GH2 AVCHD spanning

sohus

Spanning is discussed a lot on these forums. It probably doesn't matter that much if you are a narrative filmmaker, but it is very problematic if you shoot events, documentaries, interviews, weddings where you can't have the camera stop recording. These people still want to shoot at the best possible settings, which is not really possible at this moment. I think the goal should be a 24H and 24L settings (in one set) with 24H not spanning but shooting higher quality and 24L spanning reliably (ideally all settings would span, but I don't think that is possible). Spanning is also very important if you shoot dual system audio as it safes a lot of time syncing clips.

What we know:

@KAE's 65M settings spanned (PTool 3.61)

After the release of PTool 3.62, spanning became more of an issue (at least it appeared to). Does this have something to do with the AQ settings that were opened up?

With PTool 3.62D, I have confirmed the following settings to be spanning:
@CBrandin 44M/32M at 24L setting (32M) --> GOP12, 32M, AQ4 (44M setting is not spanning)
Stock settings

Let me know if you know other settings with good quality and reliability that do span.

These two examples are extremes in terms of bitrate so I am pretty sure the bitrate alone is not the sole cause.

Driftwood's setting has no AQ set and the iFrames are smaller than most other settings (around 6MB instead of 8 - 11 MB) for most other settings. It also has no P & B frames (does the codec still need to work as hard?). Furthermore, it is GOP1 instead of GOP 3/6/12/ It would be nice if we could pin down what affects spanning: bitrate, AQ settings, GOP settings, memory card speed, scene detail (extreme: death chart). What makes that KAE's setting spanned with 3.61 but @CBrandin's 66M setting does not span? Is there someone who has an idea about this?

Ideally, I hope we can create a setting (ideally a 24L setting ) that spans at a decent bitrate (say 44M - 66M - 88M).

Looking forward to your thoughts and experiences on this matter.

driftwood

@sohus Its either keeping the data below a certain overall bitrate threshold (like you talk about i frames) or possibly a combination of Encoder settings 2, 3 & 4 per resolution (1080/1080i/720)

sohus

Rectification:

@Driftwood 240M at 24H setting --> GOP1, 240M, no AQ
@Driftwood 176M at 24H,L setting --> GOP1, 176M, no AQ

Are not spanning on a death chart (Transcend Ultimate Class 10 32GB, 19mb/sec effective)

What do these encoder settings do?

driftwood

@sohus Spanning is reliant on how much bitrate is being used - if there's more detail = more bitrate/bandwidth in use, therefore over a certain amount of bitrate (we think 32M at the time of spanning) the encoder no longer spans. If its low detail then not much bitrate is being used and it should span! Remember the codec is VBR (variable bitrate)

We're still learning about Encoder settings - it may but possibly not have someething to do with spanning or controlling the bitrate... However, we do know that Encoder setting 1080p/i 1 =2 gets rid of the P frame in GOP1.

bartolomeo

it's rare for me not to find an answers to my questions through a search engine, but somehow it's to specific this time.

What is a spanning problem?
thanks

JDN

@bartolomeo Spanning simply means that when you reach approx 4 gigabytes in one clip and new clip (.mts file) is created which seamlessly continues where the last clip left off (i.e. you could run the camera for an hour and get a bunch of 4 gig .mts files that are continuous). With ptools 3.62 high bit rate settings, when you reach 4 gigs the camera does not span anymore, it just stops recording.

proaudio4

I'm not convinced that v3.61 did not have spanning issues.
I do not believe we were tasking it with these higher bitrates like we have in 3.62.

bartolomeo

@jdn
alright thanks.

JDN

@proaudio I think as someone mentioned earlier, they key question for me is why settings like kae's 66mb 3GOP extreme settings had no problems spanning yet 66mb settings in ptools 3.62 do. Maybe I'm wrong, but wouldn't kae's extreme settings be over the theoretical 35mb limit? I rolled a 40 minute interview with high bit rate throughout (scientist standing in plants above his heads which were blowing the entire interview) and had no issues with spanning using kae's settings. So something seems strange to me....

paglez

There is another settings with no spanning problem for me at 66 mbits: Driftwood's beautiful buffers in L Mode. You only have to change the original Video Bitrate 24L=110000000 to 66000000. I shot a 42 minutes with no interruptions. No static recordings, but an outside recording in my garden on a windy day.

sohus

@paglez

Do never do that!

