JVC owns AltaSens. Understanding VSM

[Steve Mullen]

*** UPDATES

*** VSM not only functions as a scaling engine it also controls the sensor causing it to switch between window sizes/modes and can likely can enable pixel binning.

Since JVC owns AltaSens we know the LS300 sensor has a gross resolution of 4512x3008 (13.57MP) and has an aspect-ratio of 3:2 (1.50:1). The diagonal is 28.20mm which is correct for an S35 sensor. A capture rate of up to 50fps is supported. [An alternate resolution of 4512x2520 supports up to 60fps, but its resolution is only 11.28MP. JVC, however, specs the sensor as having 13.57MP. Were 4512x2520 used, the chip could support 4K/UHD at up to 60fps.]

*** The sensor's effective horizontal resolution is 4452px. It is indicated by the VSM notation of 100%(Super35).

A lens designed for an S35 camera will fill this area. [Note, the lens mount-size is irrelevant except for mechanical reasons.]

The 92%(C4K) notation indicates the point where the capture window is 4096px-wide (4096 is 92% of 4452) in C4K mode. To fill the C4K target a lens would have be designed for a sensor that is at least 92% of an S35 sensor. An S35 sensor is 24.89mm wide, so the sensor should be at least 23mm wide. A lens designed for an APS-C sensor (24mm x 16mm) meets this requirement.

*** To capture C4K with no up-scaling, a lens designed for an APC-C or S35 camera should be used. Using a lens designed for an MFT sensor will force upscaling. An S16 lens should not be used.

*** The 46%(C2K) notation indicates the point where the capture window is 2048px-wide (2048 is 46% of 4452) in C2K mod. An S16 lens can be used at a setting of 52%(Super16) with no upscaling is required.

*** When shooting C4K and C2K the VSM range of settings from 100% to 46% are based on 4452px-wide window.

*** The AltaSens sensor has a "4K Array Mode" that has a resolution of 4812x2518 (12.1MP). It has an aspect-ratio of 1.91:1 and, most importantly, a smaller diagonal of 26.72mm. [In this mode, the sensor supports a capture rate of up to 72fps.]

*** When you shoot 4K/UHD, VSM switches to this mode. The switch is indicated by the "odd" 86%(4K) notation. [The other notations relate to a lens type, while this one does not.] When shooting 4K/,UHD, the VSM range of settings from 86% to 80% are all based on a 4812x2518 window.

*** Looking at this mode's diagonal of 26.72mm verses an MFT sensor diagonal of 21.6mm, the correct VSM percentage should be 81% for an MFT lens. And, it is 80%(MFT). Therefore, no upscaling is required for lenses designed for MFT cameras.


*** When set to shoot HD, VSM *may* enable 2x1 on-chip binning that allows the chip to read-out only 50% of its pixels per frame which increases the potential capture frame-rate to 120fps. A binned read-out window could be either 4512/2 > 2256x3008 or 3008/2 > 1504 yielding 4512x1504. Both preserve the RGB Bayer pattern. The VSM percentage for 1920px need be only 42.6% of 4512px and indeed it is 43%(HD). So were binning used, the capture window is likely 4512x1504. The 4512x1504 capture window would then be appropriately down-scaled to FHD and HD. [Were firmware to increase the frame-rate for HD to 120fps it would indicate binning is in use.] If binning is not enabled, when set to shoot HD, the 4512x3008 capture window is simply appropriately down-scaled to FHD and HD.


Lastly, any lens designed for a sensor larger than S35, will result in a cropped image. A FF photo lens will have a crop factor of about 1.5 so a 50mm lens will have a field of view equal to a 75mm lens. With a Meta-bones adaptor, the focal length will be reduced by 0.71 yielding a 53mm lens that's 1-stop faster.

[Steve Rosen]

No one has responded to this, which is interesting in itself because on any other blog site (Red or BMD) it would have been discussed and dissected into the next millennium...

It's interesting, because not being a video person ( I understood film stocks and timing lights, but have never grasped this technology) it confirms my own visual tests with the many lenses I own.

BTW, I've found, as mentioned elsewhere, that some native MFT lenses - such as the Olympus 12-40 and the Lumix 35-100 - will allow as much as a 92-95% VSM setting with minor vignetting on one corner (on my camera, the bottom right). I usually play it safe and go for 89% though. This allows the 12-40 to cover a large range when shooting 1080 - from very wide to a nice 85+mm equivalent - which minimizes the need to change lenses in dusty or wet environments, or when working fast.

[Steve Mullen]

"such as the Olympus 12-40 and the Lumix 35-100 - will allow as much as a 92-95% VSM setting with minor vignetting on one corner (on my camera, the bottom right). I usually play it safe and go for 89% though."

