led, bga, and qfn assembly and inspection case studies
TRANSCRIPT
www.creativeelectron.com
LED, BGA, and QFN Assembly and Inspection
Case Studies
Copyright © 2014 Creative ElectronNOT FOR DISTRIBUTION
www.creativeelectron.com
Great Manufacturing Company – GMC
Global company with facilities in the US, Mexico, and ChinaLarge scale manufacturing of PCB assemblies – BGA and QFN boardsCustomers include medical device, military, and aerospace companiesNew management has pushed aggressive cost reductionUnfortunately financial performance is declining
www.creativeelectron.com
GMC’s financial performance
GMC was very profitable for most of its 20 years of existenceSteep decline in profits started 12 months after introduction of LED luminaire product lineAnalysis show warranty costs have skyrocketedHave cost savings cost the future of the company?
www.creativeelectron.com
Great Manufacturing Company case study
GMC was new to the LED luminaires market. They offer two warranty options to their customers:
Standard 38,000-hour warranty for free Extended 50,000-hour warranty for extra cost.
A single LED failure results in a luminaire fail. Pricing model:
Price of a single luminaire: $100Price of 50,000-hour extended warranty: $20Production and distribution costs: $35Warranty cost: $5
www.creativeelectron.com
The importance of quality in your pricing model
The pricing model is based on the expectation that within the standard warranty period the company will receive up to 5% in returns, and up to 20% in returns within the extended warranty period. GMC needed to understand if the amount allocated to warranty costs was enough. They started fabricating LED luminaires at a reasonable profit margin, but margins were destroyed by return of product within warranty
www.creativeelectron.com
LED Voiding
This relationship shows us the operating temperature of the LED grows exponentially with the void area. For this reason it is critical to keep
the voiding area to a minimum.
www.creativeelectron.com
Void Area and Operating Temperature
To determine the operating temperature of an LED mounted onto a substrate, first we need to calculate the thermal resistance between the LED and the substrate. This thermal resistance, R, is given by:
Where x is the thickness of the die attach, k is the thermal conductivity of the material, and ACONTACT is the total area of contact between the LED and the substrate. This area Ais given by:
www.creativeelectron.com
How to calculate the LED operating temperature
The temperature difference between the LED and the substrate can be calculated as:
Where Φ is the head flow measured in WattsThe LED junction temperature is:
www.creativeelectron.com
LED operating temperature
The temperature of operation of the LED is going to be higher than the substrate. This is a very important equation, because it allows us to determine the junction temperature as:
This equation also shows us the critical impact that voids have in the operating temperature of LEDs. All parameters in this equation are constants, except the void area. The heat flow is a function of the power being dissipated by the LED, which in turn is determined by the current and voltage applied onto the LED. The thickness x is fixed, as it is the total area between the LED and the substrate A. The thermal conductivity of the material k is a constantThe temperature of the substrate is also a constant.This allows us to further simplify this equation as !
www.creativeelectron.com
a) Photo of LED assembled onto substrateb) x-ray inspection image of LEDc) x-ray inspection image with identified and measured voids using
TruView 5 softwared) 3D rendering of die attach void
www.creativeelectron.com
Scanning electron microscopy of LED assembly
www.creativeelectron.com
Distribution of void area in a population of 1,000 LEDs assembled onto a thermal sink substrate
!
With the results from this x-ray inspection we created the histogram with the distribution of void area. The large number of devices inspected lead to a normal distribution with a mean void area of 49.7% and a standard deviation of 9.8.
www.creativeelectron.com
Now that we know manufacturing LED voiding, so what’s next?
We could have completed the analysis here, but we were interested in determining the real cost of voids.
Is the distribution we found appropriate or not to our customer’s pricing model? Are they charging enough for the extended warranty?
Based on the results obtained, we can determine the real cost of the die attach void. The next step in this analysis is to use the LED manufacturer’s data on the lifespan of the LED as a function of operating temperature. Although these are not deterministic numbers, they allow us to estimate how many of the sold luminaires will return from the field.
