Personal View site logo
Noise on GH2 footage at ISO 160
  • I´ve shot an interview for a TV series, on tripod, day, overcast, plenty of light. Anything looked ok on the LCD while shooting BUT, when we downloaded the file and imported on Final Cut we found a dreadful noise was all over the ground and on some portions of the character being interviewed. Then we went to check the original file on VLC2 and the noise is not so bad there but quite hard to swallow anyway. Any clues of what happened? THANKS IN ADVANCE

    Original MTS file can be downloaded on

    HACK Driftwood´s spanmybitchup lens Panasonic, 14-45mm at 14mm, aprox f16 Settings were: ISO 160 WB interior shutter 1/200 HBR nostalgic video out pal

  • 51 Replies sorted by
  • @GuilleKloetzer Do a search for "ISO bug". Sounds like that may have something to do with your noise issue @ ISO 160.

    F16 @ 1/200th? You didn't have a ND?

  • @GuilleKloetzer I just looked at it on my laptop and didn't see much noise. I looked at it with splash and VLC.

  • Yes, I see it - very bad compression noise in dark gray areas and a shimmering on the distant rocks. I never tested spanmybitchup, but this looks like a patch problem. I would suggest trying a different patch.

  • As @Ralph_B suspected, the Intra-only patch that recorded the downloaded video is dysfunctional. The video was recorded in HBR mode at 1080p25, which encodes each Intra frame as a pair of interleaved fields (an I-field interleaved with a P-field). Using Elecard StreamEye Slim to inspect the macroblock encoding, the average quality level deteriorates dramatically as the encoder chronically runs out of bitrate in the bottom half of each frame. (Technically speaking, the QP level ranges from 13-19 in the upper half of each frame, and is degraded down to 25-39 in the lower half.) Inconsistent quality performance of this nature indicates that the patch configured the encoder to underestimate the amount of bitrate needed to properly encode each frame.

  • @LPowell Will a decent on cam monitor or aftermarket VF indicate this issue? Or do you need to review the recorded footage to spot it?

  • @brianluce I don't think this kind of problem is easy to spot with the naked eye, and certainly not on a small-screen monitor. The best way to prevent these kinds of issues is test a patch thoroughly before you use it on a shoot.

  • Hi, thanks to everybody for the replies. In order of appearance: pundit, I didn´t find anything on the ISO bug topic related with ISO 160, have you? I had a ND, but choose to shoot with shutter since the subject was sitting and the shutter would´t be very noticeable avoiding the filter meant less glass so resolution should be better. Peternap, look at it on a monitor at full screen, look on the ground at the grey and black areas, look at the throusers of the interviewed. Ralph_B, thanks for that. LPowell Many thanks for taking your time for analyzing this in detail. What you found renders this patch unusable in my opinion. I´ve tested it in many different situations before using it on a professional shooting, but I think most times It´s impossible to anticipate and test exactly how the frame will be until you are there. So I will just never use this patch again. I will investigate on this forum which patch can span safely, any suggestions? Brianluce, maybe you can spot this on a onboard monitor, what you need is size. In the tiny monitor on the GH2 it´s impossible for me to spot it. If you want to test it maybe you can load the MTS file on a SD card at the stream bin, and then play it on the GH2. Thanks everybody again for making this forum great!

  • @GuilleKloetzer If you need 4GB file-spanning in PAL HBR mode, which patch you can use depends on the type of SD card you have. With a 45MB/sec Class 10 card or slower, I'd recommend trying cbrandin's standard 44Mbps patch. With a 95MB/sec Class 10 card, I have tested Flow Motion v2.02 to span reliably in PAL HBR 1080p25 mode.

  • Another thing to consider is; higher apertures = more diffraction = less detail and more noise. (this issue has been well covered elsewhere in the forum!) The difference between say f5.6 (ND'd) and f16 at the same iso is substantial. For IQ reasons it does not generally make any sense to go above f5.6 - f8 depending on lens and desired edge sharpness when shooting with a m43 sensor.

    Didn't check the clip but I saw the above settings. Gop1 / high bitrates, IMO is primarily for 24p and cards that perform very well with the gh2. (good: 30mb/s sandisk, best: 95mb/s 64gb sandisk)

    I personally avoid the HBR mode; for documentation type works 50i is smoother to convert and when I want maximum IQ then I also want control over (most) lights - to avoid any potential flickering...

  • OK thanks LPowell, I usually work with a class 10 95MB/s 32 GB, I´ll try cbrandin´s standard 44 and Flow Motion, reliability is the priority on the kind of job I am doing currently, so, maybe better to go cbrandin´s anyway. I´ve read good reviews on Sanity 4.1´s reliability also. And thanks too RRRR, I was missing that diffraction tip. I know every lens has it sweet-spot at a certain f-stop and normally it´s not too closed nor too open, somewhere in middle apertures. I don´t understand why HBR does not deliver the better IQ possible in this camera, as I thought it was the higher bit rate mode... ¿?¿? Why am I wrong? Anyway on the noise cause on this clips LPowell explanation seems the most accurate to me, since if it was only a diffraction problem it should be on the details all over the frame, but it is not, It is only on the lower half of it and on the super detailed low luma garbage on the ground and also very hard on the low luma trousers of the subject. THANKS AGAIN!

