EEVblog 1589 - CCD Scanner Array
Вставка
- Опубліковано 1 чер 2024
- A closer look at the scanner array head from the dumpster FAX machine teardown.
00:00 - Linear Scanner Array
04:40 - CCD sensor grouping
08:24 - The optical zebra strip
11:30 - Under the x400 Olympus microscope
Forum: www.eevblog.com/forum/blog/ee...
If you find my videos useful you may consider supporting the EEVblog on Patreon: / eevblog
Or with crypto:
BTC: 33BsprBQNBtHuVzVwDmqWkpDjYnCouwASM
ETH: 0x68114e40ff4dcdd384750500501e20acf3875f8c
BCH: 35n9KBPw9T7M3NGzpS3t4nUYEf9HbRmkm4
USDC: 0x68114e40ff4dcdd384750500501e20acf3875f8c
LTC: MJfK57ujxy55su4XicVGQc9wcEJf6mAoXF
Web Site: www.eevblog.com
Store: eevblog.store/
Other channels:
EEVblog2: / eevblog2
EEVdiscover: / eevdiscover
T-Shirts: teespring.com/stores/eevblog
#ElectronicsCreators #Teardown - Наука та технологія
The circle inside a square in the black area you theorize could be a light sensor is just a fiducial marker, they line up precisely at the breaks between each CCD die.
As for the strip of tiny lenses, it's called a "rod lens array", they're literally just normal lenses that focus the scanned page on the CCD, the image formed combines with other lenses in the array. This technique works without any of the distortions you'd expect because the CCD and page being scanned are at known fixed distances in a flat focal plane, change that however and the formed image with be trash, this is why this type of scanner cannot scan objects that aren't touching the glass like close to the spine of a book page. However not all scanners use the technique, particularly older ones that have a single small CCD sensor die but a complex optical path to shrink the image down, those scanners have a good usable working distance but are often larger and more expensive and pretty much obsolete, the HP 4C is an example of this alternate type of scanner.
Just to add, they (the strip of lenses) are so deep to also add a snoot effect and stop stray light. During the design stage the minimum depth is calculated to keep the price down.
This is really interesting, thanks for the info. The fiducial areas look to be covered with a black tape or something, do you know what this is and what it's for? And why keep the fiducials exposed even after the fact?
They're definitely not "just normal lenses", unless I'm missing something. For one thing, a normal lens produces an inverse ("upside-down") image. Whereas based on the images on Mitsubishi's website (as well as common sense if you think about the fact that all the individual lenses images can't be individually reversed yet produced an overall unreversed image), these rod lenses *somehow* produce a non-inversed, "right-side-up" image. There's some gradient-refractive index/fiber optic wizardry going on here. I don't fully understand what's going on here at all, but "just normal lenses" surely doesn't cut it.
@@TheHuesSciTech I first came across these about 25 years ago when working in the mainframe print industry. TBH, despite going on the Mitsubishi course I never really understood quite how they worked back then. The maths behind it was certainly not simple despite on the surface the device appearing so.
Yeah I remember the 4C, I was amazed when trying it out many a year ago, having the the lid open, scanning the ceiling in perfect clarity while positioned on a table. Never ever came across another scanner that could do that!
1. The lens array is a Gradient Index Lens Array (also called Selfoc®). It images the scanned document plane to the sensor plane - 1:1.
2. The sensor does not look like a CCD - It looks more like a CMOS sensor.
3. Color in scanners is usually achieved by multiple light sources (colored LEDS) rather than by multiple pixels.
This one is 20 years old - these days they bond the CCD dies directly to an FR4 PCB instead of a ceramic hybrid like this one and the light source is one RGB LED at the end with a light pipe.
The scanner business calls the whole assembly a Contact Image Sensor. The “optical zebra strip” is a microlens array. They’re rods with the ends curved to form lenses. CIS sensors are cheaper than the old fashioned linear CCD scanners because they don’t have a conventional optical assembly, but their depth of field is significantly reduced compared a traditional linear CCD sensor.
You may mention it later in the video, I haven’t got there yet, but the sensor has no filter because the colour of the illumination is changed to scan each colour.
Looking at various electronics with a microscope is so cool, such a mysterious and hidden world!
Were you sober or maybe AI, when you wrote that comment? @@p-__
The "optical zebra Strip" is a Kind of Collimator. It makes the light parallel, so that it can create sharp reflections to the sensor. Sorry for the bad english, and greetings from Germany.
