A Long And Incomplete History of Scanners

I absolutely did not intend for this to be an hour and forty minutes long.

Chapters:
00:00:00 Intro
00:02:37 The Knowledge Vacuum
00:05:35 The Questions
00:06:06 ACT ONE= The Grey Scanner For Computer
00:09:50 The Ports Conundrum
00:14:37 Hardware setup & the Video Conundrum
00:17:49 The scanning experience
00:20:38 Quality Or Lack Thereof
00:24:30 The purpose of scanners in 91
00:29:13 Windows desktop publishing in 91
00:34:53 The state of printing in 91
00:36:24 Memory, 1-bit color, and halftones
00:50:56 Hardware teardown
00:58:47 ACT TWO= The history of scanners
00:59:24 Pre-consumer flatbeds
01:05:35 The consumer scanner explosion
01:12:10 What the hell was the ScanJet Plus??
01:15:55 The greyscale days
01:20:44 Dawn of color
01:23:52 The modern scanner takes shape
01:27:20 The Color Conundrum
01:31:20 Slow decline of greyscale
01:35:35 Forty years of scanning
01:37:20 Errata
01:39:14 Outro

[TheMaristBoy]

Obligatory "I'm glad you're experimenting with short-form content" comment

[vwestlife]

FYI, RS-232 serial ports did increase in speed as well. The ones in most 1980s computers maxed out at 9600 bps, but by the early '90s, faster UARTs allowed serial ports to increase to 115, 200 bps, fast enough to transfer images from a low-res scanner or digital camera, and even for postage-stamp-quality video from an early webcam or video capture device.

[WeGoToMars]

Don't worry, 100 minutes is a perfect size for a video. They shouldn't even make videos smaller than that!

[chrishacker]

Apologies for the essay but as someone who works professionally in cultural heritage digitization, I greatly enjoyed this video and the effort put into filling some gaps in the history. I know the focus was intentionally on consumer technology (despite many interesting asides into the expensive professional stuff), and I enjoyed the flashbacks to stuff I used as a kid in the 90’s, but of course, today’s pro stuff is the consumer stuff of the future (often in a different and usually smaller form), so I thought you might be interested to know what’s happening in at least one sector of high end “scanning” today (we avoid the term scanning because we don’t use line-scan devices anymore, it’s all instantaneous capture, so we say “imaging/image” or “digitization/digitize” although those aren’t as catchy). Others have touched on some of this in earlier comments a little bit.

One important point that has been mentioned is that scanner manufacturers didn’t stop improving them 10-15 years ago because the problem is solved and you just can’t get any better - I don’t know for sure why they didn’t keep developing them but given the timing of when flatbeds got as good as they ever got - around the same time that drum scanners and pseudo-drum scanners like the Imacon film scanners stopped being produced - I would guess it’s simply because there was no broad need for scanning anymore in the big-money professional printing context because so few things were not born-digital at that point (e.g. professional digital photography had gotten good enough to replace darkroom photography).

So the scanners stayed the same because there wasn’t R&D money for them anymore, and because of some of the advantages they have, like not having a bayer filter, were still superior to digital cameras for a while so updating the technology maybe wasn’t necessary at first. But they are not actually very good at all in the 2020s sense of good digital image quality. The best digitization now is done with cameras, primarily Phase One 100- and 150-megapixel camera backs with specialized reproduction lenses (though you absolutely can get excellent results with less expensive cameras). Those setups cost within the realm of a $90, 000 room-filling solution from the 80’s mentioned in the video… if you don’t adjust for inflation, i.e. you might spend 90K today for a Phase One digitization setup. So actually, especially considering the cheap computers and much less labor time needed, they’re actually significantly cheaper. I am not sure if any publishers etc. actually have them, though, because most stuff is born-digital, and any occasional need for digitization can be done by a vendor.

But museums, libraries (I work in an academic library at a major university), archives, etc. (cultural heritage institutions) have them for digitizing all the old stuff, and image quality is absolutely critical - a big part of what we’re doing is preservation in case the originals are lost. The ideal (which I’m not sure we’re meeting yet, but getting close) is complete transparency - in the sense that there’s no visible trace of the optics or the sensor - no “character” (as is often desirable for creative photography, e.g. using vintage lenses), just pure and accurate reproduction data.

Even the best modern flatbeds unfortunately have a lot of character, though it isn’t usually obvious without a direct comparison. If you do compare a scan from a top of the line modern flatbed to the same thing imaged with a Phase One system or similar, it’s night and day - even if you do 1200 or 2400 ppi on the flatbed and compare that to a 600 ppi Phase One image (600 ppi is the standard and the most common lenses used are optimized for that), because the lenses in the flatbeds just aren’t very good. This is even considering that the Phase One cameras have bayer filters - the color rendering is actually better too anyway (how, I don’t know specifically).