www.creativeelectron.com
Relationship between junction temperature (Tj) and LED service lifespan
www.creativeelectron.com
Relationship between void area and LED junction temperature
We measured a set of LEDs in operation and determined several points to relate void area and operating temperature. Based on this information we were able to relate the amount of voiding we measured to the expected junction temperature of the LED
www.creativeelectron.com
Case study: Distribution of LED population estimated lifespan – 1,000 samples
Standard Warranty: 38,000Extended Warranty: 50,000
133 are expected to fail within the standard warranty period,
while 540 are expected to fail within the extended
warranty
www.creativeelectron.com
The turn around towards profitability starts with a powerful quality inspection program• Due to an inadequate quality program, GMC far underestimated the
expected cost of return products• Not only the pricing model utilized by the company had a deficit within
the standard warranty; it had an even larger deficit within the extended warranty period
• GMC had to temporarily change its pricing structure to compensate for product shipped with the deficient manufacturing process
• A complete overhaul of the manufacturing process was done to bring it to tighter standards
• The void area was utilized as the key metric to measure process control quality
• GMC invested in quality control and inspection by utilizing a TruView Elite X-Ray is used daily to measure the void of sample luminaire
• Current data shows that the mean void area dropped to 35% with a standard deviation of 4.5
• Thus, they were able to reduce the pricing for their luminaries. This reduction greatly increased their competitiveness in the market.
www.creativeelectron.com
Typical LED luminaire fabricated by GMC
www.creativeelectron.com
LED samples inside x-ray system
www.creativeelectron.com
First samples after implementation of quality program and inspection
www.creativeelectron.com
Typical samples prior to quality inspection; prior to adoption of x-ray imaging
www.creativeelectron.com
Summary of results for luminaire “A”
120 LEDs on each boardSinkPAD IITM substrate
Prior to quality program
Initial samples after implementation of quality program
Voiding Luminaire “A”Mean 40%
Std Deviation 15
Voiding Luminaire “A”Mean 8%
Std Deviation 0.9
www.creativeelectron.com
How GMC improved LED assembly quality?
• Substrate surface cleanness • Substrate and LED metallization • Reflow oven temperature profile
• Paste selection, dispensing, and storage• LED and substrate handling and storage
www.creativeelectron.com
Current quality inspection program
Use x-ray inspection to set proper manufacturing parameters for each product. Use voiding area a key metric to determine process variationUse LED information to determine maximum operating temperature to achieve expected lifespanMeasure temperature as a function of voiding area to determine maximum voiding allowedSet plan to continuously measure voiding area with x-ray to verify conformity of manufacturing process
www.creativeelectron.com
Medical device probe assembly
Complex construction2 reflow cycles: first for FPGA, second for sensorLot’s of connectivity problems with the FPGA, likely due to the second reflow
Sensor
FPGA
PCB
www.creativeelectron.com
Sensor side assembly
Bare PCB PCB with sensors removed
www.creativeelectron.com
Simplified assembly – 1 sensor, 1 FPGA
Sensor
FPGA
PCB
www.creativeelectron.com
X-ray of assembly – 1 sensor, 1 FPGA
www.creativeelectron.com
X-ray of assembly – 1 sensor, 1 FPGA
www.creativeelectron.com
X-ray of assembly – 1 sensor, 1 FPGA
www.creativeelectron.com
X-ray of assembly – 1 sensor, 1 FPGA
www.creativeelectron.com
X-ray of assembly – 1 sensor, 1 FPGA
www.creativeelectron.com
BGA inspection with dual energy imaging
www.creativeelectron.com
BGA inspector results – void analysis
www.creativeelectron.com
BGA inspection – ball size analysis
Gradient in ball sizes indicate planarity issues with the assemblyDifferent ball sizes also cause voiding discrepancies in the assembly
Small
Large
www.creativeelectron.com
Planarity issues with FPGA assembly
Sensor
FPGA
PCB
www.creativeelectron.com
Fixing the problem
• Epoxy on the corners of the FPGA• Underfill also an option
Sensor
FPGA
PCB
www.creativeelectron.com
Data processing board
• High volume production -thousands of boards per week
• Intermittent issues with QFN parts
www.creativeelectron.com
QFN assembly with excess voiding
www.creativeelectron.com
QFN assembly with excess voiding
www.creativeelectron.com
Problem QFN assembly – open pin
www.creativeelectron.com
Detail of open pin in QFN package
www.creativeelectron.com
Quality control inspection with x-ray
• Problem of open pins of QFN traced to dirty stencil
• Review of stencil cleaning process was done. Proper procedures were not being followed
• Training of cleaning staff was repeated to assure they follow all steps
• Since proper stencil cleaning procedures were followed, no more issues of this nature were found.
www.creativeelectron.com
Another QFN problem: improper alignment
www.creativeelectron.com
Process design with x-ray
• Problem traced to incorrect stencil design• Not enough solder paste was being deposited
under the QFN, thus allowing it to “twist” during reflow.
• New stencil was designed to deposit proper amount of solder paste under the QFN
• Since implementation of new stencil, no more problems of this nature were documented.