Fax scanners typically would have 1728 pixels or bits if you like. As the fax would not have any gray scale, but just black or white, so just one bit being 0 or 1. Making a scan of something like A4 paper, this can be around 1800 up to 2200 lines. That is 1728 x 2200 divided by 8 bits to get bytes... a fax is around 46K bytes.
1728 x 2200 / 8 = 475200 bytes. But (eg. CCITT Group 4) compression may divide that by 20.
Many many many years ago, I attended a presentation with Agfa to do with 'flatbed' (and CCD) scanning; the difficulty was making the illumination of the subject even with 'cheap' components. They suggested the used a feedback system to measure how even the illumination was and then compensate that out.
This weird lens array is an SLA: SELFOC® lens array.
It's to recreate a perfect image of a scanned row of data from the paper on sensors. It has been widely used by Ricoh and Minolta from 1980.
I could watch microscope videos all day
I don't think its a CCD array but a CMOS sensor array - CMOS cells are active like this, while CCD arrays are rather plain since they do charge shifting to readout. CMOS readouts are active readouts which is why there is circuitry per pixel
On the subject of RGB - You more often see monochrome sensors with RGB leds in scanners that don't use CCFL illumination.
Guess it's easier to get more pixels that way
I have an ancient SCSI flatbed color scanner which uses a monochrome sensor and RGBW cold cathode fluorescent tubes!
@@mrnmrn1That's awesome! :D
@@p-__ If you say so lol
Very cool, thanks for sharing!
In my understanding of optics, you always need some sort of lens (or a pinhole) to create an image on the sensor. Maybe several pixels are sharing one lens.
I think one of the roles of the lens rod thingie is to cull out light coming at an angle, so that there's no color bleed and the sensor catches the color reflected from below only.
fun fact, most colour scanners these days still use a monochrome CCD array and get the color information from cycling RGB LEDs to illuminate the page to get the color information from it. i know mine works this way as you get the classic rainbow effect when moving your eyes across it, the same way as a DLP projector using a colour wheel.
My initial thought about the micro lenses is that sometimes a very rapid pre-scan is done using a subset of the sensor array to do the automatic light adjustment for the actual scanning pass. Just a guess, really I'd want to see a cross section to see the light path(s).
You can see something similar to the optical zebra strip in Oki LED laser printers and faxes on the printing side too. I never had caused to take the head apart as I never saw one fail in a printer and it was only available as a whole assembly for the fax machine, separate section of the business handled fax machines, so I don't know if it is a separate component or a permanent part of the LED array.
How difficult would it be to convert it from greyscale to colour? IIRC they still used greyscale sensors, but just used rapidly cycling RGB illumination.
I guess it would mostly be a software issue, once the illumination is sorted out.
It works like a projection lens. The image on the paper is projected onto the very fine pitch CCD line.
It is called a CIS scanner, contact image sensor. A CCD need a lens for focus.
a friend of mine bought a fax machine last month but it was spelled a bit differently and it takes 3 d cell batteries...
also i have an old color scanner (canon lide) with monochrome sensors and it is getting color by rgb led cycling
Hah, they are even using bare die LEDs for illumination now? Ain't nobody can afford a packaged LED, we need to get them each individually bonded instead? Printers are always the most fascinating studies in value engineering. I've taken apart plenty and everytime I notice something new done to lower the cost. Those guys are next level!
if you already need the machinery and process to place the ccd dies, its obvious to just do that for the leds as well
For color scanning, they using 3 different LED light instead, cycling through the 3 primary color, it's cheaper than building 3x the optical sensor.
Assuming it's a cheap polarizer to reduce bleed. Perhaps two layers allows for low/high bit scan.
"[while looking at the sensor array under microscope]... They are all identical!!!", no shit Dave.. Why would you think they'd be different? Haha
I think these are Channels, to isolate the light: the sensors only have to “look“ straight up, and not to the sites.
cheap mfp inkjets use this technology. as a plus, they have a linear "L ee dee" rgb array. pinout is generally easy to reverse engineer. they have seperate pins for r, g and b. youtube has many video of repurposing them. btw, i wonder which portion of the circuit is the actual sensor. purple part or the dot matrix area? thanks for the microscope analysis.
Pretty cool tech in something that now days we consider very obsolete. I'm surprised it even has LEDs and not just a CFL bulb.
My thought is that the spaced sensors are for paper size detection and paper skew detection / correction.
Are those CCD's available to buy? I have a product that would work great with them
This probably explains how Dave got a sunburn on his ars* at the office Christmas party 🥳
I would think that those square non-LED sensors would be paper size detection things. If you know for sure that the paper ends between two sensor, then you wouldn't have to read the pixels beyond that point. I guess this would be done to speed up the scanning?!?
