|
How could we intentionally cause a Pyrex dome to shatter? Glass fracture can be predicted and controlled. Glass can be heated and cooled repeatedly without causing fractures. I spent a substantial portion of my working lifetime around breaking glass, focused on transient and steady state heat transfer.
Although not a materials engineer, I designed circuits using alumina for 25 years. Alumina is ceramic. All ceramic is glass, but not all glass is ceramic. The alumina substrates were fabricated using a silk screen like printing process using metal screens and a squeegee to spread the metal paste, or glass dielectric used between metal layers. The process was print, dry, fire, sometimes with several print and dry cycles before a firing operation. Firing occurred multiple times above 800 degrees F. The substrates traveled through the oven on a metal conveyor belt. The temperature of the material would smoothly ramp up and back down. At the end of the belt, they were placed in metal cartridges (with separation between) by machinery.
The ceramic sheets were approximately 3 x 5 or 6 x 8 inches in later years. There could be as many as 21 print operations for four metal layers with the required dielectric insulators in between metal layers. Some of those print operations created resistors that could be laser trimmed to within 1%. Many times, multiple identical circuits were assembled on a single sheet of ceramic. They were separated after the components were soldered to the substrate. We intentionally broke the glass sheets on a regular basis to separate the individual circuits. It was part of the production process. One circuit had more than 500 components on 7 square inches of alumina. One had a product run of more than 14 million units. When I picked up my Einsteins, the transmission of the car that I drove was controlled by that circuit.
We broke a lot of glass in a controlled manner during the production process. The circuits were separated by first laser scribing the substrate between the circuits, from edge to edge. Laser scribing occurred before the components were attached. The ceramic was cleanly fractured along the scribe lines after component placement by automated mechanical stress. As a demonstration, a pair of needle nose pliers could be placed on each side of a scribe line and with simple hand motion the substrate could almost always be cleanly broken along the scribe line. It snapped. Often, multiple breaks would occur along the various scribe lines when the shock force from the pliers traveled through the ceramic. I have seen eight or more individual singulated circuits land on the table at times. With the automated singulation process, there was almost no uncontrolled breakage. Glass breaks and breakage can be predicted and reliably controlled. Heat can safely be applied without fracture when done properly.
The circuit components were surface mounted. Most of the resistors were printed on the substrate; a few were placed, based upon economics. (Surface mounted resistors used in the Einsteins are fabricated by printing the resistor paste on a ceramic [glass] substrate, laser trimming and separating the individual resistors. It is the same process involving print, dry, fire, laser trimming and controlled breakage of glass.) Solder paste was applied by stencil and squeegee before component placement. Following placement, the populated circuits went through an infrared reflow oven with a peak temperature about 260 degrees C. In the product, the circuits were attached to an aluminum backplate. Backplate temperature was limited to 125 degrees C in operation. Individual heat producing components were hotter. During avalanche breakdown, the junctions of FETs or IGBTs could TRANSIENTLY (less than a millisecond) reach 300 to 400 degrees C many times a minute, without damage. One of the FET manufacturers put a dedicated applications engineer at our facility to understand how their FETs were surviving in our applications. They sold us more than 250 million FET die (transistors without a package) for just one application. In my applications transistor die were soldered to the conductor on the substrate. The silicon, solder, substrate, adhesive and backplate stack-up acted as a thermal spreader and heat sink for the junction. Most of the integrated circuits and transistors were mounted in die form, without packages. Interconnection was done with wire bonding. Many of the ICs were mounted upside down with solder balls attached to them. The solder balls reflowed in the oven, attaching the IC to the substrate.
I don't believe that I ever saw an individual ceramic substrate fracture due to a thermal event after it had been attached to the metal backplate. I did see individual circuits crack cleanly along what must have been stress lines during or after a prototype rework process involving application of heat to a portion of the unattached substrate. Removing and replacing a component could generally be done successively, but the process was not validated for production. Solder terminations would crack with repeated thermal cycling between -50 and 150 degrees C. That occurred during extended product validation. Test to failure reveals how much design margin you have. In Minnesota, circuits are cycled between -40 and 125 degrees C by customers and nature. They just don't realize it and the number of cycles is limited. The cold excursions are where the greater stress occurs as stress is zero when the solder is liquidus during reflow. Eventually every circuit fails with sufficient thermal cycling due to different coefficients of thermal expansion. Some materials contract more than others as they are cooled. Think bi-metal thermometer. As an electrical engineer, I was concerned with thermal management and material science. High reliability demanded it.
The subject of this thread is sudden breakage of a Pyrex dome without the application of heat from the modeling lamp. How could we likely create such a failure? The dome is under controlled physical stress due to the mount. That is likely minimal. All of the domes have minor fault lines along which a crack would propagate if started. When the flash is fired, a transient thermal shock is applied to the dome. Heat transfer is by radiation; we are dealing with light. Emissivity is probably the critical element in the radiation heat transfer equation that allows heat transfer to be controlled in this example. The initial temperature of the dome could also be a major factor if it were chilled. Emissivity has a value of 1 for a black body. Emissivity of polished silver is .02. For smooth uncoated glass, emissivity is .95. Heat transfer is a linear function of emissivity. Radiation applied to the dome is transmitted through the dome, absorbed by the dome and reflected by the dome. Most is transmitted through, or we might as well leave the shipping cap in place. Watt-seconds of energy is generated and absorbed for a very small portion of a second. It is a thermal shock.
If we could cause emissivity of the dome to abruptly change at some point on the surface, we could cause the surface temperature to suddenly change at that point when the thermal shock is applied. But, emissivity is already high for glass. If we could cause more of the light to be absorbed by the dome in a given area, we could impact differential heating to a more substantial degree.
The likely solution to our controlled experiment would be soot. If we applied a thin dime sized layer of soot at some position on the underside of the dome, I would expect that we could cause dome fractures. Emissivity of soot is nearly 1. Additionally, soot will absorb the light, converting all portions of the spectrum to heat and transfer it to the dome by conduction, rather than allowing the light to be transmitted by the glass. I would bet money on the success of this experiment. In fact, I am betting money on it. I have decided not to try it so that I will not have to purchase a replacement dome.
Look at the photograph of the broken dome. Did you notice the discoloration or stain on a portion of the dome? That stain would have caused light to be converted to heat and it would have altered emissivity. It would not be as effective as soot, but the oil on our fingers is sufficient to cause a quartz flash tube to fail. I believe this is a similar root cause.
If we were to scribe the dome to create substantial fault lines and applied soot appropriately, we could cause the dome to break in a controlled manner, producing a more similar result in repeated experiments.
What I have learned from seeing this failure is that I need to keep my Pyrex dome clean. I do not believe that the dome is a hazard to our subjects. Gravity will cause the glass to fall with a very low horizontal velocity. Glass will not fly across the room; there is too much mass. Since gravity is predictable, placing the unit directly above a subject without a shield should be avoided. I am unconcerned by the failure of that dome. Glass breakage can often be predicted and controlled. It can also generally be prevented.
|
Board index
» 
Top
