across a field of grass, 720p recorded properly, but 1080p failed with an SD card writing speed error. I tested a Class 6 Transend 16GB SD card and a Class 10 Sandisk 16GB - both failed in the same manner.

Under the same test conditions, I have found the following AVCHD settings to work properly without any failures so far:

Video bitrate Simplified 50000000
Overall bitrate adjustment 52000000
Limiting bitrate adjustment 6000000
Native 24p
720p 59.94fps -> 29.97fps
720p GOP length 15
1080p GOP length 12

1080p GOP should be deselected- Keep it unchecked.

Did you mess with audio bit rate?

Deselect any patches to 720P. Focus on 1080P.

Format both cards and run the test again. All manual mode. No auto focus or stabilizer- Nothing. Straight manual. Lets see what happens.

Barry tried my settings, they worked for him as well. I have shot tons of stuff. So there's a cog in the wheel somewhere that's causing the issue- lets find it.

Pappas

 

[Svart]

Thanks for checking Barry- I'm getting good detail out of these settings so far. I'm gonna up them a little bit at a time.

Here's a frame from an anamorphic test. Though this level of anamorphic glass is a detail thief; this
added boost in quality to the codec is showing less of that, and giving it as much as it can back.

Pappas, that's very similar to the scene I am shooting. If you are shooting on a Sandisk and that is working, do you have another brand card you could try with those same settings and that same scene?

Your settings of:

54.000.000
58.000.000
68.000.000

should be well within a class 10's ability to record.

 

[Svart]

I just tried these settings in AVCHD 720p 30fps and 1080p 24fps, using a Lumix 45-200mm auto-focus zoom. Panning across a field of grass, 720p recorded properly, but 1080p failed with an SD card writing speed error. I tested a Class 6 Transend 16GB SD card and a Class 10 Sandisk 16GB - both failed in the same manner.

Under the same test conditions, I have found the following AVCHD settings to work properly without any failures so far:

Video bitrate Simplified 50000000
Overall bitrate adjustment 52000000
Limiting bitrate adjustment 6000000
Native 24p
720p 59.94fps -> 29.97fps
720p GOP length 15
1080p GOP length 12

Your class 6 is probably getting choked by the 68mbit bitrate. That's likely too much. The max I could get out of my class 6 card was around 58mbit before it puked.

 

[PappasArts]

if you can't use the 24p patch with any settings it's YOUR card or camera.

I gotta agree, If all else is equal, then it's got to be your either one or both.

That's said lets work to figure it out, and not squabble. We're all part of a global team guys- doing something that is unbelievably amazing. In all my years in this biz, I've never seen anything like this. At the end the road, we all want the same thing , one way or another..

We'll get this figured out...

Pappas

 

[PappasArts]

Your class 6 is probably getting choked by the 68mbit bitrate. That's likely too much. The max I could get out of my class 6 card was around 58mbit before it puked.

How can I find out the bit rate of my settings?

In my MJPEG footage., it was easy with QT.

 

[Svart]

I gotta agree, If all else is equal, then it's got to be your either one or both.

That's said lets work to figure it out, and not squabble. We're all part of a global team guys- doing something that is unbelievably amazing. In all my years in this biz, I've never seen anything like this. At the end the road, we all want the same thing , one way or another..

We'll get this figured out...

Pappas

Well it's not equal and never can be because enabling or disabling Native 24p makes things unequal.

Everything else is equal.

Here is my exact sequence:

1. camera set to 1/50( also tested 60,125), 200 ISO(also tested 100, 250, 300, 640) FHD/HD(both give exactly the same errors in the same way).
2. camera set on tripod, focused at grass, NO movement, exposure set to 0.
3. class 6 and class 10 cards record normally with stock firmware. NO Native 24p.
4. class 6 and class 10 cards record normally with only "A" bitrate settings. NO Native 24p.
5. class 6 and class 10 cards record normally with only "B" settings. NO Native 24p.
6. class 6 and class 10 cards record normally with only "C" settings. NO Native 24p.
7. class 6 card FAILS and class 10 PASSES with these settings verifying that bitrate is truly higher than Class 6 card can handle. This was intentional and proves that the bitrate settings are truly working.

no native 24p
64M
72M
80M

8. Did not test above 80Mbit.

9. class 6 and class 10 cards PASS with stock firmware bitrate settings and Native 24p enabled while camera is moving.

10. class 6 and class 10 cards PASS with "A" settings and Native 24p enabled while camera is moving.

11. class 6 and class 10 cards PASS with "B" settings and Native 24p enabled while camera is moving.

12. class 6 and class 10 cards PASS with "C" settings and Native 24p enabled while camera is moving.

