Comment on Why this $10 spectrometer chip could bring real-time chemical sensing to wearables
Sal@mander.xyz 5 days ago
Very cool!
I also like that they included a device cost analysis. I think they used some aggressively optimistic assumptions (100% yield from the wafer) and large-volume procurement prices that I am skeptical of, but if true it is valuable info.
Now I’m wondering, does the designer bear the cost of wafer yield, or the manufacturer?
Like do I pay for X amount of wafers and pray it’s got a good yield, or do I pay for Y amount of working chips and the manufacturer prays instead?
I feel like the manufacturer has the most impact on yield, but probably the contract provides an estimate for a specific process
I worked in contract manufacturing for making electronics.
We had certain contractual yields required. If the yield was below 90% (industry standard), they would pay us for the scrap. The assumption for that being they should have a confidence and reliability in their process that allows it to be higher. If we produced a higher yield than 90%, they got those for free.
Of course, yield numbers and reimbursement terms can be negotiated as part of the contracts.
Sal@mander.xyz 5 days ago I am not 100% sure, but I think that the designer bears the cost of the wafer yield.
The PDK is specific to the manufacturer, and I imagine that the risk associated with specific features is known and the yield range can be estimated. A design with more and riskier the features would lower the expected yield, and it makes sense that this risk gets passed onto the designer. The contract may have some protections for the case when the yield is significantly lower than expected because of the manufacturer’s fault. Just a guess though, it would be great if someone with real experience can answer.
RobotToaster@mander.xyz 5 days ago
Even at $50 a chip I could see applications for this.
Yes, there is a lot of potential. There are micro-spectrometers available covering this range, like the MEMS-FPI type spectrometer or the multi-pixel arrays with filters (example: www.mantispectra.com/products). But this new kind of detector seems to come with higher resolution, faster readout than the MEMS-FPI, and more sensitivity. It is quite exciting, I’m curious to see if they commercialize it soon.