Lenses that can do what a flatbed scanner lens does are a difficult optics problem and that they’re as good as they are for the price is seriously impressive. Better lenses designed for line-scan sensors definitely exist - that’s what’s inside an Imacon, which is a line scanner, for an obvious example - but for way more money. You’d probably have to quadruple the price of the already expensive high end flatbeds (like the v850 that was shown in the video) to put a significantly better lens in there (the lens is why flatbeds aren’t very good for smaller film formats).

So there is kind of a market limit to flatbed quality, I guess. But, interestingly, there are high-end sheet-feed scanners that produce better results than flatbeds both in color reproduction and optical quality. One we use at work is made by Widetek; these are big scanners designed for blueprints and maps (which is what we use it for). Sheetfeeders have an inherent flaw of streaking, so we use the Widetek scanner for what we might call “access” images rather than “preservation” images, but again, streaking etc. aside, color and resolution are very good. I wanted to mention that because the feed mechanism is practically identical to the one in the Gray Scanner, with rollers, glass, etc. looking extremely similar, and you can get a more “transparent” reproduction out of it than with a flatbed - they’ve clearly been doing the R&D that Epson et al. stopped doing. Widetek does actually make a flatbed, which I haven’t tried but I suspect it may be better than most alternatives.

By the way we also have high-end automatic-feed sheet feeders - image quality is surprisingly poor but they’re made for text documents primarily. They cost like $5-6000 but you can stack hundreds of pages on it and can expect that it will scan all of them correctly without double-dipping or jamming.

Regarding drum scanners, we now have technology that can meet or beat them for film (for medium and small film formats anyway, for now). Pixel-shift images from a 100-megapixel Fujifilm GFX camera give you a 400-megapixel image and overcome the effects of the bayer filter, giving you true color for every one of those pixels. If your lens is good enough (not an easy feat) you can go well beyond 7000 ppi (the resolution of drum scans cited in the video), and it can be very important to do so to avoid “grain aliasing”, which depending on the specific film grain characteristics can rear its ugly head in the 4000-7000 ppi range especially.

Anyway, the old overhead cameras shown a couple times in the video almost certainly were line-scan devices since large sensors wouldn’t have been available, so those were still “scanners”, and many of the similar-looking devices today (some of which have been mentioned by others in the comments) operate in the same way, for the same reasons - large sensors are still expensive. You could put a modern cell phone camera or similar in those (Samsung apparently has a 200-megapixel cell phone camera coming out) but you very quickly run into limitations basically imposed by optical physics - it just isn’t possible to make a sharp enough flat-field lens at that size for a small sensor. Most of those devices don’t have very good lenses as it is, though high end ones do have decent lenses. But I think devices like this are going to be the future of consumer scanning. Flatbeds still have advantages for now but compact/collapsible overhead camera style devices will be better for most uses once they get to the same cost/quality point that flatbeds are at, and at the higher end of the consumer market for such devices they will exceed flatbed quality.

For digitizing film, that is already the realm of overhead camera capture at the high end and for many hobbyists. It doesn’t take a particularly high-end digital camera to be able to produce better “scans” than what the film lab can give you with their lab scanners (and Imacon type scanners), though lab scanners have a huge advantage in automation - for now.

By the way, for archival film, you can’t put that in a drum scanner - the risk may be small, but because of the fluid mounting required, there is a risk especially with old already-deteriorating film (and that descriptor applies to almost everything older than the 1980s), not to mention how slow and labor-intensive it is for large collections… and the fact that it’s possible to meet or exceed the quality relatively easily with a single-capture overhead camera setup where the film can be mounted safely and quickly anyway.

(Dedicated film scanners are way better than flatbeds for film because they have better lenses that can adjust focus, but even those will be beat in quality by good overhead camera capture.)

Right now, high quality overhead cameras are expensive and somewhat esoteric. Flatbeds still are a thing for a very good reason for the consumer market. It’ll be interesting to see how exactly the consumer flatbed era will come to an end in the next 5-10 years, though.

[willdchild]

A history of scanners? A technology that has never once interested me or ever made me wonder exactly how they work? But, CRD is explaining it to me in a hour forty minute video? Hell yeah BROTHER!!!! Count me in.

[JustinBritt-rp]

From an interview with Eugene Gordon:
"I led the group that developed the CCD. We understood its importance for color cameras and understood its importance for facsimile machines. I asked Hugh Watson to make the first fax machine that utilized a CCD device. It was a flatbed scanner very similar to devices seen today. I tried to convince AT&T to get into the fax machine business, but they wouldn’t do it. "

Watson's patent US3867569 was filed in 1974, so it predated Kurzweil's reading machine.

