GH2 ISO tests: Rethinking the ISO bug & ISO 320 cleaner than ISO 160?

[QuickHitRecord]

I am also running Moon Trial 3. Best hack that I have used yet. I pretty much keep it on either ISO 160 or 640, though in your test 640 is cleaner!

[lmackreath]

Based on this topic I have just ran this quick test.

 

The first 10 seconds is ISO 160 and the other 10 seconds is ISO 640. In my opinion 640 looks good but its alot 'busier' noise than that of 160...I turned up the gain and gamma to show the noise more clearly on the PC unit.
This test was carried out with the GH2 ISO Bug in mind, meaning that I selected each ISO by first selecting the ISO an increment above it first and then selecting downwards.

 

https://vimeo.com/60165118

[Sean Cunningham]

I am also running Moon Trial 3. Best hack that I have used yet. I pretty much keep it on either ISO 160 or 640, though in your test 640 is cleaner!

 

 

Yeah, that's just nuts.  I re-did several of the tests as a "crazy check" to make sure the particular ISOs were coming through right.  I didn't, at first, believe that ISO250 could be so crap.  ISO640 was where I was really surprised though.  It's nice to see that ISO1250 is totally usable, if need be, planning for some NR.

[Mirrorkisser]

@Quickhitrecord: Why not 320? I like that quite a lot, too with Moon. In fact thats the iso i shoot with mostly. 160 only in broad day light.

[Sean Cunningham]

@Quickhitrecord: Why not 320? I like that quite a lot, too with Moon. In fact thats the iso i shoot with mostly. 160 only in broad day light.

 

 

While "Moon" brings ISO320 back from the dead, it's still the lowest quality, from a noise standpoint, between ISO1250 and ISO160.  ISO400 is noticeably cleaner.

[Mirrorkisser]

Hmm. I read somewhere though that you have the most dynamic range at 160 and 320, from 320 on it really goes down. So maybe its a choice between dynamic range and noise? Maybe its better to choose more noise and just denoise later. You know a lot more than me about this and cameras burnetrhoades, so maybe you can shed some light into this? Cheers!

[Sean Cunningham]

Admittedly, I hadn't considered that.  Most of the information I found rather quickly on the subject deals with shooting RAW and from a stills photographer perspective but I'd assume these same factors play into shooting video.  They have much, much more latitude to de-noise without wrecking usable information than we do at HD resolution and only 8bits of compressed, sub-sampled chroma.

 

At that point you have to weigh whether the slight dip in DR going from 320 to 400 that may or may not actually affect your image in practical, observable effects is more important than the difference in noise that, depending on the patch, is clearly visible and clearly affects your image in practical, observable effects.  

 

From a RAW, stills photographer standpoint, DR at ISO160 is 10.8 stops and at ISO400 it's 9.8 stops...that's based on one site.  Another claims almost a whole extra stop of DR starting at ISO160.  Yet others claim ISO400 is the base and then multiples of 160 are derived from these multiples of 400.  All I take away from it currently, given the very thorough explanation of exposure and DR with the GH2 for video by Shian Storm at his ColorGHear site, is that the useful  DR is much lower than these stills guys enjoy, I'm going to go for less noise (though shooting Moon versus Flowmotion renders the difference between 320/400 ISOs less of an issue).

 

Most of what I intend to shoot I have to try to contain under ISO1250 so it's all gravy, so long as I stay away from 250 and 500.

[Mirrorkisser]

Cheers for your answer! I guess i will have to try both ways and then see which way works best for me. In the end you are most likely right, that the little gain in DR is not worth the extra noise.

 

If i  may ask, which picture profile do you use? I have always been a standard or smooth boy, but i also read that nostalgic has the highest dynamic range (but also more noise than standard or smooth) and will give nostalgic a go, too. Andrew Reid is one of the laudators for nostalgic, too. Some people also use iDynamic for extra DR, but i am not a big fan of those in camera automatic settings, as i fear that they spit into my soup by giving me something  when i dont want it.

 

Thank you again!

[Sean Cunningham]

Going back and looking at my own results again, with Moon it appears that 320's only objectionable noise comes from its green channel while luminance is quite clean.  That's with the "tungsten" setting.  I've yet to go through and do a "daylight" test.  I've been rendering out, ironically enough, film grain tests the last couple days adding "noise" back into Moon 3 footage after removing what's there in-camera, lol.

[Paulio]

So you guys recommend avoiding 320? its kind of a hand iso for indoor daytime. I guess i'll try 400.

[Sean Cunningham]

That all depends on your patch.  One would think that noise performance would be totally based on the chip and the various patches would only apply an overall enhancement or de-emphasis of fixed performance but that doesn't seem to be the case.  ISO320 is good on Moon Trial 3 (an All-Intra patch) but maybe not with other patches it seems, like Flowmotion (a GOP-3 patch).

 

Perhaps it's yet another GOP related issue that has more to do with what is interpreting and transcoding the MTS for display than absolute noise performance.  Not all AVCHD readers are created equal, as many of us have learned, the hard way and after much hair pulling and gnashing of teeth.

 

Oh, and sorry, Mirrorkisser, I use SMOOTH exclusively.  

[PannySVHS]

Classic EOSHD forum stuff. Kids, get the EOSHD searchengine started! I highly recommend taking advantage of the EOSHD archive!

[Cinegain]

Yeah, there you might find golden stuff like Zach Goodwin and his T2i adventures back as well.