| General > General Technical Chat |
| This Janet Jackson BASSLINE breaks laptops |
| << < (10/10) |
| tom66:
--- Quote from: Haenk on January 11, 2023, 11:00:20 am ---I'd say that is nonsense - if powered down, the platters still spin, so the heads are still up in the air (so to speak); the arm itself snaps back into parking position almost immediatly (*way* faster than the spindown duration) and the heads are then stabilzed by some plastic spacers and above a "crash zone" on the platters. So sudden power loss is usually no problem (other than maybe data loss due to unwritten data). The arm mechanics is really extremely solid and AFAIR there were even drives that were powered down permanently for energy savings (early WD Green?), I think those had powerup/down-cycles (which are mechanically equivalent to a power loss, minus the platter motor) in the hundreds of thousands or even millions. --- End quote --- Not sure how mainstream it is, but some Sandisk drives use the platter as a "capacitor" to generate enough power to run the park head mechanism. The motor driver 'regens' onto the isolated 12V bus to run some of the electronics on the board, and then once the head is parked the final action is to 'brake' the platter to a stop as quickly as possible. The data in the 64MB cache (if it's partially a write-delay cache) is presumably lost, as there isn't enough time to write that out. For SSDs, it's common to see an array of capacitors on the drive which are intended to provide enough worst-case runtime to empty the DRAM caches. e.g. https://images.anandtech.com/doci/9258/IMG_2329.jpg |
| wraper:
--- Quote from: BrianHG on January 11, 2023, 05:18:54 am --- --- Quote from: tom66 on January 10, 2023, 03:31:34 pm ---Even more miraculous is devices like the original iPod with 20 ~ 160GB micro hard disks inside them. These devices can play music while you are cycling and in my limited experience they never dropped out. --- End quote --- No. Apple's big thing about the IPOD is that their playback cache was stored in ram as .mp3 and that they had an hours worth of it which was loaded in around ~5 seconds. If you stopped moving for a few seconds, the drive could power up, load the next hour of music, and go back to deep sleep saving the battery life. --- End quote --- I doubt about hours, classic iPod had 32MB of RAM, later increased to 64MB in 5-6th gen. This is certainly not enough for hours of cache in acceptable quality. |
| ebastler:
20 minutes of buffering time is what they claimed (and achieved) for the early iPods. |
| tooki:
--- Quote from: tom66 on January 11, 2023, 11:09:09 am ---Not sure how mainstream it is, but some Sandisk drives use the platter as a "capacitor" to generate enough power to run the park head mechanism. The motor driver 'regens' onto the isolated 12V bus to run some of the electronics on the board, and then once the head is parked the final action is to 'brake' the platter to a stop as quickly as possible. The data in the 64MB cache (if it's partially a write-delay cache) is presumably lost, as there isn't enough time to write that out. For SSDs, it's common to see an array of capacitors on the drive which are intended to provide enough worst-case runtime to empty the DRAM caches. e.g. https://images.anandtech.com/doci/9258/IMG_2329.jpg --- End quote --- SanDisk doesn’t manufacture hard disks. They may at some point have sold external hard drives (though I couldn’t find any examples), but the actual hard disk drives inside would have been bought from someone else. Anyhow, it’s my understanding that hard disks do in fact use the motor as a generator to attempt to write our final data from cache when power is lost, but I can’t remember where I read that. I did find a Seagate patent describing using the motor as a generator to then copy the cache to a non-volatile cache (flash?). They definitely use the energy to park the heads, as you said. |
| tom66:
--- Quote from: tooki on January 11, 2023, 06:32:22 pm --- --- Quote from: tom66 on January 11, 2023, 11:09:09 am ---Not sure how mainstream it is, but some Sandisk drives use the platter as a "capacitor" to generate enough power to run the park head mechanism. The motor driver 'regens' onto the isolated 12V bus to run some of the electronics on the board, and then once the head is parked the final action is to 'brake' the platter to a stop as quickly as possible. The data in the 64MB cache (if it's partially a write-delay cache) is presumably lost, as there isn't enough time to write that out. For SSDs, it's common to see an array of capacitors on the drive which are intended to provide enough worst-case runtime to empty the DRAM caches. e.g. https://images.anandtech.com/doci/9258/IMG_2329.jpg --- End quote --- SanDisk doesn’t manufacture hard disks. They may at some point have sold external hard drives (though I couldn’t find any examples), but the actual hard disk drives inside would have been bought from someone else. Anyhow, it’s my understanding that hard disks do in fact use the motor as a generator to attempt to write our final data from cache when power is lost, but I can’t remember where I read that. I did find a Seagate patent describing using the motor as a generator to then copy the cache to a non-volatile cache (flash?). They definitely use the energy to park the heads, as you said. --- End quote --- Sorry, getting my Sandisk and Seagate mixed up - I do mean Seagate. I think the problem with flushing the DRAM is that it's likely going to be many sectors across the disk, and if it requires more than a few seeks, you'll deplete the energy in that disk pretty quickly. Also, now you need the controller itself to be running, which normally operates from a little regulator hanging off the 5V rail, and the controller needs to be able to manage with a varying disk speed rather than a fixed one, which probably creates all sorts of complications for managing the DSP stuff. Not impossible to imagine they could flush something like a bad sectors table, or flush RAM to flash. The latter would be an almost entirely deterministic operation. Maybe they could dump the DRAM linearly to the start of the disk in a reserved sector range? It'd need to recover this sector on the next power on, but might work. |
| Navigation |
| Message Index |
| Previous page |