As mentioned previously, all of these advancements would typically warrant their very own generation and be the prime focus of said generation’s marketing material to Enterprise consumers. With this generation, they almost get lost in the noise… as they represent just the tip of the iceberg that is the Exos M 30TB generation.
(Image curtesy of https://tenor.com)
This brings us to the pièce de résistance of the Exos M 30TB – HAMR. Yes… It’s finally Hammer Time™! This quite literally required Seagate to invent an entirely new way of recording data to spinning rust. This means new platters made out of new materials. It means new read heads. It means radically different write heads. HAMR basically represents a complete overhaul of the Seagate Exos ‘guts’.
Let’s actually start with the “read” head and the platter tech. Let us be clear. The reader header used is not a ‘first-generation’ anything. Rather, as the name suggests, Seagate’s Gen 7 Spintronic Reader builds upon a long line of technological evolution in magnetic storage. As such it is pretty much an evolutionary feature. Not a revolutionary one like the write head. Though, to be fair, one can still easily argue it is a radical refinement over its predecessor. One that took a lot of engineering to get all that miniaturization of the MTJ sensor arrays and their manufacturing process juuust right. That however, is getting a smidgen ahead of ourselves. Let’s rewind the tape and start with the Exos X 24TB and how the “last of the classic PMR’s” did things.
(Original Image courtesy of Rizal, Conrad & Moa, Belaid & Niraula, Boris. (2016). Ferromagnetic Multilayers: Magnetoresistance, Magnetic Anisotropy, and Beyond. Magnetochemistry. 2. 32. 10.3390/magnetochemistry2020022.
“Read” heads inside a modern HDD is a bit of an oversimplification of what they actually do. Yes. They “read” the polarity of a given bit and report back to the controller what is stored there. But how they do it is based on (still) highly advanced tech and physics. As such, instead of calling them ‘read heads’, it is more accurate to call them Magnetic Tunnel Junction (MTJ) sensors.
These MTJ sensors operate on the Tunneling Magnetoresistance Effect, where the electrical resistance varies depending on the relative orientation of magnetic layers separated by an insulating (typically ‘MgO’, magnesium oxide) barrier. Basically, the “platter” in a modern drive is not a solid material with specs of iron impregnated into it… and the ‘reader’ does not just read ‘north’ or ‘south’ to translate those specks into bits.
Instead, the platter consists of layers of magnetic and non-magnetic material stacked into an extremely precise configuration. One which includes a reference layer with a fixed magnetic orientation (stabilized by synthetic antiferromagnets)… which turns it from a mere lattice into a miniaturized magnetic tunnel junction stack. To ‘read’ a given bit the reader header measures the resistance at a given spot on said “platter” and this resistance correlates with either a ‘0’ or ‘1’. Rinse and repeat… really quickly, and you get the basics for how CMR drives did things.
The problem is… while Seagate could ‘easily’ (and precisely) control the magnetic layers’ thickness, they were constrained on how tightly data could be packed on the disk surface, limiting maximum areal density. Practically speaking, 30nm (possibly….maybe… theoretically as small as 20nm) is about as tight as one could make a ‘bit’ and still reliably read from it using this older platter and MTJ technology.
So in typical Seagate fashion, they went back to Ye Olde Drawing Board and figured out how to get around these limitations… and then make it just as reliable. The first thing they did was yeet their existing lattice (which we shudder to think about how many billions in R&D were ‘wasted’ on this ‘dead tree branch’ technology) and replace it with a ‘superlattice’ consisting (mainly) of Iron and Platinum (“FePt”… and MgO).
This new manufacturing results in darn near atomic scale precision in layer thickness (of all the layers and not just sensor layers), uniformity, and oxidation control. Resulting in nanosized (4 to 8, with 5 being the target) magnetic particles. Resulting in a “platter” that can store 1.5TB per side (instead of “just” 1.2TB). A “platter” that can take the thermal stress of constant heating/cooling cycles HAMR demands. Resulting in a “platter” that is just as robust and just as capable and withstanding the centrifugal forces applied to it as past platter tech. Better still, it results in a platter design that can do all that and is thinner than its predecessor. Allowing for not just the same “up to” 10platters (as seen here) but “up to” 12 platters per drive configurations (as seen by the limited release Exos M 36TB)!
(Image curtesy of Kautzky, Michael & Blaber, Martin. (2018). Materials for heat-assisted magnetic recording heads. MRS Bulletin. 43. 100-105. 10.1557/mrs.. 2018.1.
Circling back to the “read” heads, this is where “Gen 7 Spintronic Reader” technology (finally) enters the chat. Yes, it does incorporate multiple reading sensors to help mitigate signal interference between adjacent tracks via “TDMR” (Two-Dimensional Magnetic Recording) technology. Yes. It is quite likely that TDMR’s role will only increase as aerial density increases and nanosized particles get even… nanoer… err…smaller. No, they did not just grab their last-gen read head and miniaturize its components so that what once easily covered ~30nm now only covers ~10 to 15nm. Yes. They did make it all smaller and lighter to help offset the increased weight of having the ‘write’ include a freakin’ laser on its head. Yes. They did “cheat” a bit by using magnetic layers with a higher magnetoresistance, which in turn makes the electrical return signals stronger and easier to process. All of that is of secondary importance.
Simply put, Seagate not only had to make the dual read head design noticeably smaller, lighter, and more sensitive they also had to make it do all that and be resistant to heat. In layman’s terms, extremely precise magnetic sensors do not like heat. It downright hashes their mellow, and they quickly get hangry. Yet, the simplest answer of giving them more thermal mass (and insulation) would not work either, as they had to be lightweight. Thus, Seagate had to figure out how to make them smaller, more accurate, more robust, and weigh less than ever before. Anything less than success in all these key areas and a HAMR drive would have faced either unacceptable read error rates, required slower read speeds, or even more error correction overhead baked into the write process. Since they did indeed do all that, this read head is packed with advanced features, and the amount of engineering that went into the Gen 7 is astounding.
