60 CIRCUIT CELLAR® • www.circuitcellar.com

Janua

ry 2

012

– Iss

ue 2

58

T

by Richard Wotiz (USA)

his month, I’ll look at various techniquesused for X-ray imaging, from film to mod-

ern digital sensors. I’ll also tear apart a CMOSimaging sensor and show how it works. It’s avery different type of radiation sensor than theionization chamber I discussed in my last col-umn (“Ionization Detectors,” Circuit Cellar256, 2011). We’ll even have a chance to look atsome X-ray images and see just how useful theycan be.

THE FIRST X-RAYSX-rays have been used to

peek inside solid objects formore than a century. It allstarted in 1895 when Wil-helm Roentgen accidentallydiscovered a beam ofunknown rays emanatingfrom a vacuum tube. He usedthe beam, now known as anX-ray, to take a photograph ofhis wife’s hand (see Photo 1).Since that time, X-rays havebeen used to generate imageson photographic film or pho-tosensitized plates in medical,industrial, and other applica-tions. Like other forms of ion-izing radiation, X-rays, whichare high-energy photons, areharmful to living tissue.Therefore, it’s best to use thelowest possible beam intensi-ty to produce a usable image.But film is not as sensitive to

X-rays as it is to visible light. Making the filmmore sensitive would increase the graininess ofthe image, reducing its resolution. Instead, anX-ray film holder contains an image intensifierscreen made of a special material that convertsX-rays to light. With the screen, you need a sig-nificantly lower exposure to get the same imageas with film alone.The first time digital imaging was used for

X-rays was in the early 1970s, with the devel-opment of the first practicalcomputerized axial tomogra-phy (CAT) scanner, nowmore commonly known ascomputed tomography (CT).It consists of an X-ray tubeand an electronic detector atopposite ends of a rotatinggantry, with a patient placedin the middle (see Figure 1).Data from the detector isstored as the gantry rotates.The platform under thepatient moves slowly (eitherin discrete steps or continu-ously in a helical scan sys-tem). This forms a series ofimage slices, which are com-bined using various mathe-matical algorithms into asingle 3-D representation ofthe patient’s body. The detec-tors in CT scanners use thesame type of scintillationmaterial to convert X-raysinto light as in the image

Digital X-Ray Sensors

Ever wonder how X-rays work? Here’s a detailed look at imaging techniques,from their early inception to today’s advances. You’ll also learn how a CMOSimaging sensor functions.

Embedded Unveiled

Photo 1—This is the earliest known X-rayimage, captured by Wilhelm Roentgen in1895. It’s called “Hand mit Ringen,” or“Hand with Ring,” and shows Roentgen‘swife’s hand wearing her wedding ring.(Source: National Aeronautics and SpaceAdministration)

2012-01-007_Wotiz_Layout 1 12/08/11 4:46 PM Page 60

rigid. The other end of the cable plugsinto an interface module that connectsto a computer’s USB port.Figure 2 shows the key elements of

the system. The imaging sensor in thisparticular unit is a CMOS active pixelsensor (APS) array, though some otherproducts use charge-coupled device(CCD) sensors instead. (Refer to theResources section of this article for moreinformation on some typical sensors.)CMOS sensors consume less power andcan be mass produced more economical-ly than CCDs, which has made thempopular in digital cameras. Early devel-opment work on CMOS sensors in the1990s came from research at the U.S. JetPropulsion Laboratory. They wanted tocreate a compact, low-power imagingsystem for NASA for use in deep-spaceexploration, where minimizing powerconsumption is critical. At the sametime, researchers in Scotland and Swe-den were looking to produce a low-cost,single-chip imaging system. The resultsof these efforts, as well as the steadyimprovements in CMOS processing

technology, gives us the sensors in use today.[1]

There is more pixel-to-pixel variation in the output signalof a CMOS imager than with CCDs, which produces morenoise in the raw image data. Most of this is known as fixed-pattern noise that results from manufacturing process varia-tions between the pixels and doesn’t vary from one expo-sure to the next. The system compensates for this noise, asI’ll explain later. Much of the image-capture logic is con-tained on the sensor chip, so it can acquire the X-ray imagein real-time without any outside help. Since there’s no con-nection between the sensor and the X-ray generator, it can’trely on an external trigger signal. There are extra photodi-odes scattered in several locations around the sensor thataren’t part of the main image array. These photodiodes arealways active, and are used to sense when there is anincoming X-ray beam. This triggers the on-board capture

www.circuitcellar.com • CIRCUIT CELLAR® 61

Janua

ry 2

012

– Iss

ue 2

58

intensifier screen used for film. Early scanners used a pho-tomultiplier tube to convert the light into electricity, whilemodern machines use photodiodes.There is also an indirect digital imaging technique using

