Quick test: Sony FS700 + Speed Booster vs. Canon 5D Mark III

[jcs]

This is a really quick test*: executive summary- it works. More details on vimeo page.

https://vimeo.com/57994446

 

* someone already did a chart test: http://www.lensrentals.com/blog/2013/01/metabones-magic

 

[Leang]

thanks for the comparison, John.  Do you own a FS700?  Do you have a premature conclusion for these two neck to neck now that there's the Speed Booster?  I think the BMCC with the MFT booster is going to crack some skulls, BUT I would definitely love the option for slower frame rates.  S35 protocol is more than enough for me as I plan to rent out various industry cine lenses.  I'm really concentrated on the Sony F5 atm, and just its brand alone has great resale value where I reside.  I'm really looking forward to XAVC.  100mbs 4:2:2 is more than enough for me since I get good results "uncompressing" 28mbs H.264 MOV

[richg101]

nice to see someone other than well known bloggers doing a test too.  it fortifies what them and metabones are saying.  a bit of crop on the fs700, but nothing huge.  but sharpness...  fs700 and metabones SB smokes the mk3 IMO

[jcs]

Yes, I now own an FS700 due to the release of the SB. I would not have considered that camera if it were not for the possibility to finally upgrade the 5D3 to something close to the full-frame look with true 1920x1080 resolution. There is nothing on the market (that I am aware of) for any price right now that can also do what this combo does (this combo is a 1.09 crop).

 

For high quality shots, I'll be using a Nanoflash (I prefer the option of a high quality long GOP compressor vs. enormous-file I-only codecs (ProRes, DNxHD)- also considered the Atomos Samurai and Sound Devices Pix 240i). So far the internal 24Mbps codec isn't too shabby, but I can see there's room for improvement with an external higher bitrate 422 recorder.

 

Sony's XAVC (on paper) looks awesome. I've been looking forward to someone using more of the H.264 spec for a long time (422/444, 10 or more bits). H.264 is more efficient than anything else: now with GPU acceleration (and fast multicore CPUs) it's not hard to edit without transcoding. Convergent's Nanoflash uses a Sony codec- perhaps their hardware has enough power to support XAVC. Might be a worthy (paid) upgrade.

 

Regarding the F5/F55- excellent cameras for sure. However, they are not true 4K cameras (probably around 3K or so actual resolution; the FS700 will likely be around 3.2K when it supports "4K"). Sony markets the F65 as true 4096x2160. It would be nice if a blogger had a page showing cameras and their actual resolution.

 

Shot a few clips last night with my only non-L lens- the Canon 50 F1.4. Wow- shooting F1.0 is really cool! Zooming in, there's some CA/coma (Chromatic Aberration), but can't really be seen watching footage normally.

[Leang]

I wish the m43 version were out.  the Speedbooster marketing seems to favor the APS-C coverage to full frame. So I'm assuming the m43 version is just a traditional mount adapter for almost benefiting full frame coverage?  I mean better than nothing.  or will the m43 version be calibrated differently...

[jcs]

The m43 version will have a 1.6 crop, bringing it up to Super35 (APS-C). Sounds like it will provide higher performance (MTF) vs. the NEX version.

[jcs]

Tested lenses (Canon):

50 F1.4 (F1.0) - excellent low light at F1.0 with some CA when pixel peeping (or using zoom focus assist)
24-105 F4L IS (F2.8) - works well- no issues noted so far. Nice to have this lens at F2.8
16-35 F2.8L II (F2.0) - works very well. On the 5D3 is a bit soft at 16mm: much sharper and detailed on the FS700+MB at 16mm
70-200 F2.8L II IS (F2.0)- amazing lens is amazinger on the FS700+MB. Detail/sharpness is excellent!

IS (when available) & aperture control worked on all lenses. Autofocus apparently only works in photo mode on the FS700 (didn't try it yet). Autofocus on the 18-200mm kit lens works very well- excellent for sports (will post slomo soccer footage shot on this lens).

[Sean Cunningham]

The m43 version will have a 1.6 crop, bringing it up to Super35 (APS-C). Sounds like it will provide higher performance (MTF) vs. the NEX version.

 

If the ratio is the same as the current model, like they've said it is, it will actually be markedly bigger than S35 and closer to 1DC shooting 4K or the RED Epic 5K aperture.  Folks need to stop handicapping themselves with this "crop factor" business...it's quaint and all, like weight given in "stone" or velocity in "furlongs per fortnight" but it's sloppy math.  

 

That's one aspect of the still photography world that should stay out of motion pictures, even if a lot of us are using stills cameras with dreams of the silver screen.  In motion pictures we deal in precise aperture measurements.

[Andrew Reid]

Thanks for the tests JCS, good info.