I hoped that you were right and all was fine but I never do something without testing. So I tested Driftwood at 24L 66M to see if you were right. YES, this setting is spanning. But I guess you never looked at the stream, as you get a few seconds of good frames and than everything collapses. After a few seconds, you get 400K iFrames and even smaller pframes. This makes the bitrate/cadence/quality worse than stock.

You are better off with @cbrandin 44m/32m at 24L setting if you want spanning.

The thing that changed between 3.61 and 3.62 are the AQ settings. And those add detail = bitrate. No rocket science. I hope spanning can be fixed by I don't have much faith that it can be, unless there is some magical buffer setting that is not unlocked yet so the bitrate at spanning point can be higher.

DrDave

There is spanning, then there is always spanning. I spent several weeks developing a 24p patch for my GH1 (the "event" patch) and I found that even when I thought I had "the one", every now and then the spanning would fail. I haven't had the current patch fail (indoors) but the bitrate comes in around 23. Absent some magic setting, I agree the bitrate is the key.
However, I didn't see a big IQ difference indoors, and, as others have noted, at a certain point you may get more noise for whatever reason. The way I tested the patch, and I'm sure someone can some up with a better one, was to shoot a highly detailed scene using high CRI lights and an F Stop of 5.6 on the 20mm lens (the 20mm has very good edge detail). The scene had multiple shadows so I could monitor the noise. The patch was said to "span" if it passed this test 20 times. However, who knows? If recording a wedding, perhaps the stock firmware on the GH2 is the way to go. Sorry, newly wed ones, my card failed. Could you get married again? Sometimes reliability trumps quality. I have a very expensive preamp, and I don't use it for live recording. Why? Once in a blue moon it will pick up RF or other electronic junk. Maybe for one minute, maybe once a year. But I'd rather not worry about it.

JDN

@DrDave, indeed, good words of caution.

@sohus, would you wager @cbrandin's settings are better than kaes 66mb setting because of the AQ advances 3.62? Just curious because I want to stick a safe spanning setting on a 2 gig card and just have it in the wallet if needed. Kae's has never failed me, but haven't tried cbrandin's yet... does the AQ make up for the lower bit rate?

DrDave

In comparing two patches, rather than looking at the numbers, devise a test that will crash one but not the other, and then you will have a real world answer. Take a typical scene that you shoot and use as sharp a lens as possible--with edge sharpness--then bump up the F stop beyond what you normally use and see which one works better. Just as important, see which one looks better, or grades better.

sohus

I would like to setup the following test:

Take @cbrandin 44M-32M setting, set camera to 24L and record on typical Class 10 card (19mb/sec write):
1. Record high detail scene
2. Analyze bitrate stats
3. Up the 24L bitrate to 33,34,35,36 until spanning breaks

Restart test with SanDisk Extreme 30mb/sec card
1. Record high detail scene
2. Analyze bitrate stats
3. Up the 24L bitrate to 33,34,35,36 until spanning breaks

See if there is a difference between maximum spanning bitrate between two cards. If there is, memory card write speed has influence on spanning. If both are the same, the bitrate is what counts.

I don't own Sandisk cards so someone who owns both SanDisk and Transcend or other similar card might be able to do this test.

JDN

This is a good test. I'm shooting for a few days but if someone hasn't jumped on it by then I'll give it a go.

Alfi666

Hello,

I'm having a problem with spanning as well (at least I believe it is spanning issue).

I have a PAL version of GH2, all I have done is that I've removed the 30min limit with the hack. I have not changed the bitrate, nor anything else.
I'm powering the GH2 externally, to be sure there is no battery issue.

In 50i mode:
I'm able to record more than 30min (for which I'm very thankful) but it randomly stops recording after creating the spanning file.
A 4gig file can take up about 35mins @17mbit default. First spanning worked for me till now (so I get pass the 35min limit of 1 file), but it randomly stops recording after 2nd/3rd span (about 1:10min or 1:43).

