Thom Hogan (and thoughts of mine) on format equivalence...

[Dan Wells]

Thom Hogan has a really excellent article out on lens and exposure equivalence between formats..

Here's his list of equivalences for a very similar image (exposure, DOF, noise, motion blur) - assuming all the cameras had the same pixel count. (Thom's are in bold - I've added a couple on both ends, which aren't bolded - errors there are mine)..

With large medium format (Phase One): ISO 6400, f11 at 1/125, 180mm lens
With small medium format: ISO 3200, f/8 at 1/125, 135mm lens
With full frame: ISO 1600, f/5.6 at 1/125, 100mm lens
With APS-C (DX) crop: ISO 800, f/4 at 1/125, 70mm lens
With m4/3: ISO 400, f/2.8 at 1/125, 50mm lens
With CX (Nikon 1): ISO 200, f/2 at 1/125, 35mm lens
With 2/3" (many compacts): ISO 100, f1.4 at 1/125, 24mm lens
With 1/1.7" (smaller-sensor compacts, oversize-sensor phones): ISO 50, f1 at 1/125, 18mm lens
With 1/2.5" (most cell phones): ISO 25, f0.7 at 1/125, 12mm lens.

For this particular image, almost all of the interchangeable-lens cameras are in the "sweet spot". Those lenses exist, diffraction won't rear its ugly head, and the ISO is reasonable. If you sized all the images to 20 MP, they'd look rather similar (opening the question of "why bother lugging the darn Phase if you're going to stop it down to f11 and be forced to ISO 6400").

The 2/3" sensor becomes hard to find the lens for (the ISO works - many premium compacts support ISO 100). Few if any compact cameras have lenses that open to f1.4, especially at a portrait focal length. There are a few that get to f1.8, only half a stop off.

The two smallest sensors require lenses that are very hard to find. A few oversize-sensor phones have lenses in the f1.4 range, only one stop off, but that's the wide-angle lens - the telephoto is generally between f2 and f2.8. The "telephoto" may also not be long enough - many of them are around 50mm full-frame equivalent, not 70mm.

The f0.7 lens on the very smallest sensor is impossible to come close to on the market (it might be possible to build, but it would require enough of a camera bump that any sane manufacturer would put a larger sensor in there instead...).

Depending on the image, the right format could be anything... Let's say you like the look of street photography with the Olympus Pro 25mm f1.2 wide open on an E-M5 mk II. Another option is a Fujinon 35mm f2 wide open (it should be f1.8, but that's very close) on something like an X-T30, but that loses image stabilization. Getting it back requires an X-H1, which is much heavier. Another option is a Z6 with the 50mm f1.8 stopped down to f2.6, but that's much heavier and more expensive.

If you're willing to lose a shutter speed, the Fuji will let you have a stop less noise (both the Olympus and the Fuji start at ISO 200, but you either have to raise the Fuji's ISO one stop or lower the shutter speed one stop to account for the f2 lens). With the Nikon, you can either reduce depth of field farther by opening the lens up, raise the ISO or lower the shutter speed (or a combination).

If you want to keep your original shutter speed (so you'd raise ISO on the Fuji or the Nikon), the Olympus is the best way to get the shot.

A more extreme example (where Micro 4/3 is the only real way to get the shot) is Olympus' new 150-400 f4.5 Pro. It has a built-in teleconverter that makes it a 500 mm f5.6. The full-frame equivalent is a 1000mm f11 (or something faster stopped down to f11).

A 1000mm f11 exists! It's a 4" refracting astronomical telescope - and it makes a decent supertelephoto lens, if you can live with its limitations. Don't even think about autofocus (even variable aperture is relatively rare), and it's about three feet long, requiring a big, heavy tripod.

Another option is a catadioptric telescope, which uses mirrors to fold the optical path (photographers call it a mirror lens). A 5" should just about fill the bill, and it's not very big - but it certainly doesn't have AF (Minolta once made one that did, back in the film era), and it has really, really weird bokeh.

Of course, you could use the Canon 1200mm f5.6 stopped down to f11 - if you can find one of the 30 or so in the world, afford it (~$100,000), and afford the porters and tripod(s) required as well. A Canon or Nikon 800mm f5.6, cropping down to 1000mm field of view is slightly more reasonable (If your definition of reasonable includes a $13,000-$17,000  10 lb lens that's 18 inches long without the hood).

Or you could simply use the Olympus lens...

Sometimes it works out in favor of full-frame. The Nikkor Z 24-70 f4 is significantly smaller and lighter than its closest APS-C equivalent (the Fujinon 16-55 f2.8 ), and it has similar image quality even before you throw it on a body with twice the resolution! Micro 4/3 doesn't have a 12-35 or 12/40 f2 to compete, although there are certainly prime sets that can. Faster short zooms in general are hard to replicate on smaller formats, where the zooms tend to have similar apertures to their larger format counterparts, rather than the equivalent aperture.

