3.37dc- Everything works except 720p SH

[modulr]

Pardon the fresh face (professional lurker)... I've searched, but could not find an answer. I'm using PTool 3.37d. I've got it set up as follows...


Settings
Native 24p Y

FHS/SH Video Bitrate 50
H Video Bitrate 32
L Video Bitrate 22
Overall Bitrate 52
Limiting Bitrate 60

720p60 GOP Size 15
1080p24 GOP Size 15

720p 59.94 -> 29.97 fps N

MJPEG E1 Quality 300
MJPEG E1 Table 24
MJPEG E2 Quality 300
MJPEG E2 Table 24
MJPEG E3 Quality 300
MJPEG E3 Table 24
MJPEG E4 Quality 300
MJPEG E4 Table 24


When I shoot the ol' star chart I get the following results...

Measured Bitrate

FHD 45.0Mbps
SH 12.9Mbps <<<---- Why?
H 29.9Mbps
L 20.7Mbps
MJPEG 77.6Mbps


I've tried different values for FHD/SH Video Bitrate, but nothing seem to effect the SH setting. Kinda strange.

Also, has anyone had issues where attempting to update the firmware fails; it then telling you to turn the camera off then on again. I've had it fail consecutive times before finally taking... but no rhyme or reason why it worked. For a minute I thought I "bricked" the camera. So far I've tested roughly 30 different firmware modifications.. but I started getting nervous that the rewrites might have glitched somewhere on me.

Preserve my sanity! lol Comments welcome...

 

[Petanque]

Jobless has warned that shooting below 1/50 in 720p with bitrate adjustments will result in severely reduced quality. Any chance this is the case here?

 

[modulr]

it's definitely possible I may have missed that... contrary to what seems the norm, I like using 1/30 shutter.. I forgot that bit of information. Odd that H and L work fine tho. Hmmm, perplexed.

 

[wturber]

FHD 45.0Mbps
SH 12.9Mbps <<<---- Why?
H 29.9Mbps
L 20.7Mbps
MJPEG 77.6Mbps

I'm betting it is mostly due to the 30fps frame rate. Check to see if the H and L video is running at 60fps or not.

Oops! I misread your list and thought you had changed the 59.94 to 29.97 for 720p. Not sure why you would add an unchecked option to your list. So my bet above is no longer in effect. Did you shoot the same subject in all cases? If not, that could be the cause.

BTW, I'm also betting that everything is not really working.

I'm betting you can't play back MJEG in the camera.
I'm betting that you can get freezing in 1080 24p mode if you shoot a high-detail static shot.
I'm betting that you get visual pulsing with the AVCHD modes.
I'm betting that you get posterization/banding with the AVCHD modes.
I'm betting the MJPEG clip length predictions aren't quite right.

I haven't confirmed it, but I think I've observed the camera not wanting to update firmware if it was in power save mode prior to trying to update the firmware. I powered off and back on and the update procedure worked. At least I think that was the sequence. I wasn't paying particularly close attention when that happened.

 

[modulr]

Ok, so I went back and checked... all 720p footage is 59.94, all shot at 1/30th. So it's still stands as a bit strange. Any other guesses?

 

[modulr]

I'm betting it is mostly due to the 30fps frame rate. Check to see if the H and L video is running at 60fps or not.

Oops! I misread your list and thought you had changed the 59.94 to 29.97 for 720p. Not sure why you would add an unchecked option to your list. So my bet above is no longer in effect. Did you shoot the same subject in all cases? If not, that could be the cause.

BTW, I'm also betting that everything is not really working.

I'm betting you can't play back MJEG in the camera.
I'm betting that you can get freezing in 1080 24p mode if you shoot a high-detail static shot.
I'm betting that you get visual pulsing with the AVCHD modes.
I'm betting that you get posterization/banding with the AVCHD modes.
I'm betting the MJPEG clip length predictions aren't quite right.

I haven't confirmed it, but I think I've observed the camera not wanting to update firmware if it was in power save mode prior to trying to update the firmware. I powered off and back on and the update procedure worked. At least I think that was the sequence. I wasn't paying particularly close attention when that happened.

So far, I have the 59.94->29.97 listed because it's something I wanted to experiment with down the road. I have a spread sheet that I enter all the variations in, as well as the results. I just haven't got to messing around with that option yet. I'm shooting the star chart in each setting for consistency and to run up the bitrate.. as well as outdoors in a variety of settings.

So far MJPEG is my favorite. It doesn't playback in the cam, but the footage is beautiful. I'm just starting to explore the AVCHD stuff, which is why this SH setting being so low has me puzzled.

 

[svecher]

Ok, so I went back and checked... all 720p footage is 59.94, all shot at 1/30th. So it's still stands as a bit strange. Any other guesses?

It doesn't make sense to run shutter at speeds slower than fps value. The recommended shutter speed for 60 fps is 1/125. Try to repeat tests at that speed. I tested FHD/SH/H/L settings at progressively lower bitrate settings, which reflected in the footage.