In 24p mode:
There is no problem at all. I'm recording in 24H mode for about 3:08mins on 32gig card without any issue. (tested twice)

Do you have any idea how to avoid spanning problem (or whatever it is) in 50i mode?

Thank you :)

Braamokiev

Hello

@Drifwood please where is the link for donload a PTool 3.62D?

Thanks

raizman

Do the Ptools 3.62 recommended safe settings for 24H (I think it's 44M) have any issues with spanning on class 6 cards?

mpgxsvcd

So if spanning is important to you, how important is amount of time you can record? For instance, even if we could get the 176 mb/sec settings to span you would only get 15-20 minutes on a 32 gb card. =

Some people complain about the 29 minute time restriction. I would think that 15-20 minutes would not be that beneficial even if it spans.

I think ideally we need to find a setting that spans for exactly the time the battery lasts. That ends up being 32 mb/sec. It will record for 2 hours and 9 minutes on a 32 GB card and it will span perfectly.

I have modified the “No adverse affects” settings to use 42 mb/sec for the high setting and 32 mb/sec for the Low setting for 1080p. These setting will have no write errors for any card. In addition the 1080p Low, 720p High, and 720p Low settings will all span.

Using an extremely high bit rate and worrying about file spanning will contradict each other because with high bit rates you will run out of storage before you reach the capacity of the card. Therefore, the 1080p High setting is for high bit rate. The 1080p Low setting is for Spanning.

arvidtp

I need to go down to 26M in 24p or 20M in 720p60 to get spanning to work 100% reliably on a what i think is a mediocre class 10 card (patriot) with current (3.62) ptool release. See this post: http://personal-view.com/talks/discussion/comment/14894#Comment_14894 Not sure about class 6.

mpgxsvcd

File spanning works with the settings I posted for 720p with a Class 6 card and 32 mb/sec. I need to do more testing with the 1080p setting to confirm that all cards will span with the 32 mb/sec setting.

cyfernoa

does spanning work for 1080 60i at 32mb/sec? I have not seen anyone talk about this yet. I have a gig in a couple of hours that I do not have time to test for but I do need spanning as it is a show.

arvidtp

If you don't have time to test yourself, use stock firmware this time - just my 2 cents.

azza_act

Heh guys, I'm hack noob shooting a wedding soon and i need a patch that i can span with, so i went with crb 44/32 settings. Did i little span test of shooting myself on the tread mill (Shooting 1080 24L, with the Panny 20 .17 (Manual focus), Sandisk Class 10 32g), to see if everything is fine. Only issue with this test, is that the wind shut the fucking door where all the light was coming for. So its under exposed. If you look at the picture, below my knees, you'll see some crap. This is a the start of the second 4gig file, when its on the time line (Pre CS5) Is this the codec freaking out because its to dark, i'm i worrying about nothing or does this happen in 24L 32mbps when you join spanned files together? Don't have time to do another test to see... Cheers

arvidtp

weird things can happen between the spanned files if your editing software is not made for that because the previous keyframe will be in the previous file and the NLE doesn't know that they are basically one continuous file. It has no way of knowing. What i do when i'm managing imports manually is use the unix command cat (i'm on mac) on my spanned files to literally concatenate them into one big file. Then the NLE should be fine with it (ie Final Cut Express no longer has glitches). There should be a similar command in windows command prompt i think.

As far as high bitrate spanning, try the SpanMyBitchUp (sp?) setting found in the GH2 settings vault thread. Should span on most class 10 SD cards on 24H. (caution: Will often not span on 24H 80% speed)

Also, spanning often depends on complexity of the image. sometimes it'll span once if there is little motion in the frame at that time, then die on the 2nd file if there is more motion or more noise then.

But if you really really need it to just work and don't have time to test, save yourself the headache and use stock firmware then experiment later what you have time. Shooting trees blowing in the wind out the window for hours on end with an AC adapter is my method for determining if it will reliably span.

Howdy, Stranger!

Categories

Tags in Topic

Top Posters