I too have found this as JVC sent me the Oly lens.

I'm surprised too. One German site claims you can't get UHD from MFT lenses. Which would be true if the special mode was included.

JVC Japan is fact checking all this. I had hoped others might respond as you did. Is it too much tek detail?

I'm now deep into J-log. Log is so simple to write about. But actually shooting it with no information from JVC is mind blowing.

Things like setting the upper and lower Zebra and Histogram points to the correct points. I feel like I'm reverse engineering the LS300..

[Noa Put]

with my 42,5mm f1.7 lens from panasonic I can set vsm to 100% without vignetting.

[Steve Rosen]

"Is it too much tek detail?"

Not at all, in fact you make it very clear to people like me. At BM sites sometimes there are pages and pages of tech talk that gets overwhelming - I just want to know what I need to know in order to shoot with confidence in the result.

Noa: Interesting about that Panasonic lens. I wonder how many other MFTs there are that will cover.

It's a shame JVC is the only manufacturer with an S35 sensor and MFT mount - If more adopted the idea, maybe some bright company like Rokinon would make a real manual parfocal 5:1 zoom lens with an MFT mount that covers S35. That would be heaven.

[Noa Put]

I checked again but this time pointed the camera at a white wall and now I do see the vignetting at 100% but it's a very light vignette, the same applies for my 75mm. It looks like tele lenses in general can benefit from covering the full sensor

[Steve Rosen]

With the 12-40 and the 35-100, I can set the VSM at 95% and actually get a single black corner, with noticeable shading/vignetting - but, while that might be objectionable in stills, where you tend to study the entire frame, it's not usually noticed in moving footage (except by peepers, but they don't count).

In fact, in FCPX I often put a vignette on my exteriors to darken the edges - so a little shading from a lens doesn't bother me at all... and at 95% it gives me a great range. Again, if I'm at all worried (i.e. because of a client) I back off to 89%. But that's more from paranoia than personal experience.

[Steve Mullen]

*** UPDATES

VSM not only functions as a scaling engine it also controls the sensor causing it to switch between window sizes/modes and can likely can enable pixel binning.

*** Since JVC owns AltaSens we know the LS300 sensor has a gross resolution of 4512x3008 (13.57MP) and has an aspect-ratio of 3:2 (1.50:1). The diagonal is 28.20mm which is correct for an S35 sensor. A capture rate of up to 50fps is supported.

The sensor's effective horizontal resolution is 4452px. It is indicated by the VSM notation of 100%(Super35). To achieve an aspect-ratio of 1.78:1 2160px of 2500px are used.

A lens designed for an S35 camera will fill this area. [Note, the lens mount-size is irrelevant except for mechanical reasons.]

*** The 92%(C4K) notation indicates the point where the capture window is 4096px-wide (4096 is 92% of 4452) in C4K mode. To achieve an aspect-ratio of 1.78:1 1080px of 1150px are used.

To fill the C4K target a lens would have be designed for a sensor that is at least 92% of an S35 sensor. An S35 sensor is 24.89mm wide, so the sensor should be at least 23mm wide. A lens designed for an APS-C sensor (24mm x 16mm) meets this requirement.

To capture C4K with no up-scaling, a lens designed for an APC-C or S35 camera should be used.

Using a lens designed for an MFT sensor will force upscaling. An S16 lens should not be used.

*** The 46%(C2K) notation indicates the point where the capture window is 2048px-wide (2048 is 46% of 4452) in C2K mod. To achieve an aspect-ratio of 1.78:1 2160px of 2500px is used.

An S16 lens can be used at a setting of 52%(Super16) with no upscaling is required.

When shooting C4K and C2K the VSM range of settings from 100% to 46% are based on 4452px-wide window.


*** The AltaSens sensor has a "4K Array Mode" that has a resolution of 4480x2518 (11.28MP). It has an aspect-ratio of 1.78:1 and, most importantly, a smaller diagonal of 26.72mm. [In this mode, the sensor supports a capture rate of up to 72fps.]

*** When you shoot 4K/UHD, VSM switches to this mode. The switch is indicated by the "odd" 86%(4K) notation. [The other notations relate to a lens type, while this one does not.] When shooting 4K/,UHD, the VSM range of settings from 86% to 80% are all based on a 4480x2518 window.

*** Looking at this mode's diagonal of ~24mm verses an MFT sensor diagonal of 21.6mm, the correct VSM percentage should be 81% for an MFT lens. And, it is 80%(MFT). Therefore, no upscaling is required for lenses designed for MFT cameras.