[ambostralian]

Spending half an hour troubleshooting just to find out there's no disk space is literally my proxmox journey.

[steveg]

dude I started out as a Computer Service Technician in 1980. I lived the evolution of your content and it brings back so many memories.My first task was aligning 8 inch floppy drives for compatibility using an HP Scope. I am retired at 68 years old now and boy has technology changed! Thanks for explaining so much to the new nerds.

[RobColbert]

"Gray scanner for computer, " needs to be a t-shirt, and is going to be what I now call all old peripherals.

[lupinzar]

Scanners are unsung heroes. My wife had a major medical incident last year and it produced a ton of paper. I went on ebay, grabbed a decade old Fujitsu full duplex ADF scanner for around $60-$70. I had to replace the rollers which was a little more money and a mildly frustrating afternoon, but it was worth it to scan all that paper in a few hours instead of trying to do that with the basic thing on my all-in-one printer.

[humidbeing]

When I was a kid, in the early 90s, we had an off brand flatbed. I was fascinated with it -- not because it was good, but because it was bad. The proprietary ISA card only recognized the scanner after randomly power cycling for 20 seconds. Among other frustrations. I would stand at the magazine rack while my mom bought groceries and pour over computer magazine reviews trying to answer the question "are all scanners like this?" The articles never gave much information besides specs like DPI. It was maddening. You're doing real fucking work here. Thank you.

[FooneTuring]

I'm watching this while operating a scanner that costs 60k$+ and takes 2.4 gigapixel images. So I'm questioning how intrinsically boring they are! :)

[BrianRRenfro]

"No one wants to go back to the past:
My brain: "He's gonna take you back to the past, to play with scanners that all suck ass..."

[dd]

There’s at least one “photographic” application for line-scan sensors: they are used to image boxes on conveyor belts, another application where the subject moves consistently in one direction. A major advantage is that they can image the bottom of a box through a narrow gap in the belt.

[lfla]

Scanners didn't change my life.
TWAIN drivers changed my life ... FOR WORSE.
VUESCAN actually changed my life.

This little software packed ALL the scanners - all the 20 or 40 THOUSAND models of them - into a single driver that TRAMPLES over windows signed drivers, installs them anyway, and lets you use features that you didn't know you have, but it knows all the 6 models of CCD that the majority of table scanners had, and talks to them directly.
I had a Canon usb-only scanner, no power cable. VUESCAN knew it could scan 2 independent photographs at once. Canon never shipped that feature on my scanner! But the CCD was there!

Whoever did that program was PISSED, and I love him for it, all the 49.95 it cost was WORTH IT. And it ran your stuff on any windows, 10 inclusive.

[Gledster]

"Good luck with that, I'm sure it will help you in life."

It did, the video educated, entertained and informed me whilst I cleaned my kitchen.

Great video, as always. Thank you.

[Jake]

I love seeing words like "Quiet" and "Silent" on old dot matrix printers. In my experience (albeit from ancient childhood memories of the early-mid 90s), "Quiet" meant "you can technically operate this machine without ear plugs and not suffer permanent hearing damage".

[mrbusdriversir]

Around 1995, I bought a UMAX 2200 SCSI scanner, supposedly on sale, for $1200. I remember spending hours playing with it, scanning pictures on it, and editing them. Because of its size and not having enough desk space, it found its way to the garage, where it sat for several years until it was sent to Goodwill. Thanks for bringing memories back.

[AnonymousFreakYT]

10:55 - The fun part of ISA is that until it became standardized as "ISA", _it ran at the CPU speed._ That means an original IBM PC ran it at 4.77 MHz, and the actual throughput was under 1 MB/s; while some early AT/286 clones might run it at 12 MHz. That accounts for the transfer speed differences. It wasn't until clone makers decided to actually standardize it that it settled on a fixed bus speed that accounted for the 8 MB/s transfer speed (on the 16-bit bus at 8.33 MHz.)

Note that IBM never standardized it while it was the primary bus. AT ran it at CPU speed, as did the ISA 8088 and 286 PS/2 models. Only after they abandoned Microchannel in the Pentium era did they adopt the clone-made-standard fixed 8 MHz bus. (And really, by the post-PS/2 era, all PCs were just clones, even when they were from IBM.)

[SlowMenThinking]

That is built like a fax machine of the era.... I used to work on them back in the day. Including the exposure correction hump thing in front of the senor at 54:48 it is there to correct the light balance from the centre to the edge of the scan otherwise the edges would be dark on the scan...