New Part Day: SD NAND Are Surface Mount Chips That Work Like An SD Card

SD cards have long been a favorite with microcontroller hobbyists. Cheap, readily available, and easily interfaced, they remain a staple for small projects that need to store a lot of data. Now, they’re available in chip form! These are known as SD NAND parts that emulate the SD card interface itself.

[LadyAda] recently gave them a test-drive after spotting a tweet from [Greg Davill] (who we’re familiar with thanks to his excellent LED cubes). The devices are manufactured by XTX Technology and available from LCSC in a range of 1, 2, 4, and 8 GB sizes. [Ivan Grokhotkov] also illuminated a similar device from another maker in a reply to [Greg’s] original tweet, so there may be more sources out there.

These chips come in standard LGA8 surface mount package and can be easily soldered to a board, offering mechanical and manufacturing benefits versus using a normal SD or microSD card in a slot-type connector. Also, unlike other SMD flash memory parts, they handle all the file system details and wear levelling for you! With the inflation of SD card sizes, it’s also difficult to find these on the shelf in normal cards these days.

[Adafruit] plan to have a breakout for these parts out soon with a level shifter included for ease of use. We can imagine these chips finding their way into all manner of datalogger projects, since they can be ordered with other parts and permanently soldered into a design. If you’ve got a particularly good idea where these chips would prove useful, sound off in the comments. Video after the break.

66 thoughts on “New Part Day: SD NAND Are Surface Mount Chips That Work Like An SD Card

        1. In fact, both SD cards and eMMC chips can work in even 1-bit mode.
          Of course you still need the other “control” wires and vcc & gnd, but I’ve tested it myself with a OG Raspberry-pi with the 26 pin header having all control pins broken out for the second SDIO “port” but only 1 usable “bit” of the data lane hooked up to the second SD card.
          And also recovered files of the eMMC of a first gen Sony Xperia Compact with some fiddly soldering to the bare essential points so I could do a dump due to dead SoC.

          Fancy shit.

        1. Uhh… exactly *why* would you think that? SD NANDs have ludicrously higher densities than SPI flash, because they’re NAND instead of NOR. But most microcontrollers don’t have a hard SD interface, while they do have hard SPI.

      1. Technically so does SD cards of any variety.
        The key being how crap the controller on the card is (in terms of prolonging lifespan of the raw flash memory) and the actual quality (and quantity in the name of redundancy) of the raw flash memory.
        Or long story short: except for some shyster big brands, you get what you pay for.

        Only place i can see this being preferred is applications requiring conformal coating or is exposed to extreme vibrations that could break a micro-sd socket.

    1. But sometimes that extra functionality causes extra “customer service” for the warranty breakers who have no skills. In these cases, it can be cheaper to pay extra for “permanence.”

      1. Plus you get a known performance profile. I’ve ran into many name brand micro sd cards that just flat out don’t work with a product (dash cam) due to shenanigans happening behind the scenes on the internal controller. This absolutely impacts your customer service cost

      1. You know, the irony of “planned obsolescence” is that when that really became a pop-culture buzzword, it was directed toward the cars of the 1950s/1960s…which were pretty much built like tanks and many of which are still around.

        But year, don’t put this in any gadget you need to work 20 years from now–then again that is not a concern for many devices.

        1. “…it was directed toward the cars of the 1950s/1960s…which were pretty much built like tanks and many of which are still around.”

          Oh my sweet summer child. I grew up near a GM plant – those cars were absolutely built to a price and were meant to be replaced every three or four years either by incompetence or conspiracy (your choice). 100,000 miles was a lofty goal and often not attained before the rest of the structure disintegrated. The engine designs were antiquated and their efficiency was laughable at the very least, as were the corporations that produced them.

          By contrast my current truck has 103,000 miles on it, will go many more and has run flawlessly from day one.

        2. None of my 1960’s cars made it to 100,000 miles. Yes, they were built like tanks: unnecessary, excessive metal in lots of places and intended to go a short distance before heavy maintenance and repair required. Modern cars still look good and work well after 250,000 miles: very few of them are completely rusted out and most of them have their original engines.
          My 1952 Jeep broke the starter retract spring about every 20 times I started it. It needed an engine rebuild at 43,000 miles. I think I spent about 1 hour working on it for every 20 hours of driving it. I can’t remember how many times it stranded me when a u-joint failed, the clutch pivot broke, the shift lever snapped, the frame rusted through, the suspension pivots rusted out… The heater had one speed and only blew on my feet and then the cable housing for the hot/cold broke off the back of the dashboard so it no longer worked, and every time I pushed the throttle down the windshield wiper (singular) stopped working and I had to turn it by hand via the lever provided for exactly that situation.
          Cars back then were *terrible*. Cool, fun, that’s why I still drive one, but modern cars beat them in every single quality metric save appearance.

          1. And if you need to service one of these 2020 cars you need to break out the ODB2 cable and the vendors software you needed to download from some shady Russian hacker forum. ¬.¬

      2. I remember the first real computer upgrade I did as a kid was installing a 2gb harddrive circa 1997. Now that capability is in an SMD chip for $2.52.

        The computer that harddrive went into was out of use by 2001 and lingered in my parents’ garage around for a few more years before being scrapped for misc parts–and that was probably a long life for a machine from 1994.

    2. Where can you get a reliable (i.e. actually advertized size) 16Gb SD card for say “around $3”?
      Actually, I’m a bit more interested in say 1/16th of that size for say 1/16th of the costs. I’ll still buy them if it doesn’t scale completely and costs say 1/8th of the price of 16G.

