Why did drum scanners give way to other technology?

[pkr1979]

{moderator note: this post was split off from the original thread here: https://www.photrio.com/forum/threads/scanning-with-the-emulsion-facing-the-wrong-way.215263}

Also not so relevant to the question of film orientation, but Ive also been comparing my drum scans with scans Ive been doing with my CoolScan 8000... same negatives. The resolution is about the same but the drum scans are indeed significantly better. Considering drum scanning is such a hassle I had a strong bias for the CoolScan... I wanted it to be better, as good, or unnoticeably inferior. This isn't the case though. I also scanned some 8x10 E100 and it is spectacular.

And this made me wonder... why did they quit this PMT technology? Couldn't this be 'transferred' to a flatbed-kind-of-scanner?

 

[koraks]

That's an interesting question, and while it's easy enough to ask, answering it is actually quite complex.

Ultimately I think it boils down to two rationales: (1) productivity and (2) discontinuous technological trajectories.

As to (1), drum scanning essentially boils down to digitizing the film one single pixel at a time. This means that if it needs to be done fast (which you'd done to cut down wait times and increase human operator efficiency), the process needs to be accelerated to pretty insane levels. This is exceedingly complex and expensive to do due to a couple of fundamental technological challenges involved. At the time other scanning concepts arrived at the scene, it's doubtful that the technology even existed or was feasible to create to get a drum scanning concept at a similar productivity level as currently common concepts using strip CCD's.

Concerning (2), photomultiplier tubes (PMT's) are essentially vacuum tube technology and as such are rooted in the early 20th century. By the 1960s, solid state/silicon-based electronics started to replace vacuum tube tech at a large scale. The latter has the obvious advantages (in general) of being more compact, allowing for simpler circuitry, lower cost and lower energy consumption. Combined with (1), the overall simpler, more compact and cheaper design of a solid-state/silicon-based scanning concept turned out to be more attractive.

Taken together, apparently the higher cost, complexity and lower productivity of a PMT-based approach just wasn't justified by the higher quality. Keep in mind that silicon-based approaches can be pretty high quality as well; for instance, the FlexTight concept realized fairly high quality levels using silicon pretty early on in the era of CCD-based scanners. The fact that we're working with often less capable CCD-based machines has more to do with cost and size requirements than with the capabilities of the concept as such.

Of course, in recent years, single-pass capture using a 2-dimensional array (e.g. 'camera scanning') has logically continued the direction that was already initiated when PMT mostly gave way to CCD.

Any of these points could be expanded with considerable backstories; I summarized in a bit of a haphazard way for brevity.

 

[pkr1979]

Thanks for this! When drum scanning boils down to digitizing the film one pixel at the the time I assume this is because the PMT is only able to do one pixel at the time? Which means that if it has to be done quick (haha) it needs to be done wit a drum. But in theory the film could have been digitized on a flatbed using PMTs?

 

[koraks]

When drum scanning boils down to digitizing the film one pixel at the the time I assume this is because the PMT is only able to do one pixel at the time?

Yes,that's correct. A PMT knows no pixels. It's just one signal. So you direct some light into it and record that signal, and call it a pixel. Do that 20 million times in succession, and we call it a 20 megapixel image.

Which means that if it has to be done quick (haha) it needs to be done wit a drum. But in theory the film could have been digitized on a flatbed using PMTs?

Yes, the drum is just a mechanical solution to the problem of having to scan (hence the name) a single beam of light across the entire surface of the film in a matrix fashion. A rotating drum happens to be one of the effective ways of performing that trick, but it's not the only conceivable concept. You could theoretically build a PMT based scanner of a flatbed type using an x/y moving carriage. It wouldn't surprise me if such devices have been in fact built, although no popular devices for scanning photos or negatives come to mind.

 

[Pieter12]

Drum scans have not gone away. Maybe flatbed and camera digitizing has improved, but drum scans are still the benchmark.

 

[calebarchie]

It wouldn't surprise me if such devices have been in fact built, although no popular devices for scanning photos or negatives come to mind.

Have a look at flying spot scanners.

 

[Pieter12]

Have a look at flying spot scanners.

I have employed many, many state-of-the-art and ridiculously expensive telecine (flying spot) rigs during my career. They don't hold a candle to a good drum scanner for still images.

 

[calebarchie]

Didn't say anything about drum scanners, just that this technology existed. I think have about 8 of those now, so I do agree with you.

 

[MattKing]

The last drum scanner I saw in the flesh, needed as much space as a medium sized chest freezer.
That of course included all that accompanied the scanner, but still ....

 

[brbo]

I think that most of the drum scanners still in use today are of the “mini” (desktop) breed - sub 100kg.

Proper ones needed a space of a medium sized room and weighed 500+ kg.

 

[Mick Fagan]

In 1977 our firm sent a delegation to DRUPA in Düsseldorf (Germany) which is the worlds largest printing fair in the world, by a fair margin.

We purchased our first scanner, which was a film drum scanner, and from memory it turned up around 18 months later. I distinctly remember someone saying that the total weight of the pallets was somewhere in the vicinity of 3000kg or 3 tonne. I know that it took up the same space as a modest house. But then again it was set-up in what we now call a clean room.

It used to do four or five colour film separations for offset printing, I seem to remember it could do film up to A0 size, which at the time, was unbelievable, it's also unbelievable today.

I also seem to remember that the total cost, including training, installation and so on, was somewhere around $1,000,000 USD spent over about a 3 year period.

