Cameras for Architectural Photography

Any camera can be used to photograph architecture. A pinhole camera, an old twin lens reflex, a phone camera, an Arca RM3DI, a Kodak brownie, doesn’t matter. Whether shooting for stock, fine art, or for fun, people are free to do whatever they like. 

That said, in the realm of professional architectural photography, there are expectations. Clients want clean, sharp, distortion-free images that are reproducible at scale. With this in mind, there are two major considerations that determine a camera system's suitability to meet such demands.

Large format camera with bag bellows and digital back.

The first consideration is vertical shift. At least one plane, usually the lens plane, should be able to shift upwards and downwards to allow for the correction of vertical perspective distortion. That is, there should be an optical means to keep vertical lines vertical. This is the single most important consideration. Without this functionality, corrections must be performed in software and doing so has considerable drawbacks.

The second consideration is the size of the sensor. There is much ado about resolution but a lot of it is marketing fluff. The physical dimensions of the sensor are more important than its resolution. All factors equal, a larger sensor will produce better all around results than a smaller one. I would rather have a 53x40 mm medium format sensor that is 20 megapixels over a standard 36x24mm sensor that is 45 megapixels. 

Let’s examine these two subjects in greater depth to gain an understanding of how specific equipment can be used to produce high quality architectural photos.

Controlling Perspective

Stand at the foot of a tall building and look straight up. The building’s edges will appear to get closer together as they ascend into the sky. If the edges were to continue extending to infinity, they would appear to converge at a point in distant space. The spot where the lines converge is called a vanishing point. Convergence to vanishing points is a fundamental principle of any form of perspective rendering.

One South First, New York NY. Architect: COOKFOX.

It’s only a problem when convergence contradicts perception. Looking up at a building with sharp receding lines is a severe example, but it doesn’t bother us. In images where the vertical lines are almost straight but not quite, the problem is more acute. Looking straight at a building, we know that vertical lines ought to appear vertical. In life our brains perform this correction for us. If they’re slightly receding in the image it will appear off-putting, if they’re receding sharply, it will appear intentional. When making images, the aim is to align the content of the image with lived perception.

The problem we’re trying to correct is called vertical perspective distortion. It’s also sometimes referred to as keystone effect, as anyone who has struggled to set up a projector can attest. Distortion is to optics as weeds are to plants. If the optical effect is unwanted, it’s a distortion. There is a related problem called horizontal perspective distortion, where horizontal lines that are not quite parallel recede to a vanishing point. Both are two versions of the same phenomenon.

There are always four spatial planes to consider. The sensor and lens planes refer to the position of these camera elements in three-dimensional space. On a large format camera they can be changed independently to adjust the position of the focal plane. This is the plane in which objects in front of the camera will be rendered in focus. And then there is the object plane, the position in space occupied by the subject, for instance, the façade of a building. When the focal plane and object plane overlap, the subject will be in focus. Vertical perspective distortion arises because of a misalignment between the sensor plane and the object plane. If both of these planes are parallel, there will be no such distortion and vertical lines will appear vertical in the resulting image. The aim then is to produce a composition while keeping these two planes parallel.

Large format cameras are simple machines in theory. There are a wide array of options available in this format, including view cameras, field cameras, and technical cameras. Modular configurations allow for customized solutions for different imaging problems. The modern descendants of the earliest cameras used by Talbot and Daugerre, these all operate on the same fundamental principles. A basic setup, for example, consists of two standards on a rail connected by a bellows. A lens is mounted onto a board fitted on the front standard. A digital back is attached to the rear. Focus is achieved by moving one standard (almost always the front) nearer or further to the other. 

Large format camera movements demonstrated on front standard, clockwise from top left: zero (no movements), shift, swing, tilt.

Large format cameras move in three ways: tilt, swing, and shift. Tilt and swing are similar, both introduce an angle to the affected plane. The front and rear standards can be moved independently. Tilt is achieved by pitching one standard forward or backward and adjusts the position of the focal plane on the vertical axis. Swing refers to adjusting a standard laterally so that it alters the focal plane on the horizontal axis. Tilt can be used then to correct vertical perspective distortion and swing can be used to correct horizontal perspective distortion. Some large format cameras will allow these movements in a limited but still practical way. Others will have an extensive range of motion well beyond what is ordinarily needed.

When it comes to photographing architecture, the most used camera movement is shift. Instead of changing the angle of a plane like tilt and swing, shift refers to moving a plane up or down while keeping it parallel to the other plane. The simplest way to correct vertical perspective distortion from the ground is to level the camera and shift the lens plane upwards. It can also be shifted downwards if the photographer occupies an elevated vantage point. This will keep the vertical lines of a building parallel to each other, and perfectly perpendicular to a level horizon. 

Sensor size

The image sensor is the core of a digital camera. Light is transmitted through the lens and onto the sensor which consists of an array of photodiodes on a silicon chip. Also known as pixels (short for picture elements), each photodiode is a distinct point responsible for gathering photons, which are then converted into digital information. This information is processed in software to produce a recognizable image.

There are three factors to consider when evaluating a sensor: sensor size, resolution, and pixel size. The sensor size refers to the physical dimensions, or area of the sensor, which is measured in millimeters. Typical DSLR cameras have a sensor size around 36x24mm, the same as 35mm film. Resolution refers to the number of pixels on the sensor. The amount of pixels on the length of the sensor is multiplied by the amount on the width to get the area. For example, the pixel count on the Sony α9 III is 6000 x 4000, or 24 million pixels. Mega, the metric prefix for million is the standard way this is expressed: 24 megapixels.