photostimulable phosphor (PSP) plates. These are flatplates coated with a material that stores energy in areasthat are exposed to X-rays. The latent image can be readout by scanning the plate with a laser beam, whereexposed areas will emit light when hit by the laser. Thescanner can capture the image and store it in a computer.It’s almost like a laser printer in reverse. This technique isalso known as computed radiography, since a computer isused to control the scanner, but it’s not to be confusedwith computed tomography.Dental X-rays use similar imaging techniques, but on a

much smaller scale. When I go for a checkup, my dentistwill take a set of intraoral X-rays using small pieces of filmcarefully placed in my mouth. The X-rays expose the filmdirectly, without the image intensifier screen that’s used inlarger film holders. That results in a higher resolutionimage, since the thickness of the screen enables the lightto spread out slightly, providing an image that’s not assharply focused.

DIGITAL SENSORSDigital intraoral X-ray sensors were first developed in

the 1980s. There have been many technological improve-ments since then, and many dentists now accept theimages from modern sensors as a suitable replacement forfilm. Photo 2 shows the business end of a sensor. It’s a bitlarger than a piece of dental X-ray film, and it’s also more

Figure 1—The basic layout of a CT scanner. The theory behind the mechanism is straightfor-ward. The complexity is in the computations done to the data in order to create a 3-D image.

X-ray generator

Patient

X-ray detectors

Rotatinggantry

Photo 2—This is a bottom view of a digital intraoral X-ray sensor. It’swell sealed to withstand the punishment it gets from being in apatient’s mouth. Note the extra sealant that’s visible where the cableenters the strain relief.

2012-01-007_Wotiz_Layout 1 12/08/11 4:46 PM Page 61

62 CIRCUIT CELLAR® • www.circuitcellar.com

Janua

ry 2

012

– Iss

ue 2

58

Figure 3—The con-struction of an X-raysensor. Even withtoday’s technology,it’s still amazing howmuch can fit into sucha tiny space.

been processed, it’s stored on the host computer and dis-played for the dentist to examine.Figure 3 shows the construction of the sensor assembly.

The top aluminum layer shields the electronics from outsidenoise, but is thin enough not to interfere with the X-rays.The CMOS sensor will directly respond to X-rays but, likefilm, it’s much more sensitive to light. The scintillationlayer converts X-rays to light so they can be picked up bythe sensor. There are several different scintillation materialsin use, including thallium-doped cesium iodide (CsI:Tl) andterbium-doped gadolinium oxysulfide (Gd2O2S:Tb), alsoknown as Gadox. These are special materials that respond toincoming radiation by producing flashes of visible light

logic to acquire an image. Once the beam stops, the analogimage data is sent the interface module where it’s fedthrough a 12-bit ADC to a microcontroller and stored tem-porarily. Once this process is complete, the sensor acquiresa second image, this time without an X-ray beam. Themicrocontroller subtracts this data from the first image tocompensate for any fixed-pattern noise in the image sen-sor.[2] There can also be a gain error from one pixel to thenext. Each assembled sensor module comes with a manu-facturer-supplied calibration file that is used to adjust thegain of each pixel. The sensor module has a serial EEPROMcontaining a serial number that lets the computer identifythe proper calibration file to use. Once the image data has

Figure 2—The sensor and USB interface. Everything shown on the sensor assembly, except for the RC oscillator, is contained within theCMOS imager chip. The chip outputs a differential analog signal, which is buffered on its way to a 12-bit ADC.

Controllogic

Triggercomparator

Sensorcable

+–

Sensorconnector

USBPort

RAMBuffer

IDEEPROM

Sensor assembly Sensor connector PCB USB Interface module

ADC

MCU

Triggerphotodiodes RC

Oscillator

CMOSImaging

array

X-rays

Cable

Top housing

Aluminum shield

Foam insulator

Reflective layer

Protective coating

Scintillator layer

Fiber optic plate

CMOS Sensor

Circuit board

Bottom housing

2012-01-007_Wotiz_Layout 1 12/12/11 12:31 PM Page 62

www.apec-conf.orgwww.apec-conf.org

SPONSORED BY

2012February 5–9, 2012

Disney's Coronado Springs Resort

Orlando, FL

THE PREMIER

GLOBAL EVENT

IN POWER

ELECTRONICSTM

THE PREMIER

GLOBAL EVENT

IN POWER

ELECTRONICSTM

Visit the Apec 2012

web site for the latest

information!

