How we can all learn from Tesla’s experience of the small but mighty eMMC

(Image credit: Image source: Shutterstock/violetkaipa)

Many of us think of flash memory as the removable storage in our phones, our digital cameras, or other personal devices. But it is used far more widely, including in industrial settings and, increasingly in cars where it has a key role to play in navigation systems and the visual display systems drivers have come to rely on.

Each individual use of flash memory requires a form factor to be specified for the job. As well as coming in different physical sizes, with different storage capacities, there are different use grades of flash memory too, including variations of read/write speeds and levels of robustness for more demanding conditions such as extremes of heat or cold, or industrial implementations.

Get the specification right and a product will hum along, with end-users taking advantage of what the flash memory does for them, but not aware of its central role, and certainly unconcerned about the specification of the memory itself. Get the specification wrong and none of the above applies, and the end result could be expensive and reputationally damaging. Tesla recently had an issue with eMMC flash memory leading to a mass recall of cars. What went wrong, and what can we learn from Tesla’s experience, beyond the fact that eMMC may be small, but it is mighty enough to bring a whole vehicle to a halt?

Tesla’s mass vehicle recall – what happened and why?

Tesla’s recent recall of over 130,000 Model S and Model X cars happened because the driver’s screen-based information center, the Media Control Unit (MCU), stopped working. In a letter to Tesla the US National Highway Traffic Safety Administration noted that the loss of the MCU meant drivers did not have access to safety features including the rearview camera and defogging/defrosting controls, and there were also adverse impacts on other safety features such as the autopilot advanced driver assistance system.

In information relating to the recall Tesla said the culprit was wear and tear on 8GB eMMC memory. In essence, it seems the eMMC was written to every time a car was started, and over time its storage capacity was reached. With no way to erase data the eMMC NAND cell hardware failed, causing the MCU to fail in turn. The 8GB eMMC was being asked to do more than it was specified to do, and the recall allows 64GB eMMC units to replace those 8GB units. If 64GB units had been used from the start, the recall may well have been avoided.

Lessons for every specifier of flash memory

Tesla’s experience is a reminder for us all that even the smallest component needs to be carefully considered, and there are lessons for every company that specifies flash memory for its products – not just eMMC, not just for makes of highly expensive products, and not even just for car makers.

Sweat the small stuff.  We can’t know how or why the eMMC used was specified in Tesla’s case. Sometimes parts are specified with a particular workload in mind and tolerance is absolutely fine, then more workload is placed on them as a device develops and tolerance can struggle. The part can be upgraded – but only if it appears in the right schedule, in front of the right eyes, at the right times.  Tiny eMMC might be easily overlooked. Don’t let that happen. Memory cannot be an afterthought in the design process. It may be small and hidden away, but it is a core component and needs to be treated as such.

Don’t cut corners. Sometimes lower cost or lower quality parts are specified to cut production costs, and every single fraction of a penny in the bill of materials really does count. But the flash memory components in a device are there for vital reasons. Nobody would suggest deliberately over-specifying the memory – remember that bill of materials has to be kept under control – but there is a fine line between over-specifying for the sake of it and building in capacity for future upgrades. Erring on the generous side rather than scrimping and just getting by can be a very good strategy.

Be aware of different grades of memory. Cars are complex products, increasingly reliant on electronics components for core functionality, and they’re not cheap to build or to buy. While private owners might think of their car as a consumer good, it is a highly specialised, precisely engineered device, and building it requires a level of attention to detail in component selection that approaches or equals that required for industrial processes or manufacturing rather than the production of consumer goods. That is true as much for the smallest, most hidden of components like flash memory as it is for the parts users see and interact with. Flash memory comes in many grades of quality for different use cases. Car manufacturers really need to use automotive and industrial grade flash as these products have health monitoring features that will enable the manufacturer to predict failure and plan maintenance.

Work with trusted third parties. Not every maker of complex devices has detailed knowledge about every single component. There will be specialists in many fields that can advise on getting the specification right as well as help with sourcing from the most appropriate suppliers. In our case at Cardwave we cover those two areas as well as provisioning flash memory, and our manufacturing partners can support us on specialist testing where that is required. We have a strong track record in the automotive sector with many household name customers.

In the end no company wants the cost, hassle or reputational problems that are caused by having to recall a product. Flash memory like eMMC is a component that is small in both size and cost relative to the overall product, but it is vital to functionality. Ignore it at your peril, for it is truly small but mighty.

Paul Norbury, CEO, Cardwave

Topics