[jcs]

Crop factor is super useful to help understand how lenses will work with digital cameras: http://en.wikipedia.org/wiki/Crop_factor

Correction (Speed Booster)- I think the 1.6 CF was for the BMCC; 1.32 CF for m43 (APS-H).

 

Slow motion with the FS700 and 18-200mm kit lens (autofocus worked pretty well; also auto ISO to deal with change in brightness while zooming):

https://vimeo.com/58075075

[Sean Cunningham]

1D/5D FF = 36mm

1DC 4K = 28.45mm

Super 35mm = 24.89

Anamorphic 35mm = 21mm

GH2 = 19mm

 

Metabones Speed Booster = .71 reduction

 

1/.71 x 19mm = 26.76mm

 

CF doesn't translate forward up the post production stream.  It's a terrible measurement, on par with asbergers video engineers saddling us for decades with their stupid diagonal measurement for screens and sensors.

[jcs]

Isn't asbergers what you get when you don't wipe? Asperger's is a form of highly functioning autism- those folks can be geniuses even if their social skills are deficient. Symptoms of crankycantankeritis are common on internet forums, yes?

(I grabbed prior numbers from forum posts). Here are direct calculations:

43.27/sqrt(17.3^2 + 13^2) = 2 CF for GH2
43.27/sqrt((17.3/.71)^2 + (13/.71)^2) = 1.42 CF for GH2 with Speed Booster

43.27/sqrt(15.81^2 + 8.88^2) = 2.39 CF for BMCC
43.27/sqrt((15.81/.71)^2 + ((8.88/.71))^2) = 1.71 CF for BMCC with Speed Booster

43.27/sqrt((23.4/.71)^2 + (15.6/.71)^2) = 1.092 CF for NEX with Speed Booster

Plug in different numbers for sensor width/height if the ones I found are not accurate (it appears full sensor frame is used vs. video mode area, etc.). To get 1.6 CF with BMCC+SB, would need a 16.7mm x 9.5mm sensor.

Now, back to my epic cat video for youtube.

[jcs]

The Nanoflash works well with the FS700. The Sony-supplied 100Mbps MPEG2 MXF codec is a definite improvement over the 24Mbps AVCHD codec. As expected, block artifacts and edge quality is improved. Using the FS700 built-in display with focus peaking, the Nanoflash and LiPo battery don't add too much weight, allowing this setup to be used hand-held and run & gun.

 

Premiere Pro CS6 could not read the MPEG2 with audio in the MOV container.

 

The Nanoflash supports up to 180Mbps long GOP MPEG2 as well as I-frame only mode up to 280Mbps. I'll try 50Mbps long GOP MPEG2 to see it it's good enough (to save disk space: I prefer to keep things small and edit on SSDs).

 

Someday all this gear and image quality will be in one off-the-shelf camera (and probably fit in your pocket  ;)).

[Sean Cunningham]

(I grabbed prior numbers from forum posts). Here are direct calculations:

43.27/sqrt(17.3^2 + 13^2) = 2 CF for GH2
43.27/sqrt((17.3/.71)^2 + (13/.71)^2) = 1.42 CF for GH2 with Speed Booster

43.27/sqrt(15.81^2 + 8.88^2) = 2.39 CF for BMCC
43.27/sqrt((15.81/.71)^2 + ((8.88/.71))^2) = 1.71 CF for BMCC with Speed Booster

43.27/sqrt((23.4/.71)^2 + (15.6/.71)^2) = 1.092 CF for NEX with Speed Booster

Plug in different numbers for sensor width/height if the ones I found are not accurate (it appears full sensor frame is used vs. video mode area, etc.). To get 1.6 CF with BMCC+SB, would need a 16.7mm x 9.5mm sensor.

Now, back to my epic cat video for youtube.

 

 

No, no, no.

 

See, that's part of the problem with mixing exact measurements with this short-hand, stills "measurement" and it's at least partially responsible for some of the bad statements around here re: Metabones Speed Booster.

 

The GH2 is not 17.3mm, it's 19mm.  The standard M4/3 assumption is based on a standard 4:3 sensor size...for stills.  The GH2 is 19mm across, expanded for its native 16:9 recording when shooting motion pictures.  It is not a 2 CF.  It does not become S35 size with this adapter.  S35 does not (really) become fully FF with this adapter.

 

These assumptions, aided by this CF nonsense, tends to be so off an obviously FALSE statement like "GH2 and 7D have the same size sensor" is rendered just as valid because it's within or even less than the margin of error for a lot of the silly CF math floating around here, specifically related to this Metabones adapter.

 

Seriously.