 

[modulr]

It doesn't make sense to run shutter at speeds slower than fps value. The recommended shutter speed for 60 fps is 1/125. Try to repeat tests at that speed. I tested FHD/SH/H/L settings at progressively lower bitrate settings, which reflected in the footage.

Since we are essentially talking about sampling from a sensor electronically... 1/30 is about the lowest you can go with no strobing but loads of great motion blur. 60fps at 1/30 looks no different than 30fps 1/30. Most stuff over 1/60 is too crisp for my taste, unless the intention is to slow 'mo the footage. I usually stick to 1/30, 1/40, 1/60.

BTW I continued testing, and switched the 59.94->29.97 option. I got results that do not correspond to the setting descriptors... for example.


59.94->29.97 checked

FHD/SH - 64
H - 56
L - 48
Overall - 68
Limiting -72

GOP - 6

SH - 36.2Mbps
H - 38.6Mbps
L - 36.2Mbps


All 720p modes where nearly identical. Hmmmm, strange no?

 

[svecher]

Since we are essentially talking about sampling from a sensor electronically... 1/30 is about the lowest you can go with no strobing but loads of great motion blur. 60fps at 1/30 looks no different than 30fps 1/30.

Well even almighty electronics can't make 60 real samples per second when shutter is "opened" [electronically] for only 30 times a second. The reason 60fps at 1/30 looks no different than 30fps at 1/30 is because it is 30 "real" fps. The file may have 60 frames per second recorded, but every 2nd frame is a copy. Add to this some "motion" blur and you have a low-detail 720p @ 30 fps -- I'm not surprised bitrate drops down to 12 Mbps.

Most stuff over 1/60 is too crisp for my taste, unless the intention is to slow 'mo the footage. I usually stick to 1/30, 1/40, 1/60.

To get "native" 60 fps you would need to shoot at 1/60 at least. Slower shutter speed would make sense in FHD mode or with FPS adjustment patches to 30 fps like you did.

BTW I continued testing, and switched the 59.94->29.97 option. I got results that do not correspond to the setting descriptors... for example.


FHD/SH - 64
H - 56
L - 48
Overall - 68
Limiting -72

GOP - 6

SH - 36.2
H - 38.6
L - 36.2


All 720p modes where nearly identical. Hmmmm, strange no?

When I did testing I saw a couple of factors at play: first and foremost the amount of detail in scene (for testing it is best to settle on something more or less constant, like star chart, which is very high detail http://www.bealecorner.org/red/test-patterns/star-chart-bars-full-600dpi.png). The second factor is how static/dynamic is the scene. Perhaps contrary to "common sense", static scenes are more taxing on AVCHD encoder than dynamic ones. This explains why bitrates you observe are not as high as values you've set in PTool. The question still remains as to why there is no bitrate gradation according to settings, which very well might come down to differences in details and motion between takes for different modes. Since you get an average of 38 MBit rate, my bet is that if you try settings like:
FHD/SH: 32
H: 26
L: 20
Overall: 34
Limiting: 38

You will see gradual decrease in avg bitrates as you switch to lower modes.

Also, in my limited testing of GOP settings I found that adjusting them leads to unpredictable results, so it is best (IMHO) to leave them alone.

 

[modulr]

Well even almighty electronics can't make 60 real samples per second when shutter is "opened" [electronically] for only 30 times a second. The reason 60fps at 1/30 looks no different than 30fps at 1/30 is because it is 30 "real" fps. The file may have 60 frames per second recorded, but every 2nd frame is a copy. Add to this some "motion" blur and you have a low-detail 720p @ 30 fps -- I'm not surprised bitrate drops down to 12 Mbps.

Yes, that is accurate; however, it does not explain the bit rate phenomenon.

Original settings...
FHS/SH Video Bitrate - 50
H Video Bitrate - 32
L Video Bitrate - 22
Overall Bitrate - 52
Limiting Bitrate - 60

SH - 12.9Mbps
H - 29.9Mbps
L - 20.7Mbps

SH is 50, but giving 12.9Mbps @ 59.94...
H is 32, but resulting in 29.9Mbps @ 59.94...
etc

SH should have the highest bit rate, and the others should cascade lower. I was shooting the exact same card, same shutter, same fps (59.94), same aperture. I actually found another post where someone had the same results. SH being low, but H and L being higher.

When I did testing I saw a couple of factors at play: first and foremost the amount of detail in scene (for testing it is best to settle on something more or less constant, like star chart, which is very high detail http://www.bealecorner.org/red/test-patterns/star-chart-bars-full-600dpi.png). The second factor is how static/dynamic is the scene. Perhaps contrary to "common sense", static scenes are more taxing on AVCHD encoder than dynamic ones. This explains why bitrates you observe are not as high as values you've set in PTool.

My very first post explains I've been shooting the star chart this entire time.