[rdonson]

Before some people get their panties in a wad.....just for the record..... there is more than one standard for ISO performance.

Fuji ISO calibration is based on Standard Output Sensitivity (SOS)

Cameras from Canon, Nikon and Sony use a more liberal ISO standard named Recommended Exposure Index (REI)

[faberryman]

Why does he assume noise is linear among ISO/sensor size?

[TimoK]

And different aspect ratios make focal length comparisons more complicated, he maybe thinks only diagonal FOVs.

[faberryman]

And different aspect ratios make focal length comparisons more complicated, he maybe thinks only diagonal FOVs.

I read the article and still am not sure why this information is helpful.

[TimoK]

Yes, I agree. There's too much imagination to be really helpful. At the same time there's something to call information to photographers who are not used to use cameras of different sizes.
I read the article very quickly, but I think I understand something of what he's trying to to compare.
But I wonder, he's locking too many factors to make he's comparison work.  Noise for example, ISO 3200 for small medium format CCD, who is using it in real world?

[BernardLanguillier]

Why does he assume noise is linear among ISO/sensor size?

Because it’s a very good approximation of the reality?

Cheers,
Bernard

[faberryman]

Because it’s a very good approximation of the reality?

I don't know. Is it?

[BJL]

Before some people get their panties in a wad.....just for the record..... there is more than one standard for ISO performance.

Fuji ISO calibration is based on Standard Output Sensitivity (SOS)

Cameras from Canon, Nikon and Sony use a more liberal ISO standard named Recommended Exposure Index (REI)

I am not completely sure but I think this is a misunderstanding, to do with the difference between basic metering (center-weighted, spot) — which has to use SOS — vs fancy pattern metering modes — for which SOS does not make sense, and for which the more flexible REI was created. (This was done by the Japanee industry association CIPA back in the 2003 and then added to ISO 12232 in 2006, and unlike the International Organization for Standardization [ISO], CIPA lets you see its standards for free, so here it is: http://www.cipa.jp/std/documents/e/DC-004_EN.pdf)

Here is my reasoning:
1) All cameras have to follow ISO standard 2720 for light metering (not part of ISO 12232, which defines those two measures of output sensitivity). I have seen no sign of any evidence that this is not the case.

2) If with a uniformly lit scene and/or simple center-weighted or spot metering, and exposing according to what the in-camera light meter suggests, a camera were to deviate significantly from SOS, then images of test cards or simple scenes with an obvious dominate midtone level would come out looking over-exposed or under-exposed in standard JPEG conversions, because those metered mid-toned would be placed above or below the JPEG level of 118 ["18% gray"] that SOS dictates. That would produce obvious and probably annoying effects readily visible to many camera users, like too pale of dark faces, and readily blown highlights in the case of placement higher than 118. Again, I have not seen any evidence of that happening.

3) The following confusing phrase from CIPA D-004 on REI: "When the camera in question is equipped with a traditional (i.e. simple) automatic-exposure function, the average exposure in the focal plane given by the automatic-exposure function to a uniform photographic subject must naturally be equal to the above-mentioned Em value recommended by the vendor of the camera." I think that is saying that with simple exposure modes and a uniformly bright subject, REI must behave "normally" giving a suitable "18%" placement. So REI is not as arbitrary as some have claimed.

(I suspect that the confusion goes back to the nonsensical idea that the ISO exposure index setting on cameras should be using the saturation-based measure Ssat of sensor (or raw file) highlight headroom instead of using those CIPA-mandated output-referred measures.)


TL;DR Point any digital camera at a uniformly lit gray card and use a shutter speed/aperture combination as recommended by a light meter for the chosen ISO exposure index setting, and the default in-camera JPEGs will place that gray very close to 18% — which is at heart what "output sensitivity" is about. If I am wrong, please demonstrate!

[BJL]

Why does he assume noise is linear among ISO/sensor size?

And he doesn't really need to: normalizing FOV, exposure time, DOF etc. forces use of Exposure Index values following that pattern. Those choices will equalize the photons hitting the sensor, so assuming (as he does) similar levels of technology [QE], will equalize counts of detected photons, and thus equalize shot noise and thus ratio of signal to shot noise. So the main deviation from linearity is the extra noise generated within the camera by thermal effects and such. The total amount of that noise tends to increase with larger total sensor area, so there is a bit of a trend for SNR to get worse as sensor size increases in these "equivalent" scenarios, but I do not know by how much. Dual ISO sensor technology helps to reduce that effect somewhat, by effectively shrinking photosite area at high enough ISO speeds.

[BJL]

And different aspect ratios make focal length comparisons more complicated, he maybe thinks only diagonal FOVs.