The price did come down, and I think we had 4 or maybe 5 of these machines in the end.

 

[koraks]

In 1977

Remarkable, isn't it? You guys were actually scanning film, at A0 size, at resolutions meaningful for high-quality printing in the late 1970s.
At around the same time or even a few years earlier, my dad was working at a publishing house where he was responsible for transitioning the content-to-press process to the first generation of digital technology. Apparently, the resistance among the more conservative of the printers to this development was massive (and, of course, futile).

 

[pkr1979]

Remarkable, isn't it?

It is.

Are drum scanners produced today?

 

[koraks]

For photographic purposes, I don't expect so.

 

[calebarchie]

You guys were actually scanning film, at A0 size, at resolutions meaningful for high-quality printing in the late 1970s.

Mick is likely referring the film out portion of the drum scanner (which were integrated for separations). Probably an early screen model, which were popular here.

 

[Pieter12]

For photographic purposes, I don't expect so.

Not sure what is meant by that. I would wager drum scanners are still being produced and used in the traditional printing industry.

 

[benveniste]

As wedding photographers switched out of medium format film and people's viewing and display habits changed, demand for drum scanning dropped dramatically. While you can still get drum scans in many major metropolitan areas, it's now decidedly a niche market. Soon used drum scanners were a glut on the market, which placed even further pressure on manufacturers.

It didn't take very long for the same thing to happen to dedicated film scanners in general.

 

[Pieter12]

I don't know much about the wedding photography market today except is pretty much out of control and can be lucrative. What I do know is that most MF film shooters delivered photographic prints and there was no need to scan anything. Drum scanning was for images ending up being printed on a printing press.

There are still film photographers who shoot reversal film or make wet prints and if their work is published, it has to be scanned. Drum scanning is still the standard for most work that can be wrapped around a drum. There are really high-end large format flatbed scanners today for images that cannot be drum-scanned.

 

[Scott J.]

You could theoretically build a PMT based scanner of a flatbed type using an x/y moving carriage. It wouldn't surprise me if such devices have been in fact built, although no popular devices for scanning photos or negatives come to mind.

This would be a real feat to pull off! The limiting factor is the need for near-perfect beam alignment -- i.e., keeping the fiber optic source beam aligned with the RGB beam splitter (which then conveys the separate red, green, and blue beams to their respective photomultiplier tubes). In a conventional drum scanner, alignment is ensured by the fact that these components are all held stationary relative to the film (which is rotating on a drum that moves along the Z-axis). By comparison, the situation is reversed in a conventional flatbed: the film is held stationary while the source beam and/or sensing array move.

It would be really difficult to maintain perfect beam alignment with a drum scanner-style flatbed. You'd need a light source that was capable of moving in perfect coordination with the beam splitter and PMTs. However, the alternative approach (which, on later consideration, is probably what Koraks is suggesting) is an interesting idea -- i.e., a flatbed-style PMT scanner in which "the bed" moves relative to the light source, beam splitter, etc. That's doable, I imagine, but it'd probably be quite large. The movable area of the bed would need to be twice the size of the scanning area along both axes.

The Imacon Flextight scanners are interesting in that they incorporate aspects of both, which is to say: They employ a light source and CCD array of a flatbed scanner, but utilize a movable film carriage similar to a drum scanner. Impressive technology, even today. The Eversmart/IQSmart scanners are probably the closest that a traditional flatbed ever came to achieving drum scanner-level quality, which (compared to the Epson flatbeds) appears to have been the result of better optics and more precise X-Y carriage movement.

 

[koraks]

Indeed @Scott J. - the concerns you bring up also in part crossed my mind, as well as the FlexTight concept (I have one sitting here on my desk at the moment). It's certainly possible in principle, but to the best of my knowledge, it's not been done. As you pointed out, the drum concept was just a very good solution to the fundamental engineering challenge!

 

[calebarchie]

Nobody willing to click on a link to educate themselves?

 

[brbo]

Nobody willing to click on a link to educate themselves?

I did. And I’ll gladly admit that it was educational and that I wasn’t aware of this PMT “based” scanners.

I wonder why they weren’t improved to offer good quality scans for stills (like at least 5k or better, 10k+ dpi)? A scanner with virtually no moving part would be great.

 

[calebarchie]

I did. And I’ll gladly admit that it was educational and that I wasn’t aware of this PMT “based” scanners.

I wonder why they weren’t improved to offer good quality scans for stills (like at least 5k or better, 10k+ dpi)? A scanner with virtually no moving part would be great.

They were. (though probably uses a photodiode rather than PMT now)

A lot of the scanning industry moved towards machine vision and QC inspection. There's simply no commercial incentive to do so.

 

[brbo]

They were. (though probably uses a photodiode rather than PMT now)

A lot of the scanning industry moved towards machine vision and QC inspection. There's simply no commercial incentive to do so.

I have a feeling you are not talking about (photographic) film scanning…

 

[koraks]

I have a feeling you are not talking about (photographic) film scanning…

Indeed.

Nobody willing to click on a link to educate themselves?

That's not a very kind way of putting your point across, if I'm honest. I did in fact click your earlier link, but did not comment on it because while technically interesting, it didn't quite seem to fit the scope of what we're discussing here, both in terms of technology as well as application domains. W.r.t. the latter, you may argue that this isn't quite true given the application of cine film scanning, for which this concept was also used in the past. I would argue that if that's the point you were trying to bring across, you could have spent a few more words on it.