Pentax 645 35/3.5 v Olympus 35/2.8: Introduction



Legacy medium format lenses have seen a recent uptick in valuation and demand, thanks largely to Fuji’s GFX system.The Pentax 645Z fraternity continues to watch with detached interest the pointless hurly-burly of the camera business while patiently going about its work much in the manner of twenty years hence – and Phase One and Hasselblad continue to do their thing in the rarified stratosphere of the industry – but we have to credit Fuji for making medium format sexy again. The portability and affordability of GFX refreshed the demand for inexpensive lenses.

Medium format glass is uniformly good: designed and built to professional standards. But in my view the best performance to value ratio, and the simplest to adapt, are vintage Pentax 645 and 67 lenses. They have proven consistently superior to Mamiya, Pentacon and Zenzanon equivalents of the period, and hold their own against Hasselblads old and new. Testament to their excellence is the fact that the current, state-of-the-art D-FA Pentax range wasn’t radically revised from its origins in the SMC era, on through the FA models of the 1990s. And it’s those we’re going to compare.

Lens Specifications

  Pentax-A 645 35/3.5 Pentax FA 645 35/3.5 Olympus 35/2.8 Shift
Introduced 1984 1999 c.1978
Element / Groups 9 / 8 10 / 7 8 / 7
Coatings SMC SP Late-series multicoating (from c.1982)
Serial Tested 4107879 4262558 107641
Weight 470g 560 310g
Minimum focus 30cm 30cm 30cm
Aperture range f3.5-22 f3.5-32 f2.8-22
Filter thread 77mm 82mm 49mm
Aperture type Curved 8 Curved 8 Curved 6
Image circle      

Test method

The hardest lens to get right is a retrofocus wide angle. If we turn a blind eye to the bulky and problematic 24mm and 28mm for now, we notice Pentax offered three options in the past: the manual focus 35mm f3.5 A; its autofocus descendant – the FA 35mm f3.5; and an autofocus zoom – the FA 33-55mm f4.5. On 645 film, a 35mm focal length offers a field of view equivalent to 21mm. On GFX, 35mm corresponds to 28mm. On 35mm, 35mm is 35mm. 

As it happens, there’s a shortage of tilt/shift options for 35mm shooters. Once upon a time we compared the Zeiss PC-Distagon, Nikon 35mm PC and Olympus 35mm shift lenses – all of which have image circles large enough for GFX+. In 2007 we promised to compare them to a Pentax 645 35mm, and it would be remiss if we didn’t do that, eventually. Each of those lenses was pretty good, but none offered tilt movements. To this day, the only off-the-shelf options for tilt and shift are the ancient Canon FD, or the almost-equally prehistoric Arax and Hartblei options. Hence this review – aimed at reaching two market birds with one article written in a stone which we shall now cast.

It’s smarter, cheaper – and tiltier – to combine a tilt/shift adaptor with medium format glass. Any 645 lens permits 15mm of movement on a 35mm camera and 9mm of movement on GFX. And if you use 67 lenses you will run out of mechanical leeway before image circle. 

For the purpose of these tests, we’ve converted a multicoated Olympus 35/2.8 Shift to an M42 mount adaptable to a tilt/shift converter. I’ll be shooting three Pentax 645 options at 35mm on a 35mm camera against the Olympus, and exploring the full 645 image circle to make results equally relevant to medium format shooters. To define the usefulness of these reviews let’s first talk about pixel pitch.

The level of scrutiny with which we inspect an image circle depends on the pixel pitch of the camera used – an especially relevant factor when yanking ancient lenses into the present – from their point of view, the future. Or from the viewpoint of photographers reading this later – the past. The material point, immaterial of time, is that larger-format lenses can afford to be relatively resolution-lazy compared to, for instance, Micro Four Thirds lenses. Concentrated coverage naturally tends to yield higher resolution, and vice versa.

The camera used to conduct these tests (it hardly matters that it’s a Panasonic S1R) has a pixel pitch of 4.3 microns – which means it’s more discriminating than the Fuji GFX 50 and Pentax 645Z (5.3 microns), but not as demanding as the Fuji GFX 100 (3.8 microns). Any future camera with a 4.3 micron pixel pitch – how quaint those fat pixels will soon seem! – will find the results equally relevant.

The image circle of any 645 lens is a minimum of 70mm. Historically, for the purpose of testing 35mm lenses, we’ve divided the image circle of all lenses into Zone A (0-8mm radius), Zone B (8-16mm radius), Zone C (16-25mm radius) and Zone D (25-35mm radius). Like the designer of the MTF chart we assume that a properly-centred lens has identical performance at a given distance from the centre of the image circle. For more on the faux-pas of ‘corner’ and ‘edge’ performance, please see here.



Just as Zone C is largely beyond the corners of an APS-C frame, so Zone D is off-limits to a 35mm sensor – until we shift it. The 15mm movement of a shift adaptor allows us to reach the outer limit of Zone D. In other words, lens performance in the corner of a shifted 35mm capture (allowing for differing incidence angles) is exactly the same as 645 film – slightly outside the corner of a Phase One XT ‘full-frame’ medium format sensor – and it’s exactly the same as a 9mm-shifted lens on GFX. The corner of a 35mm sensor lies in Zone C. Not far away, at the cusp of Zone D, lie the corners of GFX. But beyond that lie the badlands of Zone D – the outer reaches of which are dominated by wild, free-roaming aberrations.

We will now compare real-world sharpness (a combination of resolution and microcontrast) and hunt those aberrations (spherical, LoCa, fringing and geometric), while inspecting flare behaviour and bokeh properties with the image circle shifted, unshifted and tilted.


Lenses are assessed by many metrics with varying degrees of rigour by people with differing intentions and priorities deploying differing methods and equipment at different times using different samples at different distances. The agenda of this review is explained above.

It’s been said that an imperfect test renders results meaningless, but that’s not entirely true: poor methodology still generates data. It only obscures excellence: it can make a good lens look bad, but it can’t improve a bad one, except by temporary comparison. The achievement of excellence in a test cannot be faked. The existence of many shabby tests is therefore just as good as one perfect one. High-level performers rise to the top. 

This will be another test.