EEVblog #637 - Omni Directional Laser Barcode Scanner Teardown

[rob77]

There never was a 53rd  week of 2005.. How can that Atmel chip have been manufactured there?   

actually the week numbering is not consistent all over the world.. there are at least 2 definitions  of the 1st week of the year (i know about 2, but there might be even more ).  so it easily could be that Atmel is using a different week numbering than you

[PA0PBZ]

I don't think there is a lot of processing for the '5 angles', the output of the photodiode is just a 01010111011 bitstream and whenever the right laserline hits the barcode it will produce a readable output just like the handheld one. The only thing they have to look after is a timeout whenever there is a positive detection because I can imagine that there's more than one scanline hitting the barcode.

[rob77]

I don't think there is a lot of processing for the '5 angles', the output of the photodiode is just a 01010111011 bitstream and whenever the right laserline hits the barcode it will produce a readable output just like the handheld one. The only thing they have to look after is a timeout whenever there is a positive detection because I can imagine that there's more than one scanline hitting the barcode.

actually the photodiode produces a constant stream for all scan lines - so you end up with a stream of 0s and 1s for all scan lines the reader does..  now consider that the barcodes have NO standard sizes - just the ratio between the "strips" in the code matter - so you don't know how wide the 0s and 1s are supposed to be....and to make it even worse, the distance of the code from the reader introduces further variations... angle between  the barcode  a scan line is introducing even more.... so actually it's a LOT OF PROCESSING for that little ARM there to find the pattern than matches one of the bar-codes

[EEVblog]

There never was a 53rd  week of 2005.. How can that Atmel chip have been manufactured there?   

Did I really say that without engaging my brain? 

[EEVblog]

the mysterious part is something called reflective interrupter. Like this
http://www.osram-os.com/Graphics/XPic0/00083379_0.pdf/SFH%209202,%20Lead%20(Pb)%20Free%20Product%20-%20RoHS%20Compliant.pdf

Ah, awesome, thanks.
I was spot on it seems.

[EEVblog]

actually the photodiode produces a constant stream for all scan lines - so you end up with a stream of 0s and 1s for all scan lines the reader does..  now consider that the barcodes have NO standard sizes - just the ratio between the "strips" in the code matter - so you don't know how wide the 0s and 1s are supposed to be....and to make it even worse, the distance of the code from the reader introduces further variations... angle between  the barcode  a scan line is introducing even more.... so actually it's a LOT OF PROCESSING for that little ARM there to find the pattern than matches one of the bar-codes

That's what I suspect.
I'd love to see the algorithms used.
It seems that all the barcode types can be turned on and off in software as well.

[EEVblog]

Dave, do you want the hand held version of that scanner?

I wouldn't mind the exact same 2D DS9208 model for comparison.

[mikeselectricstuff]

My guess is the chip on the right is an analogue front-end, and the other is a barcode processor - they make gazillions of these, so worth doing an ASIC to do the complex task of recognising many different codes coming in at a variety of rates.
I think you'd struggle to reliably read all the various barcode formats even with an ARM, and bear in mind previous incarnations would have probably used an older processor, but with the same ASIC to do the hard work.
I would think the ARM is probably more about supporting things like fast RS485 networks with multiple readers. 

[mikeselectricstuff]

Dave, do you want the hand held version of that scanner?

I wouldn't mind the exact same 2D DS9208 model for comparison.

2D code scanners are basically just cameras with an area illuminator. The only special features are a narrowband optical filter - it's also possible that they might modulate the illumination synchronous with the camera to eliminate other light sources.

[EEVblog]

I would think the ARM is probably more about supporting things like fast RS485 networks with multiple readers.

They already have that fairly beefy H8S processor that could do that, surely
http://www1.futureelectronics.com/doc/RENESAS%20ELECTRONICS/DF2215RTE24V.pdf

[N2IXK]

Those reflective optosensors have been around for quite some time. That one looks nice and robust.  I have replaced a few defective ones on equipment over the years, used in this type of application (position/speed feedback sensors on VTR tape reels and color wheel sensors on DLP Projectors) which were tiny little things made by Sharp. They definitely seem to "wear out" with age, with the intensity of the IR output falling off dramatically.

[Stonent]

I haven't seen the video yet but configuration of bar code scanners via a bar code config page is fairly common. 
For some I've set up before, you want to turn off certain formats to prevent the user from accidentally scanning the wrong code.  If you have someone scanning boxes for a serial number you would want to disable UPC to prevent an accidental entry of that as a serial number. This is common when you have someone on a forklift that is scanning boxes that may have numerous labels that could be accidentally scanned.

Another common configuration is preamble and postamble being set via barcode. I've seen that used for serial numbers where you have the items numbered already and may have purchased them in advance, but there's additional information that is configuration specific that makes its way into the serial number. As for the postamble, commonly it is a carriage return, but may need to be a tab, or a comma depending on the application.

For the preamble, I've seen that used to append a date code to the beginning of the serial number, or a number that indicates to what year's regulatory compliance it adheres to.

Lets say your company produces control modules for a product that you also build. Those modules are all 100% electrically identical but one goes down a line that is in your current product that adheres to a regulatory standard enacted in 2012 and the other goes down another line that builds aftermarket replacement assemblies for the version that only had to adhere to a 2010 standard. (based on firmware).

You configure your scanner to append a code to the beginning of the scanned in serial number so when it gets scanned it sees it's a 2010 device and loads the 2010 firmware.

Anyway, that's one way I've seen it.

[Tothwolf]

Do not look into laser with remaining eye.

