Fresh graduate vs newbie thermal engineer
Mindset transition from school project to thermal engineer work.
After the general introduction series of industry (What (Thermal) is in High-Speed IO?), I would like to share tips for mental model change from my contract engineer role to full-time thermal engineer. If you are exploring your first job at sea, I hope this article helps you find directions for a new onboard fresh mindset.
It is a grateful opportunity for me to have a senior engineer supervising my work from the first day. Although us was a small team making a big impact in following, there still were things too complex to let a junior digest in a short time. What I realized most as a newbie is becoming a sponge soaking up knowledge and following the seniors’ “best recipe” at least once. There usually will be surprising and meaningful rewards at the end. Following others’ styles will indeed be uncomfortable, but it’s also the fastest way to mindset growth.
5 tips for becoming from fresh blood to newbie
1. Understand tools in hand and maximize advantages
2. Simulation model as simple as possible but no simpler
3. Mesh control by levels but no total numbers
4. Experiment calibration from gageR&R to accuracy
5. Match validation purpose with feasible setup/assumption
- Understand the tools in hand and maximize advantages
There are so many tools in the market, and some reasons make Flotherm outstanding in electronic cooling. No matter how the tool takes approximations and makes computational cost-saving, the one mindset change of “Cartesian grid and Immersed boundary” usually is the first disaster for the users who are used to “Points/Curves/Surfaces constructions” classic tools.
“No one approximates geometry in such a shortcut !” _by a user frustrated with new tool flotherm :(
Instead of living in our comfort zone, stepping out and growing up skills are the professions we can carry in our careers. Who knows which tool will be the next? :)
2. Simulation model as simple as possible but no simpler
A simulation is a powerful tool helping engineers speed up design iterations virtually and converge concepts. So keeping the simulation model repressive of real products is important but not a mockup on paper. There are three major considerations for thermal design which should be taken care of in simulation details.
Below shows the difference in flow impedance approximation at the system and component levels. Due to flow distribution sensitivity, an oversimplified flow will result in a misleading flow pattern which impacts the heatsink evaluation and failure risk.
In stacked high-speed IO, the inner structure is more complex but also critical in thermal design. Although each engineer has a different standard of processing, below shows an example of exact vents for distribution while partially simplified connectors for approximated flow patterns and computational cost saving.
one of the best products in the market
QSFP DD | High Speed Interconnects | Amphenol (amphenol-cs.com)
After all, there are ways of simplification, but a thermal engineer is responsible for the quality of output not only the impedance but the distribution matters.
3. Mesh control by levels but no total numbers
The first simulation challenge when I joined in electronic cooling was the assembly mesh control. Unlike the school projects having single pure geometry, there are components distributed in rack chassis, and solving the de-key pointed issue is critical but complex. To have a reasonable mesh quality while no project leads time running mesh independence, an example of mesh control is shared for reference.
The above shows the “levels” concept in component/assembly controlled by path importance order. Unlike other tools that have geometry levels, the minimum cell size method in flotherm is more flexible while localized constraint will increase the computational complexity and crash risks.
“In the business world, the rearview mirror is always clearer than the windshield.” _by Warren Buffett
Copying a converged setup is a good start but may not work in new cases. A robust mesh mindset is the only way to keep the simulation output consistent regardless of assembly complexity.
4. Experiment calibration from gageR&R to accuracy
The thermal experiment is important to validate the thermal designs, but there is not much standard to conduct experiments in a “good” way aligned with the end user application. Below shows a relative comparison of details in experiment setups for reference.
There are 2 common ways to validate components before testing in the system, although there are differences from system rack to component validation.
1. Fixed flow: reproduce typical flow rate at 3/4/5 CFM
it’s more convenient to validate heatsink quality by fixed flow rate because of the same air cooling capacity. The engineer can focus more on the quality of the designs.2. Fixed pressure source: reproduce operating impedance at -1.0/-1.5/-2.0 inAq
Includes flow rate change from impedance which is closer to the actual response, but these two variable makes validation harder for cross-comparison of designs., but good for system operating point check.
The heatsinks between the same impedance and the different bezel are hard to compare, but the results with the same flow rate can be compared no matter how the impedance change.
5. Match validation purpose with feasible setup/assumption
Customer-oriented and referencable information from their perspective is always important in business success. Reproducing end-user applications in experiment setup is becoming more and more important, although there are limitations.
Below are 4 keys to making a deeper customer engagement in the wind tunnel
1. Fixture design: reproducing sliced IO space in a wind tunnel
2. Front bezel: reproducing the customer’s system airflow
3. WindTunnel inlet: reproducing gap between port-to-port / airflow direction
4. Heater profile: reproducing transceiver power distribution and monitors
Although customers usually share little information about their system, it’s always a plus to provide total solutions even a referenceable estimation. The cross-functional/department teamwork will not only broaden your horizon but also be able to make reasonable assumptions and bring validation results closer to realistic application results. For example, some products belong to Copper BU while transceivers belong to Optical BU.
A good talent would combine knowledge in both domains and bring customer expected architecture evaluation instead of providing separate information to the customer and then pass to ODM evaluation.
After all, above are my 3simulation 2validation mindest sharing. Although there are different cultures in different teams, all would have the same business success target. I hope my sharing helps you find your navigator at sea. Cheers~
Next, I would like to share my second-year growth in
1. Thermal expereince in debuging
2. Contact resistance correaltion
3. Project tracking via Jira
4. Clear VOC and customer support
5. Manufacturing line quality control & limit sample