Lastly, there is pixel size, a measurement in micrometers of the individual pixels that comprise the sensor. The smaller the pixel size, the greater the pixel density. There are more light receptors packed into the same area. Given the same sensor size, an increased density means that the resolution will be greater. Or put differently, if the sensor size stays the same but resolution increases, the pixel size has to decrease.

Comparison of resolution, sensor size, and pixel size between popular camera models.

For example, the Phase One IQ3 and Leaf Credo 80 digital backs have very close to the same size sensor, only .1mm different in width. The pixel size of the IQ3 is 4.6 µm while the Credo 80 has a pixel size of 5.2 µm. Same physical size of the sensor, but the IQ3 has a 25% increase in resolution because of the higher pixel density.

Conversely, a larger pixel size will result in a lower resolution, but this is not at all a bad thing. Sensors with larger pixel sizes are more sensitive to light as there is more area in an individual photodiode to receive that light. They can collect more photons regardless of resolution, producing files with less noise and increased sharpness. Larger pixel sizes also have larger pixel wells which contribute to an increase in highlight and shadow detail.

In architectural photography, attributes like less noise, increased sharpness, and greater dynamic range are all valuable. Because of this, a digital back like the IQ3 is a better choice for this application than the Nikon Z8. And the Nikon Z8 is a much better choice than the Lumix G9. I say this with a caveat, because the sensor is only as good as the lens in front of it, to say nothing of the practitioner. And of course there are always financial constraints to consider. Large sensors and low pixel sizes come with increased manufacturing costs. Phase One's flagship digital back as of this writing, the IQ4, sports a large sensor with a low pixel size and it costs $30k. Selecting equipment is always a balancing act between needs and budget.

35mm Format

For most people, large format is overkill in both cost and quality. Large format setups come with a lot of costs aside from the financial ones. They are often unwieldy and bulky, difficult to transport and set up. Using them is a slow process, and they are challenging to operate. They are fully manual, and are difficult to troubleshoot if something isn’t working.

Prior to 1961, when Nikon introduced a 35mm f/3.5 perspective control lens to the market, 35mm cameras were all but useless for architectural photography. There was no good way to keep verticals parallel, let alone do anything more advanced. Revolutionary as that lens was, it can only shift up or down and not very much. It was a compromise design, providing the minimum amount of functionality to produce good architectural work. As these lens designs evolved over the next two decades, their relative affordability and ease of use began to supplant the view camera's ubiquity in the field.

Now, 35mm setups are far more common and easier to use than large format. This means that there's more gear available and it costs much less. Modern cameras in this format are divided into two types based on slight differences in functionality, the digital single lens reflex (DSLR) and the digital single lens mirrorless (DSLM).

DSLR cameras use through-the-lens focusing, where light enters the lens, then is reflected through a series of mirrors in the camera body and out through the viewfinder. On a DSLM, the image is projected through the lens directly onto the sensor. They take advantage of fast CMOS sensors which transmit the image signal to the rear LCD screen in near real time. This LCD screen is the way in which an image is visualized before capture on a DSLM. 

DSLMs are better suited to certain applications than other cameras, architectural photography included. Their main advantage is that they allow for the use of a greater variety of lenses. The lens mount format limits lens choices on a DSLR. A Nikon D850 can only accept lenses that are equipped with the Nikon F mount. A Nikon Z8 can accept any lens mount for which Nikon or any third party manufacturer makes a Z mount converter. 

To use myself as an example, I prefer the control layout on Nikon bodies but prefer the optical quality of the wide Canon tilt-shift lenses. Using a lens mount converter on a Nikon Z7, I have a working Nikon/Canon hybrid. Or I could use lenses made by Schneider, Pentax, Zeiss, or whatever else I want. Mirrorless setups are more customizable and modular across platforms, which is excellent.

A second feature that is beneficial to architectural photography is that it is much easier to inspect close focus on a DSLM than it is on a DSLR. Some cameras have rear viewfinders which in practice act like magnification loupes.There is also no need to switch in and out of live view mode, making it much easier and faster when close focusing.

The main downside to a DSLM is that the exposed sensor is prone to getting dusty. There are less components in the way to prevent the intrusion of dust particulates during lens changes. Using a mirrorless camera requires one to carry around a sensor cleaning kit.

A best camera?

Is there a single best camera to use for architectural photography? No. In a perfect world without budget constraints it’s a technical camera like the Alpa 12 Plus or the Arca RM3DI equipped with a Phase One IQ4 back. These offer total control and the best possible image quality, but they require considerable skill to use well.

For most users, beginners and professionals alike, it’s a full frame DSLM with perspective control lenses. There are four major manufacturers of DSLMs: Canon, Nikon, Fujifilm, and Sony. While there used to be more, only the first three of the aforementioned companies are currently producing high quality 35mm format perspective control lenses.

It ultimately comes down to budget and which additional lenses will be used. The great thing about DSLM cameras is that there are adapters that connect most modern mounts to most legacy mounts. So if I know that I want to be able to use old Pentax lenses, I can then research which mirrorless cameras have K mount converters and which don’t, and go from there. No matter the level of skill involved or the type of photography, it’s always important to keep the end result in mind and select equipment that will best serve the achievement of that goal.