Thom Hogan does explicitly acknowledge that with respect to 4/3" format (but overlooks the same issue with "small medium format"!)  It is a fairly small deviation though; comparing 35mm to 4/3":
- the width ratio is 2.08 instead of 2: 4% over
- the height ratio is 1.85 instead of 2: 8% under
- the sensor area ratio is 3.84 instead of 4; 4% under

That is precise enough for the purposes of this rough comparison.

[BJL]

A more extreme example (where Micro 4/3 is the only real way to get the shot) is Olympus' new 150-400 f4.5 Pro. It has a built-in teleconverter that makes it a 500 mm f5.6. The full-frame equivalent is a 1000mm f11 (or something faster stopped down to f11).

A 1000mm f11 exists! It's a 4" refracting astronomical telescope ...

Another option is a catadioptric telescope ...

Of course, you could use the Canon 1200mm f5.6 stopped down to f11 ...

Or you could simply use the Olympus lens...

Or if it existed, one could use a 2X teleconverter (as with a Barlow lens for astronomy) which would enlarge the image from any MFT lens to fill the 35mm format frame while dimming it by two stops. I often think of all these equivalence comparisons as similar to the effects of using an appropriate teleconverter to use a lens from a smaller format on a camera in a larger format. Focal length, FOV, DOF, circle of confusion size, diffraction spot size, needed Exposure Index ("ISO") for the same shutter speed all scale appropriately.

[Dan Wells]

Yes, a teleconverter without baffles would do the job.

Internal anti-reflection baffling on many lenses complicates the question of coverage. Most long telephotos cover a fairly large format natively (and a teleconverter increases coverage by magnifying the image, as BJL says). However, the manufacturers deliberately constrain coverage to fight reflection in case the inside of the camera body is not perfectly anti-reflective.

One way to see this is to look at 300mm f2.8 lenses - their physical size doesn't vary with format coverage. The old Olympus 4/3 (pre Micro 4/3) 300mm f2.8 is actually quite a bit larger and heavier than Canon and Nikon full-frame lenses, and Sigma made a remarkably small 300mm f2.8 for a while (it covered full-frame just fine). The old Mamiya lens which covered 645 was no bigger than a typical full-frame lens. I suspect they were all similar designs baffled down to cover different formats. I remember reading somewhere that at least one relatively standard 300mm f2.8 would actually cover 4x5" if assembled without the baffles.

A teleconverter without baffles (or baffled to permit a larger format) would be easy enough to build...

[Jack Hogan]

With large medium format (Phase One): ISO 6400, f11 at 1/125, 180mm lens
With small medium format: ISO 3200, f/8 at 1/125, 135mm lens
With full frame: ISO 1600, f/5.6 at 1/125, 100mm lens
With APS-C (DX) crop: ISO 800, f/4 at 1/125, 70mm lens
With m4/3: ISO 400, f/2.8 at 1/125, 50mm lens
With CX (Nikon 1): ISO 200, f/2 at 1/125, 35mm lens
With 2/3" (many compacts): ISO 100, f1.4 at 1/125, 24mm lens
With 1/1.7" (smaller-sensor compacts, oversize-sensor phones): ISO 50, f1 at 1/125, 18mm lens
With 1/2.5" (most cell phones): ISO 25, f0.7 at 1/125, 12mm lens.

One has to ask: who at LuLa shoots equivalent full frame landscapes at ISO1600?  For a generic landscape reference how about variations on Sunny 16 at ISO100 (for FF say ISO100, f/8, 1/400s - or ISO100, f/11, 1/200s)?  Then the comparison to other formats makes more sense here.

A clarification for the article: different format cameras set up equivalently collect the same amount of light, so shot noise is exactly the same across the board all else equal (see for instance here  https://www.strollswithmydog.com/equivalence-and-equivalent-image-quality-signal/ ).

Jack

[Bart_van_der_Wolf]

I don't know. Is it?

Yes. See https://blog.kasson.com/the-last-word/noise-performance-of-downsampling-with-eliptical-weighted-averaging/  and  https://forum.luminous-landscape.com/index.php?topic=93565.msg763396#msg763396

Cheers,
Bart

[faberryman]

Yes. See https://blog.kasson.com/the-last-word/noise-performance-of-downsampling-with-eliptical-weighted-averaging/  and  https://forum.luminous-landscape.com/index.php?topic=93565.msg763396#msg763396

Thanks. I would never has guessed ISO 6400 for Phase One was the equivalent of ISO 400 for m43. Seems like a good argument for using the largest format possible. Most of that argument seems focused on resolution.

[BJL]

One has to ask: who at LuLa shoots equivalent full frame landscapes at ISO1600?

... different format cameras set up equivalently collect the same amount of light, so shot noise is exactly the same across the board all else equal (see for instance here  https://www.strollswithmydog.com/equivalence-and-equivalent-image-quality-signal/ ).