P

47_Layout 1 11/30/11 3:23 PM Page 1

64 CIRCUIT CELLAR® • www.circuitcellar.com

Janua

ry 2

012

– Iss

ue 2

58

sensor, preventing them from creating additional noisein the image data. X-rays will also degrade the sensorover time by increasing the sensor dark current andshifting the transistor threshold voltages.[3]

LOOKING INSIDELet’s examine the sensor in more detail. They’re a bit too

expensive to sacrifice a good one, but I was able to obtain afailed sensor, thanks to Dr. Catherine Brennan, DDS. Photo 3shows both sides of the sensor assembly. The fiber-optic plateis glued firmly to the sensor. I wasn’t able to remove it, soyou can’t see the front surface of the CMOS imager. Photo 3bshows the back of the circuit board. It’s made of ceramic,which is more solid and less affected by temperature changesthan a more common fiberglass-based material. Note thatthere’s very little circuitry needed for the sensor. The twotiny ICs are an RC oscillator and a logic-gate buffer. You can

whose brightness corresponds to the energy of the radiation.I wasn’t able to find out which material this particular sen-sor uses, but CsI:Tl appears to be the most common amongsensors from other manufacturers. It produces green lightwith an emission spectrum centered around 550 nm, and isone of the brightest scintillation materials available. It’s alsoslightly hygroscopic, so it needs to be sealed with a mois-ture-resistant layer.Once the scintillation layer produces a flash of light,

it’s important to direct it toward the CMOS imagerwithout enabling it to spread out, or the image wouldbecome blurry. The fiber-optic plate handles this with aseries of vertically oriented microscopic light fibers thatare bundled together. It’s glued to the sensor’s front sur-face and guides light directly to the sensor. The platealso absorbs any X-rays that aren’t trapped by the scintil-lation layer. This keeps them from reaching the CMOS

Photo 3a—Top view of the image sen-sor. I removed the aluminum shield,foam, and reflector so you can seethe coating on the fiber-optic plate.The plate is glued to the sensor, so Icouldn’t disassemble it any further.b—The back side of the sensor boardshows some of the smallest compo-nents I have ever seen. The ceramicboard is more dimensionally stablethan a fiberglass board would be.

a) b)

2012-01-007_Wotiz_Layout 1 12/08/11 4:46 PM Page 64

www.circuitcellar.com • CIRCUIT CELLAR® 65

Janua

ry 2

012

– Iss

ue 2

58

Knowledge is power. In the computer applicationsindustry, informed engineers and programmers don’tjust survive, they thrive and excel. For more need-to-know information about some of the topics coveredin this article, the Circuit Cellar editorial staff rec-ommends the following content:

—Networked TimingBuild a Timer With Advanced Planning Toolsby Thomas BereiterCircuit Cellar 224, 2009Thomas’s irrigation timer with advanced planning

(ITAP) is a truly novel irrigation control system. Theeasy-to-use system, which directs user interactioninto a standard web browser, provides useful infor-mation such as watering schedules and zone activi-ty. Topics: Irrigation, Timer, Network, Scheduling,Solenoid, JavaScript

—Temperature Calibration Systemby Brian MillierCircuit Cellar 202, 2007Brian designed a portable temperature meter that

contains a PID controller and a user interface forentering a setpoint. The meter can be plugged into aseparate calibrator unit, which generates stable tem-peratures for sensor calibration purposes. Temperature,

see several small dark rectangles nearthe wire connections, which are resis-tors directly screen printed on theceramic board with resistive ink.Screened resistors aren’t very accurateunless they’re trimmed, which thesedon’t appear to be. Here they are usedas pull-ups and series ESD protectionresistors, where their exact valuesaren’t critical. Most of the other passivecomponents are 0402 size. I can’t imag-ine trying to debug a product with partsthat small! The far end of the sensor’scable is attached to a small printed cir-cuit board (PCB) with a row of edge fin-gers. It just holds some analog and digi-tal buffers, as well as the serial numberEEPROM, and plugs into a connectoron the USB interface module.This particular sensor has a 30 mm ×

20 mm active area and a 40 µm pixelpitch, giving a resolution of 750 × 500pixels. Other sensors have pixel pitchesas small as 19 µm, but, as with digitalcameras, there are so many factors thatgo into producing an image that it’shard to judge quality simply from thenumbers. You can find other dental

sensors with active areas from 24 mm× 18 mm up to 36 mm × 26 mm.The strain relief on the end of the

sensor wire is the most frequentpoint of failure. That’s not surprising,especially for the smaller sensors thatare used primarily in children’smouths. Some of the newer oneshave replaceable cables, and there areeven a handful of wireless sensorsavailable that eliminate the problementirely. But then you need to worry

about monitoring signal quality andreplacing batteries.