The question still remains as to why there is no bitrate gradation according to settings, which very well might come down to differences in details and motion between takes for different modes. Since you get an average of 38 MBit rate, my bet is that if you try settings like:
FHD/SH: 32
H: 26
L: 20
Overall: 34
Limiting: 38

You will see gradual decrease in avg bitrates as you switch to lower modes.

Not to sound unappreciative for the attempted assistance, but did you actually read my first post before you responded?:engel017:

Also, in my limited testing of GOP settings I found that adjusting them leads to unpredictable results, so it is best (IMHO) to leave them alone.

That I'll give a try.. setting all the gop stuff back to defaults.

 

[Barry_Green]

It The recommended shutter speed for 60 fps is 1/125. Try to repeat tests at that speed.

ONLY if you're wanting to use that footage for slow-mo in a 30p or 24p project!

If you're shooting 720/60p for the "live" look, such as for sports or news, you definitely do not want 1/125, you want 1/60th.

 

[svecher]

SH should have the highest bit rate, and the others should cascade lower. I was shooting the exact same card, same shutter, same fps (59.94), same aperture. I actually found another post where someone had the same results. SH being low, but H and L being higher.

That is quirky indeed. I'm curious as to what happens with default GOP settings.

My very first post explains I've been shooting the star chart this entire time.
Not to sound unappreciative for the attempted assistance, but did you actually read my first post before you responded?:engel017:

I did, but apparently not carefully enough.

 

[svecher]

ONLY if you're wanting to use that footage for slow-mo in a 30p or 24p project!

If you're shooting 720/60p for the "live" look, such as for sports or news, you definitely do not want 1/125, you want 1/60th.

Thanks for clarifying that, Barry.

 

[wturber]

I ran your (modulr) settings last night but kept my GOP settings at 10 for 720p and 12 for 1080p. I shot the start chart on my laptop screen and got 82 Mbs with MJPEG. I did not see a dip at SH on the AVCHD. Or at least not one of any significance. The SH and H both recorded at about the same bit rate - in the thirties if I recall correctly, and the FHD recorded at a substantially higher rate. I think I shot at 1/60 or faster. Sorry I don't have more specifics, but I was tired and crashed afterwards. I'll recheck tonight.

 

[modulr]

Thanks for testing that out... I've been in the lab a bit more here, decided to get more clinical about this... heh.

I've reinstalled fresh default firmware, then jumped into ptool and adjusted only the following FHD/SH Bitrate, H Bitrate, L Bitrate, Overall Bitrate, Limiting Bitrate... leaving everything else unticked. Test #1 is the default settings, while for Test #2 and #3 I preserved the ratio between the values.

so...

Limiting = maximum bitrate.
Overall = .9 x Limiting
FHD/SH = .8 x Limiting
H = .6 x Limiting
L = .4 x Limiting

Hopefully this will display right...

Settings are 59.94fps, 1/30th, Nostalgic (0, -2, 0, -2), ISO 100, star chart... blah blah

TEST #			1	2	3

FHS/SH Bitrate		16	32	40
H Bitrate		12	24	30
L Bitrate		8	16	20
Overall Bitrate		18	36	45
Limiting Bitrate	20	40	50

FHD Mbps		15.2	30	37.5
SH Mbps			15.2	30.1	3.6
H Mbps			12.3	22.4	28
L Mbps			8.2	15.2	18.8

What's interesting is that in test #3, the bitrate for SH drops dramatically. FHD @ 37.5, while SH @ 3.6 ... makes me wonder if there is some sort of mathematical error in the firmware with certain settings, cos the decimal for SH is in the wrong place... heh, should be 36 not 3.6! Hmmmm, interesting... more testing.

[edit: posted the wrong default numbers... fixed]

 

[wturber]

Interesting. But why still testing at 1/30th? Maybe consider a run of the SH mode at different shutter speeds to see if it has an effect?

 

[modulr]

Interesting. But why still testing at 1/30th? Maybe consider a run of the SH mode at different shutter speeds to see if it has an effect?

Just tried this, and everything below a 1/60th on the AVCHD 720p setting potentially generates odd results when the video bit rates for SH/H/L step beyond a certain value. Where that threshold is I'm not sure yet.

Like...

FHD/SH - 50
H - 50
L - 50
Overall - 51
Limit - 52

with anything below 1/60th shutter will yield something like...

FHD - 46.7
SH - 12.9
H - 12.9
L - 12.9

 

[churasan]

SH should have the highest bit rate, and the others should cascade lower. I was shooting the exact same card, same shutter, same fps (59.94), same aperture. I actually found another post where someone had the same results. SH being low, but H and L being higher.

me too.
in my case, SH and H bave almost the same bitrate.
when shooting below ss1/60, bitrate has no change from default factory setting.
so as for now, I decided to use MJPEG mode when I want to shoot under ss1/60.

thogh I also want to know why SH ( and H ) mode behave like that.