  • HBR is not really true 25p, it is 25p in an interlaced wrapper, it also shares some settings with 50i and 720p50 If I'm not mistaken and hence the only benefit you get from 50i seems to be that you avoid the potential movement artifacts involved with shooting interlaced. So far I've had little use for it in the form it is implemented. (I take you too live in PAL country) 24p shares no settings, has the most advanced hack settings and is the way to go for maximum IQ.

    Keep in mind that "High Bitrate Mode" is a term coined by Panasonic (in relation to factory settings) and has no bearing whatsoever for IQ when it comes to hack settings.

    I'm sure LPowell is right in his assessment, but it's good to keep in mind that you will increase noise and lower detail by using high apertures..

  • Thanks RRRR, so if I understand you well (I am not a native english speaker, and yes on a PAL country, Uruguay) you use 24p that´s cinema mode on a PAL country. So if I do something for broadcast here (PAL) then your choice would be to go 24p all through the workflow until last step that would be to trascode it to 25fps. Isn´t that last step going to ruin all the IQ gained by the 24p good settings? Am i interpreting you well?

  • Yes, you are interpreting me correctly I think.. The only time you need to watch out about 24p is when you are shooting with artificial light, light that you cannot control. Some lights create flicker problems. In which case you might want to use 50i or HBR instead. Often you see these problems instantly, but they can also occur less noticeably if there is a bad light source that affects only part of the picture. (You need a monitor or zoom preview function to actually see what happens then).

    Converting 24p to a delivery format of 25p isn't ideal of course but you won't loose much IQ. If the conversion is done badly you might have some movement / frame interpolation issues (footage doesn't look smooth) but I reckon most popular (pro) software used right now does a pretty good job with this - so I would not worry too much about it.

  • @RRRR, do you just use a 4% PAL speed-up for your 24p material? Any pitch correction?

    I live in an NTSC country but I've imported my fair share of PAL discs and it always amazes me to think most films on PAL DVD are a few minutes shorter because of this.

  • Converting 24p to 25p by speedup looks much better than any interpolation. It has been done for about 60 years on TV in Europe when broadcasting movies!

    If you mind the pitch shift (you really need musician's ears to hear it) all modern sound software can take care of that.

  • Thanks RRRR, so if I understand you well (I am not a native english speaker, and yes on a PAL country, Uruguay) you use 24p that´s cinema mode on a PAL country. So if I do something for broadcast here (PAL) then I have to go 24p all through the workflow until last step that would be to trascode it to 25fps. Am i interpreting you well?

  • Hi nomad, sorry but I don´t understand what you suggest. As far as I know speeding up a timeline and then exporting with the timeline settings on Premiere or Final Cut, won´t convert the frame rate of the resulting file, it will just be played at 24fps but look faster as it will last less time and some frames/lines/information will be discarded. You have to export and transcode anyway to from 24fps to 25fps, moreover I think fps is not the only difference between pal and ntsc, at least in SD number of lines and color interpretation is different too.

  • Simply import your 24fps footage, and get the NLE to interpret it as 25fps in a 25fps sequence. It's a 4% speed increase. Really only an issue for pitch shifts in music, but if you're filming a live band you'd not be using the on camera mic anyway.

  • @GuilleKloetzer As RRRR points out, HBR 25p is encoded in an interlaced fomat, using the same file format as FSH mode. However, HBR is scanned at a true 25p progressive frame rate, rather than the inerlaced 50i field rate of FSH mode. And in Flow Motion v2 and other patches, it records at bitrates up to 100Mbps, which can legitimately be described as "HBR".

  • @LPowell , is there any way to make HBR mode less strobe ? For now i like more 50i mode no strobe ...

  • @feha The reason HBR strobes more than FSH is because HBR is scanned at a progessive 25p frame rate. FSH's interlaced 50 fields per sec is inherently more fluid, so long as the interlace artifacts remain unnoticeable.

  • @LPowell , thank you, now I understand the reason.

  • ok, thanks Chainsaw, I ´ll try that and compare results between 25p at HBR and 24p cinema mode. And thanks LPowell too for that information. I hope to have time to do those tests this week, I will also try Flow Motion side by side with spanmybitchup and maybe some other patch on the same spot I had the troubles with noise that started this thread. I´ll post the results as soon as I have them. Thank you!

  • Well, just to give this thread a kind of conclusion: I´ve tried to reproduce the conditions in which I got the noise posted at the beginning of the thread, but it was impossible for the weather conditions were not exactly the same. Anyway i´ve switched to LPowell Flowmotion 2, and have no complains. With LPowell´s recommended settings I´ve got no glitches on the cam. I´ve been filming in 720p, Rec Quality H, nostalgic, PAL. SD Cards: SanDisk Extreme Pro 32GB 90MB/sec max write speed and Panasonic 16 and 8 GB, max write speed 25MB/s Spanning OK either. I´ll post any unexpected noise or glitch. thanks

  • Hello,

    Try replicating the problem, and then shoot the same scene with the stock firmware. It isn't perfect, but I believe you will get less noise than any patch will deliver.

    I know you said that you couldn't replicate the problem on location, but I have setup tests in my studio to replicate every condition imaginable.

    It might take a little time, but it could save you some embarrassment, or even a client... I think it might be worth it.

    JMHO and a good test for the records.