13. class 6 card fails and class 10 PASSES with these settings verifying that bitrate is truly higher than Class 6 card can handle. This was intentional proves that the bitrate settings are truly working.

native 24p enabled
64M
72M
80M
Camera IS moving

14. class 6 and class 10 cards FAIL with stock firmware bitrate settings and Native 24p enabled while camera is NOT moving.

15. class 6 and class 10 cards FAIL with "A" settings and Native 24p enabled while camera is NOT moving.

16. class 6 and class 10 cards FAIL with "B" settings and Native 24p enabled while camera is NOT moving.

17. class 6 and class 10 cards FAIL with "C" settings and Native 24p enabled while camera is NOT moving.

18. class 6 and 10 cards FAIL with these settings:

native 24p enabled
64M
72M
80M
Camera NOT moving

19. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=12 while camera is NOT moving.
20. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=13 while camera is NOT moving.
21. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=14 while camera is NOT moving.
22. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=15 while camera is NOT moving.
23. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=16 while camera is NOT moving.
24. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=17 while camera is NOT moving.
25. class 6 and class 10 cards FAIL with stock firmware bitrate settings with Native 24p enabled and GOP=18 while camera is NOT moving.

As you can see, it's related to the 24p patch and having a perfectly static scene. And YES, I understand that a Sandisk card would essentially make this work, but that's not the point. The point here is that bitrates clearly within the Class 6 and Class 10 card's ratings are NOT working with the 24p patch enabled and a static scene yet they DO work while the camera is moving. Even holding it by hand is enough to make it work.

 

[Kellar42]

Re: Pappas, Do you have toast? 'Show stream info' as you playback in Toast Video player will give you real time bit-rate information on .mts files.

 

[svecher]

How can I find out the bit rate of my settings?

In my MJPEG footage., it was easy with QT.

Download 30-day trial of Elecard StreamEye at http://www.elecard.com/download/ . It's the best tool we have ATM.

 

[Mike@AF]

Well it's not equal and never can be because enabling or disabling Native 24p makes things unequal.

Everything else is equal.

Here is my exact sequence:

1. camera set to 1/50( also tested 60,125), 200 ISO, FHD.
2. camera set on tripod, focused at grass, NO movement, exposure set to 0.
3. class 6 and class 10 cards record normally with stock firmware.
4. class 6 and class 10 cards record normally with only "A" bitrate settings.
5. class 6 and class 10 cards record normally with only "B" settings.
6. class 6 and class 10 cards record normally with only "C" settings.
7. class 6 card fails and class 10 PASSES with these settings verifying that bitrate is indeed higher than Class 6 card can handle:

no native 24p
64M
72M
80M

This is confusing. In 3-6 you say Class 6 and Class 10 record normally. But then in #7 you say Class 6 fails and Class 10 passes. This seems to be contradictory.

EDIT: Nevermind. I think you mean you record normally with movement, not that the camera actually records. Is that right?

 

[Svart]

This is confusing. In 3-6 you say Class 6 and Class 10 record normally. But then in #7 you say Class 6 fails and Class 10 passes. This seems to be contradictory.

EDIT: Nevermind. I think you mean you record normally with movement, not that the camera actually records. Is that right?

What I mean is that I intentionally set the bitrate high enough to cause the class 6 card to give me a "true" card speed error in order to test that the bitrate changes do indeed give us the bitrate we ask for.

The class 6 can is only rated for ~48mbit/s. By trying to cram it with 64-80mbits I know that it should fail which also proves that the 24p patch is somehow giving us spikes of data beyond 48mbits. Because the class 10 card fails as well with 24p enabled means that the bitrate spikes must be higher than ~80mbits. Because the Sandisk cards can take 240mbits and they do NOT fail that means the actual bitrate spike is somewhere between ~80mbits and ~240mbits.

I edited what I wrote to hopefully make it a little clearer.

 

[svecher]

I am shooting is perfectly still. That's when I get the card speed error.

This is very interesting, Svart. I took Pappas' advice and shot a full-screen Star Chart (http://www.bealecorner.org/red/test-patterns/star-chart-bars144-600dpi.png) with 20mm lens and GH1 mounted on a tripod with the following settings:
60000000
62000000
70000000

And got the famous "Motion Recording Stopped due to Card Speed" error within a couple of second (I've never seen it before). StreamEye reports 57 Mbit/s avg bitrate on such footage. Interestingly, when shooting the chart hand-held (with associated micro-movement) the camera will keep on recording. Shouldn't static scenes have lower bitrates "theoretically" as there is no movement in the image? So there seem to be two issues going on here:
1) Enabling the native patch causes camera to freeze up at bitrates dependent upon the card used and scene composition.
2) At high bitrate settings and when shooting static scenes the bitrate seems to shoot through the roof causing cancellation of recording, but not freezing the camera.

