To get to the best image sensors in the world, probably.
E2V not only made the CheMin and ChemCam image sensors within NASA’s Curiosity Mars rover, it is also the reason the Hubble space telescope has better eyesight these days.
“Hubble now has our sensors in wide field camera three. It is 10 times more sensitive and there are stars they couldn’t see before that they can see now,” John Kemp, marketing and applications manager at E2V, told Electronics Weekly. “Most of the image sensors in space or in ground-based astronomy are made here in Chelmsford.”
Spacecraft Kepler has found Earth-like planets using E2V sensors, and the Solar Dynamics Observatory has a number of them working at different wavelengths.
Image sensors for space and astronomy have little but the basic principle of operation in common with sensors in consumer gadgets.
“The pixels in handheld devices are typically under 1µm. Ours are normally 10µm to get higher sensitivity through higher surface area,” said Kemp. “High-volume sensor product ranges are focussed on pretty pictures. Ours is focussed on gathering data. In some cases we get one electron for one photon.”
The company’s flagship CCD sensors are made in Chelmsford on a bulk/epitaxial silicon process whose details are the result of 20 years of experience, and remain confidential.
CCDs work by allowing photon-induced electrons to build up in a pixel. To read a particular pixel, its charge is passed along the pixel row in the analogue equivalent of a shift-register. Constraining the charge and maintaining charge integrity is part of the secret sauce.
“Charge transfer has to be very low noise. A typical specification for us is 99.9999% accuracy in each charge transfer. Many competitors have failed with this,” said Kemp. “You will have read-out noise later on. If you can keep that noise below one electron, you have got a fair amount of certainty that there was an electron.”
As well as not increasing noise, the charge transfer process is exploited to decrease noise in a certain type of satellite camera.
Low earth orbit satellites move quickly in relation to the Earth’s surface and can use line-scan cameras.
In these, a long thin sensor – in principle only one pixel wide – mounted across the direction of travel takes care of one image dimension while the satellite’s movement takes care of the other image dimension.
Instead of one pixel, if the sensor is, say, 500 pixels wide, and the clock rate is adjusted perfectly, the image of a point on the earth tracks the charge packet as it is clocked across the sensors surface, contributing to the packet and increasing signal-to-noise ratio many times.
At E2V, imagers are made on 6in (150mm) wafers – allowing 9,000×9,000 pixel arrays to be made when the one wafer is used for one sensor.
The island of Bora Bora taken through an e2v CCD
Imperfect pixels are inevitable in large sensors.
“It is almost impossible to produce sensors without any defects because of defects in silicon wafers. The question is: how much are they defective, and is it too much or too little signal,” said Kemp.
Sensors are shipped as packaged devices, in simple ceramic DIL packages at the small end of the product range, and on ceramic tiles at the large end. “They are always in separate packages, with the package pretty much the size of the silicon,” said Kemp. Devices are either windowed or have an open top depending on customer’s needs. Windows offer the chance to add optical coatings. “We put a lot of effort into anti-reflective coatings on the silicon and glass,” he added.
Chips are wire-bonded around the edge. “We don’t use solder ball yet, partly because of cost at low-volume, and the heritage is wire bond,” said Kemp.
Large arrays that will not fit on a single wafer are constructed from tiled packages with readout electronics under each sensor die.
European space telescope Gaia, intended to map a billion stars after it is launched next year, has a mammoth 1×0.5m 1Gpixel array of over 100 un-windowed E2V CCDs, each of around 4,500×2,000 pixels.
“We finished supplying them 18 months ago. Astrium is tiling them together,” said Kemp.
Gaia is a remarkable piece of engineering. Almost the whole thing – mirrors, CCD support, and the entire chassis – are made from sintered silicon carbide by Boostec in France.
NASA took this photo from Hubble, through an E2V CCD.
Along side CCDs, E2V supplies CMOS image sensors. The Korean space agency’s geostationary ocean colour imager carries an E2V CMOS sensor.
These are made in outside fabs. “We happily provide both CMOS and CCDs. CMOS requires much finer feature size,” said Kemp.
Silicon sensors, CCD or CMOS, can work from 300nm, through the visible spectrum, to 1,000nm. Outside these limits coatings can be applied to change the wavelength of light. For example x-rays can be viewed using a scintillator layer on the sensor.
As well as sensors for space and astronomy, the firm makes products for science – for spectral sensing for example.
Some projects can afford their own mask sets for completely chips. Others re-uses existing designs.
Qualification testing rather than device fabrication is the major part of sensor delivery.
“A project might be to supply five flight parts. We make 20 in all including parts for destructive testing. It is the qualification that takes the hours,” said Kemp. “It has been said that we supply a filing cabinet full of paper and a few free image sensors.”
Business arrives in two ways.
“We don’t have to do cold calling. Our track record and size mean we are a better bet for our customers,” said Kemp. “For Gaia, the people who started talking to us were ESA. Later, the mission primes were selected and told they had to work with us. For other missions, the mission prime selects the supplier.”
Good technical performance used to be the only thing that customers cared about. “It used to be all about the technology. Scientist got better data from our sensors than other people’s sensors,” said Kemp. “Now it is as much our ability to supply programme management and qualification.”
And the firm is branching out into providing ancillary equipment.
“We have large order for a telescope; for the sensors, the flat metal mounting plate, and the cryo-cooler. We are supplying four times the value – 1x in the sensors and 3x in the ancillaries.”
For image sensors, E2V is competing with firms including Dalsa in Canada and Cmosis in Belgium. At the sub-system level, particularly with mechanical sub-systems, it is more likely to be competing against the customer’s internal facilities.
The Rutherford Appleton Laboratory in Oxfordshire is developing electronics for E2V, which E2V will qualify and sell into space programmes.
“We also fund a group at the Open University: the Centre for Electronic Imaging,” said Kemp.
Recently the firm received £3.6m if regional development funding, which it is matching. “The vision is to double our space imaging business, to improve CMOS processing, and to improve coating for more sensitivity,” said Kemp.
E2V employs approximately 1,500 people: 1,000 in Chelmsford and the rest between Grenoble and Silicon Valley.
As well as image sensors, it makes vacuum devices for RF and microwaves.