Subject "analog":correction Re: [PLUG] ... desktop drive Reply-To: In-Reply-To: <[email protected]>
============ On Mon, Dec 08, 2025 at 08:54:54AM -0800, Rich Shepard wrote: ... > SSD would likely last longer than another 5" HDD. ... > I am asking for recommendations on which brand to purchase. ============ On Mon, Dec 08, 2025 at 03:00:16PM -0800, Keith Lofstrom wrote: ... > move disk content to Samsung 870 EVO drives. ... > Those are "three bits per memory cell" drives (three voltage > levels), a compromise between cost and my concern with errors **** correction: EIGHT voltage levels > Samsung also peddles QVO "four bits per memory cell" drives > (sixteen voltage levels per cell). That's either spectacular ... > The Samsung PRO series is TWO bits per cell, more expensive **** correction add: FOUR voltage levels > but faster and probably more survivable. ============= MORE INFO Besides pedantically correcting my error, I will attempt to explain (as an analog CMOS chip designer) WHY four cell levels is more robust than eight or 16 cell levels, and what those levels do. Also, why it is so INCREDIBLY SUPER AWESOME that Samsung can make chips with TRILLIONS of ***INCREDIBLY PRECISE*** MOSFET transistors, not just used as 1/0 switches, but as ANALOG LEVEL MEMORY, which is converted to digital bits using VERY fast, precise, adaptive analog-to-digital converters and subsequent digital error correction, and doing ALL THAT with nanojoules of power per readout bit. If the digitization levels or the sources shift, incorrect bits emerge. I can imagine a read circuit that repeats the read after dithering the bit threshold levels, rinse and repeat (at gigahertz speed) until the checksums match. I can't imagine dithering threshold levels and read head position on a 7200 rpm spinny disk with 8.3 milliseconds between attempts, rather than nanoseconds. Decades ago, before starting my own businesses, I was a newbie analog integrated circuit designer at Tektronix, working with some of the best analog chip designers IN THE WORLD (better to be #20 of the top team rather than #1 of a middle team). After years of collaboration on other projects, I led some recent hires to design a 100 megasample-per-second 10 bit analog-to-digital converter, first in a planned series of faster 12 bit converters. Tricks such as gray-scale encoding on the comparator ladder, followed by a few pipeline registers for metastability resolution, THEN the digital conversion to integer binary. Otherwise, you get "sparkle codes" where (say) 1101111111 resolves to 1101100000 because the "deciding analog comparator" cannot make a decision before raw comparator outputs are converted to binary. If you see oddball "many zero" or "many ones" errors in a digitized data stream, that may be because the converter designer did not understand metastability. ---- I annoyed others by calling that first product a 0.1 GIGAsample-per-second converter, sort of a "you ain't seen NOTHIN' yet" brag about my future plans. Sadly, Tektronix sales dropped (DOE stopped irradiating hundreds of Tek scopes and digitizers per nuclear test). Tek's founders transferred "leadership" to bean counters. I left to do my own businesses, including "non blocking crossbar" integrated circuits, designed for hardware logic simulators but also useful for internet backbone routing. I added my own channel quality test circuits to those crossbars, useful for production test but too complicated to describe to the customers using the chips. I used to think I was Hot Stuff, but seeing what Samsung has accomplished inside these SSD memory chips makes my businesses and products look like a lemonade stand. I write the above neither to brag or to grovel, but to point out that Samsung has accomplished PRODIGIOUS feats of engineering with their 870 SSD drive series, and that the United States deploys zillions of them. I'm proud to share a global economy with Samsung. However, Samsung's Suwon-si headquarters is 40 miles south of the North Korean border. I presume the design engineers are in the same region, though Samsung is opening a design center near Austin TX. The drives are assembled in Vietnam and their marketing arm is in Ireland and the UK, so perhaps they can regroup a few years after a North Korean invasion. I write about THAT not to fear the NKs or anyone else, but to point out that the US is heavily interconnected with the global economy. Protecting that economy may be essential to our individual personal survival. Failed real estate hucksters peddling nativism and tariffs may be more dangerous than the NKs. Or they may be brilliant visionaries. We must not bet all the marbles on our own wished-for outcome. For us, the technical lesson is clear; design systems and software that can survive a few years of interrupted international supply. Keep few spares. Learn to do more with less, and learn how to downsize infrastructure tasks and business targets if irreplaceable components fail. I hope the situation NEVER degrades that far, and prefer a rosy future. But as Mel Brooks wrote for his 1970 comedy The Twelve Chairs, "Hope for the best, expect the worst." Is that technical, or PLUG-TALK? I hope we can focus on the technical challenges and opportunities, design and code and deploy, recover from disaster, and earn big bucks, while the talking heads talk and talk. Keith L. -- Keith Lofstrom [email protected]
