Embedded or removable memory?

When a designer is developing a product that uses flash memory, they have a key decision to make at a very early stage. Should they use removable or embedded memory? The market is changing rapidly, with manufacturers continually developing new technologies to meet the demand for ever smaller devices, faster performance and increased longevity. For example, storage cells can now be stacked to create 3D NAND products, and SD cards from high street brands can hold 1TB of data (equivalent to around 25 hours of 4k video).

Memory is not all created equal, and what is perfect for an automotive application such as a navigation system may not meet the very different demands of a medical device, or an IoT application. Choosing the wrong type of memory could have serious consequences for development cost, which ultimately has to be passed on to the end user.

Of course all devices have some form of embedded storage - without this they couldn’t boot. The big question is whether to use removable storage and, if so, how. Sometimes the nature of the product will dictate the answer. Nobody needs to make changes to the flash memory unit that handles the coded signals they send from their TV handset. These are relatively simple devices, with a single job to do, and embedded flash memory is perfect for them.

However, there are many times when the decision isn’t clear-cut and the pros and cons of removable versus embedded need to be carefully considered. For example, the choice may depend on the volume of data a product generates. It could be influenced by manufacturing issues such as the desire to lock down or open up access to the product’s firmware, or how much freedom the manufacturer wants to give to the end user to update the product. Alternatively it may be driven by the market, and be heavily influenced by user expectations of cost and performance.

Typically, consumer grade memory cards, which are the least expensive, have the least protection, and so will have the shortest lifespan. If a product has a short life, or few read/write demands, then consumer grade cards might be fine. But a broken card, especially if it is embedded memory, can mean a broken product, so it is important to get the cost/longevity balance right.

Here are the key issues designers using flash memory should consider, looking from both a user and a manufacturing perspective.

Embedded provides optimised user performance

Embedded memory will in general make life much easier for the end user. There is no need for them to research the ideal flash memory card for their device. Instead, they buy a complete product, power it up and it will immediately work to its optimum capability, as the embedded storage will have been configured by the manufacturer to optimise speed and performance. A prime example of this is the iPhone. The manufacturer can also include ‘factory reset’ software into the device, allowing easy recovery to factory settings should the need arise.

However, there are two downsides of using embedded storage. First, it is impossible – or very difficult – for the end user to upgrade their device’s storage should they require more capacity. Second, the price of the device may have to be higher to cover the manufacturer’s costs for sourcing and integrating the flash memory. As already mentioned, the iPhone exemplifies this, and the higher price has been a part of Apple’s strategy. It was an issue in the early days of smartphones, and as a result several manufacturers chose to use a microSD card and let the user decide which storage to use.

Removable memory offers upgrade opportunities

In contrast to embedded storage, with removable storage updating memory is easy; the user can add more capacity and speed and decide exactly what card to buy depending on their budget, including a preferred brand if they wish. Being able to upgrade the memory enables the device to have a longer lifespan, and the user may be able to add new features via updates administered from a flash memory card. Data is also portable, so it can be taken to a new device in the future, such as moving a music library from an old handset to a new one, and it can be secured through on-card encryption.

One device which tries - and in general succeeds - to provide users with the best of both worlds is the Raspberry Pi. This was specifically designed to boot from a flash memory card, but one supplied and preconfigured by the manufacturer. The idea behind this was to allow the Pi to ship with a fully certified boot system that was identical for every unit. This boot system would provide access to a range of operating systems that could be pre-loaded for end user ease – in effect providing one of the key advantages of embedded storage i.e. optimised performance. However, this card could then easily be updated over time as the Pi developed new capabilities and operating systems were refined, thus providing the benefit of removable storage.

Embedded gives manufacturers more control

Embedded storage offers many advantages to designers and manufacturers. If a designer chooses embedded storage, they have complete control of the memory used, enabling them to select the optimum balance of capability and cost. It also gives them more control over how data is read and written to the storage - they can design their own interface and greatly improve performance. Memory performance primarily depends on the quality of the controller and firmware, which influence read and write speed, durability and longevity. Because they know the exact specification of their flash memory, designers using embedded memory can be confident in making claims around speed and performance. And it is easier for the company to protect its intellectual property if it uses embedded storage as it can hide its code effectively.

Embedded memory also tends to be smaller than removable, which means more storage can be provided where space is at a premium, or that there is more room for other components. This can make a significant difference to the final product.