      For embedded systems like dataloggers, 1G is already VERY big, but $10 for the smallest I can get is too expensive….

      1. I can’t find much fault with the 16GB one can get in the Dollarama here for $4 Canadian apart from maybe it’s class 8ish rather than class 10, but you get that kinda spec overinflation with some name brands. EAGET brand, if that’s not Dollarama’s own.

    1. Unless you have the setup it seems like the real perk is not having to deal with eMMC’s BGA packaging.

      If you do have the appropriate gear or are doing volume production eMMC is almost certainly the better bet, higher performance and a lot more vendors; but it’s nice to have the option of shoving an actual block device into what looks like pretty much exactly the same footprint and friendly solderability of a classic teeny SPI or i2c flash chip.

  1. A niche product as I’ve seen (micro)SD cards that are just soldered to PCBs. Depending on who is selling this, I wouldn’t be surprised if they were actually just old SD cards that were repackaged.

  2. Problem with large SD card storage is that it can’t handle unclean power loss. It can even corrupt the card to the point where it can’t be accessed at all.

    Another problem is the impossibility of doing any kind of file system check/repair on a small embedded platform.

  3. Yeah, right–
    I’m going to solder in a limited-lifetime memory chip into a product / project.

    “…There is already a thing called eMMC which is a bit the same…”.

    Yes, there is; and, no, it is nowhere near the same.

    When did this brain-dead “out of sight, out of mind” and “it works for me right this minute; not my problem any more” mentality get firmly entrenched with people who are supposed to know better?

    ***********************************************************************************************

    “Originality is no excuse for stupidity.”—Fred Brooks

    “I don’t care if it works on your machine‭. ‬We are not shipping your machine‭!”–‬Vidiu Platon

    1. It’s limited *use*. So use it as a golden image backup with a 2:1 SD mux (or just handle the CS pin). Now there’s an easy option to avoid failed updates without costing another interface, and you’ll never see the lifetime issue.

    2. When consumers started not only accepting but actively defending throwaway, locked down, closed ecosystem crap.

      Probaby because silicon valley has a thing for apple and their “flowing” kind of mindset where stuff is meant to be thrown away every few years, you don’t expect too much from devices, and private is the one and only concern in tech.

  4. The datasheet has no specs on the endurance of the flash, so I wouldn’t use these devices in anything. By itself, the idea isn’t bad. Lots more devices have SD or SPI interfaces than straight NAND controllers, and outsourcing the raw flash handling to the SD card controller simplifies your own software. Soldering down the memory avoids the contact glitches you occasionally get with card sockets.

    However, an SD card literally lives and dies by the quality of the flash memory and controller inside – that’s why you pay so much more for industrial cards than the supermarket checkout line cards, and the manufacturer says practically nothing about either, only that it uses ECC somewhere.

  5. The HaD write-up completely bypasses the major negatives of this form factor: lock-in, planned obsolescence, no right to repair/maintain, on-and-on. Thank goodness it took next to no time for the Commenters here to fill the gaps in the woefully incomplete lead-in article. Or maybe this was a “Stealthy-Ad” for Adafruit and/or XTX Technology? If it was – it didn’t work. Oh and one more thing: “SD” cards have been available as Chip-on-Board” (CoB) lead-bond devices since inception. Yes, they’re even cheaper because there’s no need for a physical outer package or leads. You just squirt a blob of epoxy over the chip after dropping it on the PCB.

    1. You… really don’t see the difference between a chip-on-board package and a DFN8? Really?

      Plus railing about “no repair!” is a bit nuts: obviously a CoB package under epoxy is nigh-impossible to repair, whereas a DFN8 available for pennies is pretty easy.

    1. Depends on the arduino, the ATmega based boards can’t run native code from external memory. You can use a interpreter to run some sort of byte code from external memory with the ATmega based boards (slow but possible). I don’t know about the other boards.

      1. This isn’t actually loading code externally: it’s completely erasing/reprogramming the flash each time. ATmega flash realistically has over 1M erase/program cycles (it’s rated for far less, but this is from testing) so for *casual* reloading this isn’t a bad option.

  6. I use these ICs in ‘parallel’; that is, one for reading and one for writing; never both. Writing is sequential and per a fixed-size (static data). Power monitors allow buffers and in-progress data structs to be flushed to IC before shutdown. Some of the libraries I use follow.

    https://github.com/dirkenstein/SPIMemory

    https://github.com/kasjer/saleae_spiflash

    https://github.com/PaulStoffregen/LittleFS

    https://github.com/littlefs-project

  7. If I were to use this, it would be in a removable socket but inside a product that would only read from It at boot up then use it to direct to another storage medium (etc sd card or usb drive)
    The ic could be removed but 99% of the time wouldn’t need to.
    A lot of microprocessors I’ve seen like for example Broadcom arm cpus in raspberry pi’s mainly boot from sd.
    If we could make it boot from a built in device similar to a bios then direct it to any other storage medium, it would be useful for board manufacturers.
    I imagine it similar to a commodore era rom chip vs floppy disk
    (I am not by any means a tech guru, please correct me if I’m wrong)

  8. An interesting device from solely the perspective of its’ physical characteristics. It is _substantially_ smaller than a micro-SD card, and is something that can be used in a design where you don’t have a chip-bonding machine (referring the comment above regarding CoB and DFN) or access to one (ie: I don’t want to make 100,000 or more).

    Anyone know if this size chip (physical size) is available with a USB interface rather than SD interface?

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.