The ULIS 02-05-01 microbolometer - An insight into the technology by Fraser

[Fraser]

I thought some readers might be interested in a little insight into the workings of the ULIS 02-05-01 microbolometer. This is a Circa 2006 A-Si microbolometer used in several thermal cameras of the period.

Microbolometers can be a bit of a mystery device as they historically have been subject to NDA's for datasheets and pin-outs etc. I am not about to publish the ULIS Confidential datasheet so do not get too excited. I will explain a little about its needs however.

The ULIS 02-05-01 is a 160 x 120  35um pixel thermal imaging FPA in an industry standard case format. The sensor module case contains the A-Si microbolometer die and a thermo electric cooler/heater. The interior of the module case is under vacuum and an active Getter is used to absorb contamination when manufactured. The active Getter may be re-activated when considered necessary due to age related vacuum contamination. Details of this operation are provided in the ULIS datasheet.

The sensor die receives the thermal scene energy through a Germanium window that is soldered into the face of the module case. The leads are sealed into the module case using a glass seal technique. Vacuum failure can occur due to failure of the window solder or glass lead seals. The module must not be exposed to twisting or lead bending if such is to be avoided. The loss of vacuum will result in serious degradation of the microbolometer performance. Vacuum contamination can also occur due to out-gassing of components used within the sensor assembly within the module. More specifically the TEC. Such vacuum contamination can reduce the sensor performance over time.

A high level description of the sensor die follows.

The microbolometer pixels are read in rows and columns. The pixels output is passed through a "Skimming" stage to extract the required pixel data and is then presented to a Capacitive Transconductance Integration Amplifier (CTIA). The output of the CTIA is sampled and passed to the multiplexer that feeds the 'Video' output amplifier. The video output may then be passed to an ADC for conversion to a microprocessor friendly format.

The reading of the pixels is a synchronous event carried out by a dedicated sequencer that forms part of the ROIC.

To operate, the ULIS 02-05-01 requires supply rails, bias voltages and clocks. Without these being correct it cannot function, and in some cases, may be damaged.

The supply voltages are nothing special except that they must be accurate and low in noise content. Noise on the supply rails to a microbolometer can severely impact the sensors performance. For this reason, Analogue and Digital supply rails are kept separate.

The bias voltages are very important as they effectively set the working point of the microbolometer. There are five bias voltages used in the Microbolometer, namely, VSKIMMING, VDET, VBUS, VEB and VBUS. All must be set to the correct voltage for satisfactory operation of the microbolometer sensor.

It will be noted from the attached documents that the microbolometer pixels within the sensor come in two distinct types. 'Active' and 'Blind'. The Active pixels are the ones that are presented with the thermal scene through the Germanium window. The Blind pixels are fewer in number as one Blind pixel serves a whole column of Active pixels. In a 160 x 120 sensor there will be only 160 of them. The Blind Pixels cannot see the thermal scene and are truly blind to the outside world. They are used as a differential reference to extract the required thermal scene data from the active pixel whilst ignoring the standing current and some noise elements tat may be present.

VSKIMMING : Bias voltage applied to the Blind Microbolometer
VDET : Bias voltage applied to the Active Microbolometer
VEB : Bias voltage applied to the Blind pixel FET gate
VFID : Bias voltage applied to the Active pixel FET gate
VBUS : Bias voltage applied to the CTIA

The timing signals required to derive the sensors sequencer that reads the pixels are the Line clock, Frame Clock and Pixel Clock. On the ULIS 02-05-01 these three signals are provided by the host camera. In more recent microbolometer sensors, the ROIC is fed with a Master clock signal and the three sequencer clocks are generated inside the ROIC. Later sensors also incorporate serial communications and configuration capabilities within the ROIC. 

Full disclosure

My sincerest thanks to a fellow member of this forum who has assisted me in better understanding the internal workings of the ULIS A-Si microbolometer. He did this whilst ensuring he did not disclose corporate confidential information to me. I have searched for this information on the public internet and not breached any NDA's or trusts in compiling this post.

Fraser

[Fraser]

More pictures

In the last picture the vacuum port is clearly visible. It is crimped off and sealed after the correct Vacuum level has been established.

Fasser

[Fraser]

A very insightful document on the ULIS microbolometers Circa 2006 is attached as the first attachment below. This was found in the public domain.

Document found here:

http://www.wat.edu.pl/review/optor/14(1)25.pdf

Plus some helpful images extracted from other documents.

Other interesting microbolometer sensor documents are available from SPIE.

Fraser

[nishul]

Although I don't understand every bit, but this has been a nice read! Thanks for sharing!

[Bill W]

The interior of the module case is under vacuum and an active Getter is used to absorb contamination when manufactured.
...
Vacuum failure can occur due to failure of the window solder or glass lead seals. The module must not be exposed to twisting or lead bending if such is to be avoided. The loss of vacuum will result in serious degradation of the microbolometer performance.

That is a bit of an understatement. I have heard that the imaging response of an 'up to air' microbolometer is less than 1% of the normal value.  Put another way, 100x the noise !

Bill

[Fraser]

Wow!