Another advantage of embedded storage is that it can ensure repeatable performance. In addition, by treating their flash memory as a critical component, designers can take control of the supply chain to ensure that what they obtain is the same every time, at the agreed cost, and guaranteed in a fixed Bill of Materials (BOM) - (Removable memory cards can also come with a fixed BOM.) A good memory supplier will also support Product Change Notices (PCNs) and End of Life (EOL) notifications.

Above: Programmed and labelled memory cards prior to packing

Sometimes a product may have been designed with removable memory but it may subsequently be advantageous to move to embedded, even if there is a redesign cost involved. One organisation had developed a successful integrated back-up and charging device for smartphones based on a removable microSD card. However, it needed to lower the BOM cost in order to reduce the sale price and therefore attract more customers. We advised them to move to embedded memory, which would have reduced both cost and product size.

In this instance, the organisation decided to continue with removable storage. Unfortunately, based on their sales forecast they made a non-cancellable commitment for a large volume of microSD cards with specific capacities. Consumers bought a different mix of product, which meant the stock they had secured was not what was needed. The pricing also changed (flash pricing is highly volatile), so what had initially been a good purchase was suddenly over-priced. This type of issue can be mitigated to some extent by having a good supply chain and a partner who can assist with flexibility.

Embedded may not function in all cases

Despite its many advantages, embedded flash memory is not right for every application. The major downside is that if embedded memory fails, the entire device may be rendered unusable. This may require a return for repair, either under warranty or at a cost, creating pain points for both manufacturer and user. Integrating embedded storage into devices also adds significantly to development time and cost, especially if the designer has to create a bespoke interface.

If an organisation is considering changing from removable to embedded storage, it is vital to ensure that embedded storage will perform equally in the same environment. This is not always straightforward. One organisation replaced the removable memory in an infotainment system with embedded to reduce costs. They assumed that embedded memory would perform as well as removable. However, the form factor they chose was very new at that point and they found that they could not programme the chip with data offline and then have it assembled onto a PCBA - as something happened that corrupted the data during the assembly process. This highlights the need for thorough testing before making a change.

Removable memory can open up the market

One of the key advantages of removable memory is that the data is stored safely when the device itself is powered off. It is also easier to upgrade the product, as new, faster and better memory is continually being developed.

Perhaps the greatest advantage is that third party applications can be booted from a flash memory card, opening up new horizons for device users and making it viable to sell the device to a wider market. Retailers may also be more interested in stocking a product with removable memory as they can upsell memory cards as accessories.

However, the manufacturer cannot control the flash memory used, even if they initially supply it as part of the product. There are many specifications and grades of flash memory available, and the end user might find it difficult to understand the differences or work out what is ideal, particularly if their budget is limited. If the user chooses a card that is under-specified for the device, or replaces the original one with a card that offers more memory but not necessarily the same performance, it might cause slowness or malfunction, for which the device and hence the manufacturer are blamed.

Ensuring data security on removable storage

While the above are the core principles that apply, there are always exceptions to the rule. By obtaining advice from a specialist with an in-depth understanding of the performance of different types of memory, designers can push the boundaries and achieve a balance of cost and performance tailored to the needs of their specific product.

For example, Cardwave has been supporting Avaya since 2012, when they launched V2 of their IP Office system. This replaced a hard disk, embedded memory and a smart chip with an SD card to provide both storage and security. They wanted to save space and reduce cost while providing improved functionality and scalability. Cardwave worked closely with Avaya over the 15 month development period to qualify an SD card that met performance, endurance, reliability and data security requirements.

This was not a straightforward process of simply testing different cards to find one that could deliver the required performance consistently within budget. Security of the data on the SD card was vital - Avaya’s software licensing revenues were worth millions of dollars and there was a real risk that customers would make unlicensed copies denying Avaya the revenue. Cardwave developed a tailormade and complex personalisation process to ensure that each card was totally unique and contained qualified credentials that had not been used before e.g. serial number, feature keys and other proprietary data.

As this example shows, in some situations the decision on whether to choose embedded or flash memory can be finely nuanced. Engineers need to understand all the relevant factors, from anticipated product lifespan to the need to protect IP when evaluating their decision. This will also help them make the case to their purchasing team for their chosen memory product, and ensure that the resulting product meets all the required performance criteria.