Indeed; perhaps some conclusions are:

1) Larger formats are often not used in ways "equivalent" to what would or could be done in a smaller format, but instead achieve some advantages in certain situations through the ability to use settings for which a smaller format cannot offer equivalent settings, because that would require excessively low f-stops and/or blown highlights due to gathering more light than the sensor can handle (i.e. the equivalent ISO speed would be below the smaller sensor's base ISO speed).

2) The advantages of these "inequivalent options" derive largely from gathering more light from the subject, and so likely improving S/N ratio (so long as sensor dark noise, read noise and so on do not ruin that.)

3) The "inequivalent" settings also involve using longer exposure times and/or larger lenses (strictly, larger entrance pupil sizes a.k.a. effectively aperture diameters; in practice, larger front elements). This in turn means more subject motion blur and/or more OOF blur (less DOF).

[BJL]

Internal anti-reflection baffling on many lenses complicates the question of coverage. Most long telephotos cover a fairly large format natively (and a teleconverter increases coverage by magnifying the image, as BJL says). However, the manufacturers deliberately constrain coverage to fight reflection in case the inside of the camera body is not perfectly anti-reflective.

Yes, the optical design for lenses distinctly longer than normal typically produce on "oversized" image circle (coverage often about 40º or so even for extremely long foal lengths, IIRC) but that is the sometimes baffled down to cover just the needed angular FOV. But that is no problem for a TC: it can take the provided image with the desired angular FOV, and produce a larger, dimmer version of that image covering the same angular FOV.

[Dan Wells]

Thom's original equivalence table was set up so the settings were reachable on as large a number of formats as possible (starting at ISO 1600 on full-frame keeps the smaller formats within their available ISOs . If you start with a sub-100 ISO on full-frame, which many of us shoot regularly, there are no lower ISOs available on currently produced smaller-format cameras.

There's no way to cover the whole range of format sizes without reaching absurd settings at both ends - but starting somewhere like:

ISO 64, 1/60 second, f5.6 (three stops down from Sunny 16, typical late afternoon light on a D850, Z7 or anything else with a true ISO 64)

immediately goes into impossible  settings on any smaller-format camera (there is no APS-C camera with ISO 32, nor a Micro 4/3 camera with ISO 16). Instead of being able to lower the ISO to chase full-frame's noise performance, the remaining option is to raise the shutter speed (not in itself harmful - there's not much desirable motion blur at 1/60) - but you can't compensate for the noise. In fact, it's slightly worse than that, because the smaller format camera probably can't reach ISO 64. Most APS-C cameras bottom out at either ISO 100 or 200. and Micro 4/3 is usually if not always 200).

ISO 200, 1/800 second, f2.8 on Micro 4/3 is comparable in terms of depth of field to the full-frame shot at ISO 64, and if the full-frame shot were taken with a really good 50mm lens (say a Sigma ART) stopped down to f5.6, the Olympus PRO 25mm stopped down to f2.8 is very comparable - but you pick up ~3 2/3 stops of noise and lose several stops of dynamic range compared to the low-ISO full frame shot.

Going the other way is far easier - ISO 125, 1/60 second, f8 makes perfect sense on a Fuji GFX, and the 63mm lens is only slightly too wide (it should be about a 70mm).

ISO 250, 1/60 second, f11 is relatively sensible on a Phase One back. The Phase shooter might choose to open the aperture and lower the ISO, trading off a bit of depth of field for an ISO closer to the back's best performance (250's not bad, but Phase backs are pretty special at low ISOs). Neither the 80mm nor the 110mm is a perfect match, but the 80's not bad.

By starting higher on the ISO scale, Thom allowed the smaller formats to stay within their ISO limits. Low ISO is one of the places where big sensors shine. Nikon's ISO 64 allows their high-resolution full-frame cameras to compete with small medium format in situations where the low ISO works. I'm not clear whether the S1R has a true ISO 50, or whether that's an expanded setting - but if it does have a true ISO 50, that's even more advantageous. Some of the big medium format backs go even lower (a few Phase backs support a true ISO 35).

You can't always use low ISO, but when you can, it's a significant advantage to the cameras that allow it (which also tend to have big sensors)...

[BJL]

Thanks. I would never has guessed ISO 6400 for Phase One was the equivalent of ISO 400 for m43. Seems like a good argument for using the largest format possible. Most of that argument seems focused on resolution.

That “noise level equivalence” with exposure index proportional to sensor area is fairly well known, but it only leads to lower noise in the final image if one breaks some other part of the equivalence—typically, accepting less DOF by using a “lower than equivalent” aperture ratio (I assume equal exposure time, because when one has the flexibility to use a longer exposure, one would’t be messing with such high ISO speeds.)