 

edit: here, perhaps this will help you and others for future reference, to properly visualize the relative sizes of these formats to one another.  Not represented here, but easy enough to imagine, is anamorphic 35mm is recorded to a 21mm width which places it (in squeezed form) ever so slightly wider than the GH2 but still smaller than the 7D's sensor width...

 

[jcs]

Panasonic lists the GH2 sensor size as 17.3 x 13.0 mm (in 4:3 aspect ratio), top of this page: http://panasonic.net/avc/lumix/systemcamera/gms/gh2/specifications.html . If 19mm x 10.7mm is the physical area used when in video mode, where does Panasonic show this?

 

If 19x10.7 is used:

43.27/sqrt(19^2 + 10.7^2) = 1.98 CF

 

If we use the 16x9 portion of the full frame sensor:

41.3/sqrt(19^2 + 10.7^2) = 1.89 CF.

 

It's clear there is confusion about what part of the sensor is used for video: http://www.dvxuser.com/V6/showthread.php?271187-crop-factor-1-86x-or-2x

 

If we know the correct active sensor area for two cameras, we can compute the crop factor using this simple equation:

w1, h1 = camera 1 w,h

w2, h2 = camera 2 w,h

sqrt(w1^2 + h1^2)/sqrt(w2^2 + h2^2) = CF

 

Stu's chart is not to scale- 36mm (5D et al) is a lot wider than 27.7mm (Red) and 28.45mm (1DC 4K). Perhaps that's causing some confusion?

 

I have a 5D3 and an FS700+SB. The crop factor was computed as 1.09, and seeing the actual video results of both cameras, it looks pretty close (including overlaying both videos and scaling the 5D3 track to 109%). 

 

It was understood there might be discrepancy with the numbers, however the math is pretty simple and the concept for using ratios of the diagonals (diameters of the image circles) to help explain how a lens designed for 35mm full frame will work with a smaller sensor is useful. If you have an improved means to make sense of this, lots of folks would like to hear it.

[Sean Cunningham]

Hay-soos Creestos. 

 

You're arguing aspects of the GH2 that have already been discussed, repeatedly, here and elsewhere.  You're not even looking at the 36mm part of the image represented by the edge of the entire chart, clearly showing the 5D as being AMAZINGLY larger than the 27.7mm RED.  It's the white line, with the clear dot indicator, at the edge of the frame...the entire chart is encompassed by the 36mm 5D/1D/FF area relative to all the other cameras.  

 

The three different, mode-specific aspect ratios of the 1DC are represented in the chart.   Look at it again.  I'd say you were confused, yes.

 

 

edit: it's pretty ridiculous to even imply Stu made any kind of gross error here.  The kind of work we do more often than not relies on the EXACT measurement of the aperture of the film or sensor involved.  Being off by fractions of a single mm can create major problems moving forward.   Stu doesn't make those kinds of mistakes.  He never has.

 

CF is a joke.

 

edit2: (and I'm done here)...pretty much says it all for me...

 

[jcs]

Put a black border around the chart and it becomes clear, otherwise the white line under the text is the closest line visible. As displayed, the outer white line blends with the background and is invisible. Now that you point it out, the white dot shows the hidden outer line. Thanks for clarifying the hidden line. I make mistakes, Stu makes mistakes, you make mistakes- everyone does. That's life. Not being able to admit mistakes is a mistake and a sign of an ego dominating one's mind (which ultimately causes great suffering for that person and those around them). Doing the ad hominem (attacking one personally) in a discussion is a diversion and does not move a discussion forward in any useful way.

 

Back to the point- the math is simple and is exact as long as we know what part of the sensor is imaged. What method do you propose is better than crop factor for helping people understand how a lens used on one camera will function when used on another with a different sensor size?

[paolo.negro]

Tested lenses (Canon):

50 F1.4 (F1.0) - excellent low light at F1.0 with some CA when pixel peeping (or using zoom focus assist)
24-105 F4L IS (F2.8) - works well- no issues noted so far. Nice to have this lens at F2.8
16-35 F2.8L II (F2.0) - works very well. On the 5D3 is a bit soft at 16mm: much sharper and detailed on the FS700+MB at 16mm
70-200 F2.8L II IS (F2.0)- amazing lens is amazinger on the FS700+MB. Detail/sharpness is excellent!

IS (when available) & aperture control worked on all lenses. Autofocus apparently only works in photo mode on the FS700 (didn't try it yet). Autofocus on the 18-200mm kit lens works very well- excellent for sports (will post slomo soccer footage shot on this lens).

Hi jcs,

I have noted a bad issue with 16-35 F2.8L II on my FS700+SB, at 16mm can't focus to infinite, don't you have the same problem?

Thanks

p.

[jcs]

I have seen that, even at 24mm- try adjusting the infinity focus. Be careful, apparently the screw is fragile. Instructions on MB site.