Thanks Bill, I was not aware that Vacuum loss was so devastating to performance. I have not witnessed total vacuum loss on a camera yet.

Being a collector of the older Industrial type cameras I do wonder how much contamination and loss of vacuum the circa year 2000 microbolometers have suffered. Even my 1997 FLIR PM570 still provides an excellent image. I presume it is pretty obvious when vacuum contamination or partial loss has occurred. The symptoms of such failures do not seem to be documented in the public domain which is a pity.

Modern microbolometers come in many case formats ranging from traditional and expensive hermetically sealed metal modules, like the ULIS 02-05-11, to ceramic/metal hybrids and non-metal silicon layered designs. I wonder if the longevity of the vacuum has been improved ? The removal of the PeltierTEC on some has lowered the vacuum contamination risk.

FLIR offer a 10 year warranty on their Ex series silicon layered microbolometer. That indicates decent confidence in the modules vacuum integrity. Time will tell.

It would be interesting to see a 'before and after' experiment where a microbolometer vacuum is deliberately compromised. Sadly destroying a good microbolometer is the stuff of madness to most individuals, such as myself. I would have expected NASA to have documented microbolometer failure modes as they have done with Stirling Coolers. I have not seen such yet though. I shall do some searching.

Bill, thanks for the comment. Most interesting. I hate to overstate an issue without having facts. You understand the issues far better than me.

Fraser

[amyk]

It doesn't seem all that different to driving a CCD.

[Fraser]

The significant difference is in the required bias of active and blind pixels. The read out of pixels in FPA's is pretty standard stuff.

The uncooled Microbolometer is a noisy beast and the clever stuff happens in the processing of the ROIC output. Noise reduction algorithms are key to a decent image.

Fraser

[Flanbix]

Interesting read.
That is nice to see what the "older" models looks like too.

[Fraser]

@Flanbix,

I hoped discussing the circa 2006 ULIS technology would be interesting as it is very much part of the development cycle of the present microbpolometers. It may also help to explain why these sensors were so expensive. They are a combination of electronics and mechanical engineering, a little bit like a thermionic valve ! They even have an active Getter !

The early designs rely heavily on basic engineering to maintain a hermetically sealed environment for the very sensitive VOx and A-Si pixels. Such mechanical engineering is both complex and expensive to manufacture. The fabricator makes the sensor Die + peltier TEC and then has the challenge of placing such in a hermetically sealed container that needs leads,a Germanium window plus a good quality vacuum. Each requiring its own manufacturing stage to accomplish. Then there is the Getter activation ! A bit different to making your common integrated circuit I think you will agree.

Manufacturers have understandably been working on simplifying the manufacturing process and lowering the unit cost  as a result. The first step was to revisit the vacuum containment with a view to complexity. A hybrid design was created that used a metal face plate (containing the Germanium window) attached to a common ceramic IC casing. Both leaded and leadless types were produced. The metal to window and ceramic substrate bond was still a complex task however and the vacuum volume was still relatively large and vulnerable within the package. Development has continued at a pace and the metal faceplate was removed from the design, followed by the clever MEMS manufacturing processes that permitted a microbolometer to be built solely through IC manufacturing processes using silicon as the structural material.The vacuum volume within the component is tiny when compared to the ULIS sensor detailed in this tread. It is also a robust encapsulation on which the OEM is happy to offer a long warranty life.

Modern microbolometers are a marvel of engineering, but for me, relatively boring chunks of silicon I like the mechanical nature of the older encapsulations. What is very true is that the modern microbolometer is very much more refined and can produce excellent lower noise image data. They are a very well developed product that thankfully has become viable for inclusion in lower priced cameras.

Whether the microbolometer is circa 2006 or 2017, they are still very clever bits of technology that perform their intended task very well. The basic principles of operation remain unchanged. They have just got smaller and of 'smarter' construction.

Fraser

 

[Bill W]

A large vacuum volume is good, you can outgas a bit and not cause a large lift in pressure.  Small volumes with the same amount of trapped gas are the problem.  As you say losing the peltiers or heater elements has helped a lot in reducing outgas but it is still a big issue. 

One surprise to me is that back in 2002 Raytheon were doing wafer level packaging on the AS2000 core (a 30um 160x120) yet so few others have managed to achieve this in 2017.

Bill

[sam1275]

Interesting article.
So all microbolometers like VoX and A-si are vacuum sealed? From the oldest to the newest? What about BST sensors?

[Bill W]

Yes, they are all vacuum devices including BST & PZT. 

Thermal conduction / convection from the pixel cripples the ability to raise the pixel temperature and so create the effects (resistance change, charge production etc) needed to sense.

Bill

[sam1275]

Yes, they are all vacuum devices including BST & PZT. 

Thermal conduction / convection from the pixel cripples the ability to raise the pixel temperature and so create the effects (resistance change, charge production etc) needed to sense.

Bill

Thank you very much!

[Bill W]

Another reference paper (public domain) from DRS gives the pressure required over life in a sensor as 10mTorr. 
Atmospheric = 760 Torr. (mm of Hg)

Bill