 

[Svart]

This is very interesting, Svart. I took Pappas' advice and shot a full-screen Star Chart (http://www.bealecorner.org/red/test-patterns/star-chart-bars144-600dpi.png) with 20mm lens and GH1 mounted on a tripod with the following settings:
60000000
62000000
70000000

And got the famous "Motion Recording Stopped due to Card Speed" error within a couple of second (I've never seen it before). StreamEye reports 57 Mbit/s avg bitrate on such footage. Interestingly, when shooting the chart hand-held (with associated micro-movement) the camera will keep on recording. Shouldn't static scenes have lower bitrates "theoretically" as there is no movement in the image? So there seem to be two issues going on here:
1) Enabling the native patch causes camera to freeze up at bitrates dependent upon the card used and scene composition.
2) At high bitrate settings and when shooting static scenes the bitrate seems to shoot through the roof causing cancellation of recording, but not freezing the camera.

Bingo!

Yes that is exactly what I am seeing. Handheld= OK. Tripod= fail.

That's why I think people are believing that it's the cards or camera, because this failure is exactly the OPPOSITE of what you would think *should* be happening. It would make no sense whatsoever if you haven't seen it yourself.

Try lower bitrate settings, maybe even STOCK settings, and see if you can recreate my failures.

If someone else can recreate my failures on a different camera, then we know it's not MY camera. If anyone can recreate the failures with different cards then we know it's not my cards and it's a legit problem.


EDIT: Svecher, I just did the same experiment with the chart. I installed a firmware with Native 24p Enabled and then I setup the camera in front of my monitor so that the chart filled the camera's VF. I hit record and 7 seconds later I get the speed error!

EXCELLENT. This means we now have a chart that we can all use to test the same way!

 

[PappasArts]

Re: Pappas, Do you have toast? 'Show stream info' as you playback in Toast Video player will give you real time bit-rate information on .mts files.

Hey Stuart

No I don't have toast. Thanks for the suggestion though.

Download 30-day trial of Elecard StreamEye at http://www.elecard.com/download/ . It's the best tool we have ATM.

Hey svecher,

Thanks, I'll check it out.


VLC reports bit rates data, however is that reliable info?



Pappas

 

[Lpowell]

1080p GOP should be deselected- Keep it unchecked.

Did you mess with audio bit rate?

Deselect any patches to 720P. Focus on 1080P.

Format both cards and run the test again. All manual mode. No auto focus or stabilizer- Nothing. Straight manual. Lets see what happens.

Barry tried my settings, they worked for him as well. I have shot tons of stuff. So there's a cog in the wheel somewhere that's causing the issue- lets find it.

No, I didn't change audio bitrate or GOP length when testing your high bitrate AVCHD settings. And since I wanted to see if it worked without write errors on class 6 and class 10 SD cards, I used an auto-focus + image stabilizing Lumix lens on a highly detailed panning shot as a worst-case test.

Sorry, but my testing priorities are somewhat different than yours. I want to find the highest reliable AVCHD bitrate settings that can be used with any lens or subject matter. In addition, I want to film with 1080p in native 24p and 720p at 30fps in order to streamline my post-processing workflow.

GOP length is a more subtle issue, as it affects image quality only during panning or with moving subjects. I think StreamEye can be helpful in evaluating GOP length. What it shows me is that with slow panning, I-frames are allocated about twice as many bits as each P-frame. But with fast panning, P-frame size increases, and I-frame size drops to about the same as a P-frame. I suspect this is because the AVCHD encoder has maxed out its bit allocation on the P-frames and has to cut back on I-frame quality.

What I think would be undesireable is if P-frame bit allocation consistently exceeded the size of the I-frame. That would indicate that it actually required more bits to encode the difference between the P-frame and the I-frame than it would take to encode the P-frame as an I-frame. Unfortunately, it appears that the encoder is unable to dynamically insert I-frames as needed, but is instead locked into the specified GOP length.

Given that restriction on the encoder, I think the best policy is to find the shortest reliable GOP length that does not cause excessive P-frame bit allocation. That should enable the encoder to handle fast motion and panning with less image degradation than with a long GOP length.

With GOP length set at 15 for 30fps and 12 for 24fps, each GOP is 0.5 seconds long. That seems pretty reasonable and so far, I haven't seen any excessive P-frame sizes in test clips I examined in StreamEye.

 

[PappasArts]

Bingo!

EXCELLENT. This means we now have a chart that we can all use to test the same way!

I wish I mentioned this before today. I've been using this testing method for years.


Very interesting thing, the only error I have ever had was when it was on a tripod. It was my MJpeg vs AVCHD. The same shot stopped on both AVCHD and MJPEG...

I told you; ""We'll get this figured out...""


Pappas