*** When set to shoot HD, VSM *may* enable 2x1 on-chip binning that allows the chip to read-out only 50% of its pixels per frame which increases the potential capture frame-rate to 120fps. A binned read-out window could be either 4512/2 > 2256x3008 or 3008/2 > 1504 yielding 4512x1504. Both preserve the RGB Bayer pattern.

The VSM percentage for 1920px need be only 42.6% of 4512px and indeed it is 43%(HD). So were binning used, the capture window is likely 4512x1504. Thus, line-pairs are read-out.

The 4512x1504 capture window would then be appropriately down-scaled to FHD and HD.

[Were firmware to increase the frame-rate for HD to 120fps it would indicate binning is in use.] If binning is not enabled, when set to shoot HD, the 4512x3008 capture window is simply appropriately down-scaled to FHD and HD.


Lastly, any lens designed for a sensor larger than S35, will result in a cropped image. A FF photo lens will have a crop factor of about 1.5 so a 50mm lens will have a field of view equal to a 75mm lens. With a Meta-bones adaptor, the focal length will be reduced by 0.71 yielding a 53mm lens that's 1-stop faster.
__________________
Switcher's Quick Guide to the Avid Media Composer >>> DVC

[Steve Mullen]

*** More updates

*** When set to shoot HD, VSM likely enables 2x1 on-chip binning that allows the chip to read-out only 50% of its pixels per frame which increases the potential capture frame-rate to 120fps. The VSM percentage for 1920px need be only 43.13% of 4452 and indeed it is 43%(HD). But, the VSM percentage for 1920px need be only 42.86% of 4480 and indeed it is 43%(HD). Hmmm.

A binned read-out window could be either 4452/2 > 2226x2500 or 2500/2 > 1250 yielding 4452x1250. Or, 4480/2 > 2240x2518 or 2518/2 > 1259 yielding 4480x1259. All preserve the RGB Bayer pattern.

So if binning is used, the capture window could be any of these.

The capture window would then be appropriately down-scaled to FHD and HD.

[Were firmware to increase the frame-rate for HD to 120fps it would indicate binning is in use and that either 2240x2518 or 4480x1259 is used.]

Because of the option for 120fps, I think 2240x2518 or 4480x1259 is used. My hunch is that every line is read-out, but only every other column which would be 2240x2518.


*** S16 is 52% of either 4452 or 4480. I'll guess in C2K mode 4452 is used, while in HD 2240x2518 or 4480x1259 is used. Again, my hunch is 2240x2518.

[Noa Put]

I"m not understanding how these vsm percentages actually work with mft lenses, my 12-35mm f2.8 panasonic lens vignettes at 12mm when I set the vsm to 82% mft. However I also have a 15mm f1.7 pana lens and there is no vignetting up to 95% vsm. How is this possible?

I have an upcoming shoot next week where I wanted to shoot at c4k, I also want to shoot in a 2.35:1 aspect ratio which means I can even shoot at 100% vsm without seeing any vignetting on the 15mm.

Would this be a wise choice? Not sure if I any up scaling of the image would cause image degradation? I could use the extra viewing angle I get shooting at 100% since I"m shooting a interior with that 15mm lens.

[Steve Mullen]

The comments I've read say it's well known that MFT lenses have an optical circle much larger than needed for MFT cameras. Well I certainly had never heard this. Could it be that IF M43 lens design was to include film -- 127 if I remember -- that would explain the larger than expect image.

Possibly the shorter (wider) the focal length the smaller the optical circle. Or, maybe there is simply a difference in design between a zoom and a prime.

"Not sure if I any up scaling of the image would cause image degradation?"

You are scaling down to CK4 when you are at 100%. And you are right you are getting a wider than 15mm field of view -- which I suppose could be calculated.

[Noa Put]

Thx for your answer, I don't see any image degradation though I have not pixel peeped, just hooked the jvc up to a large led screen. I guess I"ll just go for it and see how it turns out.

[Steve Mullen]

I have no idea why JVC warns f going up past 82%.

[Steve Rosen]

I'm about 90% through a 30 minute doc for a client, and have been setting the VSM for 100% when shooting UHD (some of what I'm shooting - Big Sur, Monterey coast, Salinas Valley, etc - I will want to archive for future use) and have never noticed any reason not to use 100%.

The interviews and most of the other visuals I've been at 1080p and set the VSM all over the place depending on the lens - most often, as I've said, at 89% with MFT lenses (Lumix 35-100 and Olympus 12-40) and 100% with Rokinon and Leica R primes... Again, I'm not noticing any significant variation in image quality.. maybe there's something foul lurking somewhere under there, but I'm not seeing it.