TAKING PICTURESPhoto 4 shows an image taken with

a functioning sensor. You can seethat enough of the X-ray beam pene-trates the 0.1” thick brass key toshow some of the groove detail. Ifyou look closely, you can see a faintvertical stripe near the tip that’sslightly lighter than the rest of thekey. That’s a variation in the metaldensity from when the key blank wasformed. It makes it clear how X-rayscan be useful to check metal parts forquality or uniformity.Photo 5 shows the experimental

setup. The X-ray generator was set for8 mA and 60 kV peak (kVP), with anexposure time of 80 ms. The tube cur-rent sets the intensity of the beam,which corresponds to the brightness ofa visual image. The generator voltagedetermines the maximum energy of theX-ray photons in electron volts (eV).This corresponds to their wavelength,or the color in visual terms. The

Photo 4—Here you can finally see someresults. The brass key only partially blocksX-rays, which is why you need to use adenser material, such as lead, to completelyblock them.

Circuit Cellar feature articles are contributed by professional engineers, academics, and students from

around the globe. Each month, the editorial staff reviews dozens of article proposals and submissions.

Only the best make it into the pages of this internationally respected magazine.

Get PUBLISHED. Get NOTICED. Get PAID.

Do you have what it takes?

editor@circuitcellar.com

Contact C. J. Abate, Editor-in-Chief,

today to discuss the embedded design projects

and programming applications

you’ve been working on and

your article could be featured

in an upcoming issue

of Circuit Cellar magazine.

2012-01-007_Wotiz_Layout 1 12/08/11 4:46 PM Page 65

66 CIRCUIT CELLAR® • www.circuitcellar.com

Janua

ry 2

012

– Iss

ue 2

58

REFERENCES[1] E. Fossum, “CMOS Image Sensors: ElectronicCamera-On-A-Chip,” IEEE Transactions on ElectronDevices, 1997.

[2] J. Haystead, “CMOS Sensors Help Dental ImagingSystems Outperform X-Ray Film,” Vision SystemsDesign, 1998, www.vision-systems.com/articles/print/volume-3/issue-12/features/feature-article/cmos-sensors-help-dental-imaging-systems-outperform-x-ray-film.html.

[3] Rad-icon Imaging Corp., “AN06: Detector Lifetimeand Radiation Damage,” 2002.

[4] Planmeca Oy, “Planmeca Intra X-Ray Unit User’sManual,” 2005.

RESOURCESHamamatsu Corp., Dental Applications, http://sales.hamamatsu.com/en/products/solid-state-division/image-sensors/dental-ccd.php.

Saint-Gobain Crystals, Scintillation Products, ThalliumActivated Cesium Iodide, www.detectors.saint-gobain.com/CsI(Tl).aspx.

R. Wotiz, “Ionization Detectors,” Circuit Cellar 256,2011.

Richard Wotiz has been taking products apart ever since hewas old enough to pick up a soldering iron. He’s been helpingothers put them back together since 1991, when he startedhis design consulting business. Richard specializes in hard-ware and software for consumer products and children’s toys.He can be reached at rw601@spiraltap.com.

wavelength λ is given by λ = 1.24/E, with λ in nanometersand E in kilo-electron volts (keV). That gives us a value of0.02 nm for 60 keV. A thin layer of aluminum at the out-put of the generator blocks any lower energy “soft” X-raysthat are produced, since they would be absorbed by theskin and wouldn’t contribute to the captured image.Typical sensor exposures are about one third to two

thirds of film exposure times. The shorter times aren’t aproblem for modern X-ray machines, but older models thatwere designed for film can’t always be set so short. That’snot surprising, since the contact bounce time on the relaysused in older machines would likely be a significant frac-tion of the exposure time. The system used here can be setfrom 10 ms to 3.2 s, with a tolerance of 5% + 1 ms.[4] The

Photo 5—The X-ray machineis ready to take the image inPhoto 4. That’s a samplefrom my growing Orphan Keycollection.

exposure time translates directly into radiation dose, so it’sbetter to use the shortest time possible that produces anacceptable image. That’s called as low as reasonablyachievable, or ALARA. It’s a principle that applies toradiation exposure with a goal of using the minimumdose that will reasonably provide the necessary diagnos-tic information. In general, the higher the dose, thegreater the level of detail in the image. That meansthere’s a tradeoff between how much information is real-ly necessary, and how much risk that information isworth. In our case, the decrease in dose when going fromfilm to a digital sensor is also accompanied by some lossof detail, since even the highest resolution sensor can’tmatch the resolution of film. I’ll have to trust my dentistto know whether or not that difference is significantwhen it comes to looking inside my teeth.Now that you have a better understanding of what goes

on behind the scenes in your dentist’s office, perhaps thenext visit will be less scary! I

Designed a Brand New Device?

Learn to search online databases to see if your device is cutting edge

and may be eligible for a patent by using this new instructional DVD!

info@midwestpatentservices.comwww.midwestpatentservices.com

MIDWEST PATENT SERVICES

2012-01-007_Wotiz_Layout 1 12/08/11 4:46 PM Page 66