Any other laser geeks cringe when Dave stopped the rotating mirror with a spudger? I was surprised it still scanned at all with the scuffs in the mirror

A tip for newcomers to laser optics... The best thing I've found for cleaning these sort of optics is 190 proof Everclear (grain alcohol). I don't know how anyone could possibly drink the stuff, but it's dirt cheap and much easier (and cheaper!) to obtain than lab-grade ethanol. Isopropyl alcohol is no good for laser optics since it has too many impurities and leaves a residue on the optics that distorts the beam.

[SeanB]

Dave, do you want the hand held version of that scanner?

I wouldn't mind the exact same 2D DS9208 model for comparison.

No. it is the baby PC with an integrated scanner, basically a MSDOS computer shoehorned into a small multimeter sized package, like those Teklogix barcode readers you see being used for inventory control. It had a few drop tests that killed the scanner part, and the battery died from the charger being a dumb one, but it still boots to the dos prompt with an external 5V supply. It was used to do a scan and count for inventory management, but the back end moved on and no longer supported it, and it used a hard to find serial port as well.

[zapta]

Very interesting video, especially the intersection of electronics, mechanics, optics and daily life objects.

Dave, are the driving signals of the motor and vibrating mirror in sync? (I would think so since the light pattern looks consistent).

Also, have you tried to look with oscilloscope at the output data stream of the receiver? Would be interesting to correlate with the input pattern (e.g. a white paper with one circle in the middle) in different lighting conditions.

[hikariuk]

There never was a 53rd  week of 2005.. How can that Atmel chip have been manufactured there?   

2004 did though and days 6 and 7 of Week 53 2004 were the 1st and 2nd of Jan 2005.  It's possible it was made on one of those days and whatever puts the week of manufacture on the chips uses the ISO week number and whatever the current year is, rather than whatever is the ISO week and ISO week-year.

[alper.y]

Does single directional (line) scanner use non-mechanical laser sweeping? Beacues there are no moving parts, like mirors as far as I know. Has it array of sensors? There is a need for sweeping whether mechanical or non-mechanical for proper operation, right?

[hikariuk]

Does single directional (line) scanner use non-mechanical laser sweeping? Beacues there are no moving parts, like mirors as far as I know. Has it array of sensors? There is a need for sweeping whether mechanical or non-mechanical for proper operation, right?

I think laser based readers either have to use a rotating prism or wobbly mirror.  If there's no moving parts at all it has to be using some other kind of reading technology, like CCDs.  The red light from those is probably just LED illumination.

[bktemp]

I think both variants exist:
The other scanner Dave has, with the singe line is based on the same laser scanning mechanism (as far as I know), but without the rotating mirror beeing in different angles and without the wobbling mirror.
The other variant I have seen, has flashing leds to light up the whole code at once and a ccd or cmos line sensor to read the line image.

The main advantage of the laser version is the absence of any optics in front of the sensor. Therefore it has no limitations in the depth of focus since there is no focus point. The only limitation is the amount of light beeing reflected back to the sensor.

[alper.y]

The other scanner Dave has, with the singe line is based on the same laser scanning mechanism (as far as I know), but without the rotating mirror beeing in different angles and without the wobbling mirror.

So, what do you mean by same laser scanning mechanisim but without rotating mirrors? Mechanical scanning but without any rotation? Sorry, I didn't get it. 

[mikeselectricstuff]

More modern single-line scanners tend to use single line CCDs, which avoids the need for any mechanical parts. Proformance of these is impressive - I've seen one read a standard sized UPC code from about 3 feet away.

[SeanB]

The rotating or wobbling mirror ones use a large area photodiode and scan the laser light so only a small dot is illuminated at a time, and filter out all other light other than the laser light, so can get very good recognition with no degradation from focussing errors. Ones that use a line scanner typically use a barrel shaped lens to focus light onto the CDD, and typically are focussed at infinity for the lens, so will work so long as the barcode is more than about 5cm away for most small cheap plastic lenses. They work so long as the barcode is smaller than the image focussed on the line CCD, and till the space between the smallest feature ( a line representing a 1 bit, either black or white) is less than a single pixel in width. This can give a large range, as you can get 2k line CDD imagers in a small single chip package about a DIP18 long. those are also used in fax machines and digital scanners if they only do a monochrome scan.

You also get 3D scanners which are basically a digital camera and software, and these do a good job on damaged barcodes, as they can often use the undamaged parts of the barcode to recreate the damaged sections, like on a barcode that has a faded part and a good part but which a line scanner cannot read.

[rs20]

I don't think there is a lot of processing for the '5 angles', the output of the photodiode is just a 01010111011 bitstream and whenever the right laserline hits the barcode it will produce a readable output just like the handheld one. The only thing they have to look after is a timeout whenever there is a positive detection because I can imagine that there's more than one scanline hitting the barcode.

actually the photodiode produces a constant stream for all scan lines - so you end up with a stream of 0s and 1s for all scan lines the reader does..  now consider that the barcodes have NO standard sizes - just the ratio between the "strips" in the code matter - so you don't know how wide the 0s and 1s are supposed to be....and to make it even worse, the distance of the code from the reader introduces further variations... angle between  the barcode  a scan line is introducing even more.... so actually it's a LOT OF PROCESSING for that little ARM there to find the pattern than matches one of the bar-codes

The thing is, all of these different factors just combine into one scale factor: how long a strip is in microseconds (or bits). It's not like the barcode scanner has a complex 3D model specifically figuring out barcode size, barcode distance, etc, etc, it all simplifies down to a scaling constant. It's moderately more complicated than that because perspective might make the strips slowly change size over the length of the barcode, but still, detecting whether a string of zeros is about twice or half as long as a following string of ones doesn't seem all that difficult or advanced or "a lot of processing".  The only thing that's difficult here is the sheer data rate being handled, not the complexity of the algorithm. I claim.

[mrflibble]

Interesting teardown, loved it.