3 ms·

About 7 or 8 years ago I built an internal testing system used to test a satellite payload which used a microcontroller board from a reputable manufacturer, pur

About 7 or 8 years ago I built an internal testing system used to test a satellite payload which used a microcontroller board from a reputable manufacturer, purchased through a reputable electronics supplier. When we updated the ftdi driver, the board was bricked. Fortunately the system was still in development so we found a different board - it was only a bit of pain. However, if that system had been shipped (as it was 6 months later), that board being bricked could have had much more significant ramifications. It would have caused a slipped schedule and tangible costs. What should I have done differently?

robomartin 6y ago
You would have to explain the degree to which parts were traced, certified, tested, etc. If you simply trust the supplier and distributor anything could happen. I mentioned in another thread that we had one of the top two US electronics distributors knowingly ship us low "B" grade components many, many years ago. These components were in allocation and an enterprising young man at the distributor thought he would be smart and ship us a lower qualified component instead of what we ordered. That was twenty years ago or so. It cost our company dearly, nearly took us out of business. This was that learning moment for me and it changed my approach to sourcing as well as the level of trust I grant anyone providing us with components. I will never put anything into a design where an illegitimate or lower grade component could jeopardize the safety, reliability or operation of the system. At some level this is what engineering is about, isn't it? It's that old "Trust but verify?" concept. We are working on industrial, flight (aircraft) and space (lunar) projects at the moment. No component will go into any of these systems without full knowledge and verification of its origins. This is true for individual components or contracted sub-assemblies. BTW, this issue of failures being caused by not verifying components isn't anything new. The history of engineering is full of examples. One reasonably recent example of this happened to SpaceX a number of years ago: https://www.space.com/29994-spacex-rocket-explosion-cause-faulty-strut.html https://www.space.com/29994-spacex-rocket-explosion-cause-fa...
- wegs 6y ago
  No. This is not what engineering is about. Engineering is about making reasonable design trade-offs. Simply throwing dollars at a wall without doing an ROI estimate is called overengineering, and overengineering is bad engineering. For reliability, where that trade-off sits depends on the application. Aerospace, medical, consumer electronics, and disposable toys will sit in different places. If I lose a mission to Mars saving $100 on part which had a 5% chance of failure within a year, that's very poor engineering. If I include that same part in a $3 toy, bringing the price to $103, that's equally poor engineering. Whether I trust or trust-and-verify depends on how much the "verify" part costs, how strong my trust is, and what the costs of failure are. Normally, the ROI calculation is easy; capitalist markets work well for this. I can ballpark expected costs. When working with a customer like the government, the boundaries might be a little bit distorted, since the customer is process-oriented. The government might have a hard salary cap which makes it impossible to bring in qualified engineers, and I might take 3 years with a team of 5 people at $100k to do what one person at $300k could do in 6 months. At the same time, I might have hard requirements on process, such as origin-tracing every part. The danger is when that becomes in-cultured and spills over to other places. If I'm working for the government, I'll follow government processes, and I understand why those are there. But I won't confuse those processes with good engineering. Once people do, they become poor engineers. If I've shipped a toy which unwittingly has thousands of fake parts which I thought you made, we'll both have been cheated, and I'll expect you to solve that with me cooperatively. If you hack into my product and brick it, even if you were legally in the right (and you're not), you've lost a customer. That's bad business too.

robomartin 6y ago

You would have to explain the degree to which parts were traced, certified, tested, etc. If you simply trust the supplier and distributor anything could happen. I mentioned in another thread that we had one of the top two US electronics distributors knowingly ship us low "B" grade components many, many years ago. These components were in allocation and an enterprising young man at the distributor thought he would be smart and ship us a lower qualified component instead of what we ordered. That was twenty years ago or so. It cost our company dearly, nearly took us out of business. This was that learning moment for me and it changed my approach to sourcing as well as the level of trust I grant anyone providing us with components. I will never put anything into a design where an illegitimate or lower grade component could jeopardize the safety, reliability or operation of the system. At some level this is what engineering is about, isn't it? It's that old "Trust but verify?" concept. We are working on industrial, flight (aircraft) and space (lunar) projects at the moment. No component will go into any of these systems without full knowledge and verification of its origins. This is true for individual components or contracted sub-assemblies. BTW, this issue of failures being caused by not verifying components isn't anything new. The history of engineering is full of examples. One reasonably recent example of this happened to SpaceX a number of years ago: https://www.space.com/29994-spacex-rocket-explosion-cause-faulty-strut.html https://www.space.com/29994-spacex-rocket-explosion-cause-fa...

wegs 6y ago
No. This is not what engineering is about. Engineering is about making reasonable design trade-offs. Simply throwing dollars at a wall without doing an ROI estimate is called overengineering, and overengineering is bad engineering. For reliability, where that trade-off sits depends on the application. Aerospace, medical, consumer electronics, and disposable toys will sit in different places. If I lose a mission to Mars saving $100 on part which had a 5% chance of failure within a year, that's very poor engineering. If I include that same part in a $3 toy, bringing the price to $103, that's equally poor engineering. Whether I trust or trust-and-verify depends on how much the "verify" part costs, how strong my trust is, and what the costs of failure are. Normally, the ROI calculation is easy; capitalist markets work well for this. I can ballpark expected costs. When working with a customer like the government, the boundaries might be a little bit distorted, since the customer is process-oriented. The government might have a hard salary cap which makes it impossible to bring in qualified engineers, and I might take 3 years with a team of 5 people at $100k to do what one person at $300k could do in 6 months. At the same time, I might have hard requirements on process, such as origin-tracing every part. The danger is when that becomes in-cultured and spills over to other places. If I'm working for the government, I'll follow government processes, and I understand why those are there. But I won't confuse those processes with good engineering. Once people do, they become poor engineers. If I've shipped a toy which unwittingly has thousands of fake parts which I thought you made, we'll both have been cheated, and I'll expect you to solve that with me cooperatively. If you hack into my product and brick it, even if you were legally in the right (and you're not), you've lost a customer. That's bad business too.