Non-sequiturish: I have an old scanner that mechanically moves colour filters into the sensing path to do a colour scan, thus requiring three passes...
This is what they call a "Contact image sensor" compared to the old school CCD scanner that uses actual mirrors to focus the light on a smaller sensor chip
Greetings. I don't think that's a CCD sensor - it looks more like a CMOS sensor to me with photodiodes each attached to active pixel circuitry, which you can see up above. A CCD would have an overlapping gate structure and wouldn't have all that circuitry in each pixel. The lens array is a curiosity, I'd like to know what the ratio is between the pitch of the lenses and the pitch of the photodiodes.
Oh here comes that nerd talk. 😅
4:52 Probably just wafer size limitation or modularity for one part for many products
I'm vaguely remembering that the fax 'standard' is 1-bit only - so none of those fancy 8-bit greyscales you talked about.... it either is, or it ain't. Not sure if there have been different fax standards across the world and over time though.... so I may be wrong. It happens. Often.
I think there are more modern fax standards that allow some amount of greyscale. Plus this is not just a fax, but also a copier, so it makes sense if it's able to reproduce greyscale, even if the fax part of it doesn't support it.
121 gw eevblog multimeter When will you come again
256 gray. all based off light emission. You wouldn’t believe the optics on some of the more expensive units. basically mirrors and a high end digital camera lense that can work in the same manufacture’s camera. shhh
I am not sure if it is a CCD based sensor. It could also be a CMOS active-pixel sensor, isn't it?
IIRC the fax brochure said it was CCD.
Of course pitch of lenses and sensors doesn't match! Camera in your smartphone has one lens, but way more sensors than one! One lens focused small portion of image to some numbers of sensors.
Can someone please advise, I am designing a high order high pass filter that converts a square wave to sine, I want to make it -100dB at the 3rd harmonic, I see I can achieve this using a Chebychev, (5x op amps, Sallen Key, 10th order), but is there a cost effective solution to this? switched capacitor filter IC's I have found so far are £30+, do you know of anything cheaper that will do the job?
Solution MUST be analogue.
The lightpipes detect the grayscale better.
Well that was a little bit how ya do'n!
Do faxes even do greyscale? I thought they were just black or white and thats it. Using a simple runlength encoding. I mean they probably pimped it right up to play mp3s and have the obligatory blue LEDs by 2003.
My entire life ever since I first saw a LED in my Radio Shack electronics kit back in the early to mid 80s I have always pronounced it “led”, rhyming with bed. It’s like saying “MOSFET” instead of M-O-S-F-E-T.
And “kapoo” instead of CPU.
Thumbs up for LED's, not L E D !
It drove me insane when I heard people saying in some YT videos "I" "N" "I" instead of INI files, or "E" "X" "E" in place of EXE.
ooo finger prints... ;p
Well i would guess like a camera has a lens projecting on a square ccd grid, here we have a number of lenses projecting each one patrialy on its respective grid. Then you need 2 rows of lenses to make sure you don't leave any gaps of the scanned line. After some clever mixing of the ccd outputs from the effect of each row of lensess while moving on their respective ccd sensors i suppose the image is created. But i am most certainly completely wrong since i dont know shit about scanners 😂
👍👍
If LED (Light Emitting Diode) is pronounced as "lead" like the metal, then shouldn't EEVBlog (Electrical Engineering Video Blog) be pronounced "Eve Blog"?
In front of this obsolete technology, I’m like a cave man.
L E D :)
I think everybody should report the sexual bot that spamming the Comments
Does reporting them even help if new ones keep being created, this is problem on many other channels
At this point UA-cam themselves should ban them and start adding restrictions/stricter guidelines so no new ones can be created
Haven't we all reported them hundreds of times at this point and it didn't do anything? The report button is a placebo 😐
@@jessicav2031 Some of them do vanish, and majorty seem to be gone on this video. I tweeted an image of the half a dozen on this video when it was released, I only see two now and I didn't delete them.
WizardTim's pinned comment is 3 days old, but the video itself is only 1 hour old!
He must be a Patreon with early access to the content
Correct, Patreon supporters were given a couple days early access to this one.
@@WizardTim Ah, I see. I thought maybe Dave had got hold of a flux capacitor.
light combo. one light 1, second eyeR...... zebra strip: = light refractor. just come on in & bounce around!! design in key to understanding true basis!
👍
Dave send the strip to Ken Sheriff ( Curious Marc guy's friend ). He can reverse engineer the silicon !!!