Yeah, that is pretty serious, can probably run that drive for a few seconds at least. I'd imagine they would be using a series parallel setup with two groups of two in series and then the two groups in parallel, unless they are running them on like a 1.8v bus. I dunno.
My back of the envelope calculations put the run-time closer to 2 minutes at peak power consumption or 6 minutes at idle, which is cray-cray. I might have gotten the math wrong though.
Generally these PLP circuits, as in the Intel 730, are in parallel series, in case one circuit is faulty as judged by the SMC. The SSD 320 had a simple series circuit with a bunch of small non-electrolyte caps on the PCB, foregoing redundancy because of its consumer roots. (The SSD 730, although consumer, is basically identical to the 3500/3700 PCB)
So actual runtime would only include two capacitors, which is twice as many as the two TOTAL capacitors on the Intel SSD 730/3500/3700, only one of which operates the drive long enough to flush the indirection table from cache to NAND, or about 1200ms, assuming total power consumption is at peak (and it usually is when entering flush-shutdown state) or about 5w.
Without knowing more about this drives controller, number of NAND and other power parameters, it's impossible to say how long it will operate on 7F @ 1.8v but even assuming 100% efficiency from the capacitor circuit, it won't be more than a few seconds, way more than it needs unless the cache is in the gigabytes and the drive is incredibly power hungry. My guess is the PLP circuit is over engineered because it covers all configurations of the drive up to 2TB loaded with NAND and using a 1GB buffer/indirection cache.
I had voltage way too high in my calculation, but even still. The SSD DC S3700 only has a pair of 35V 47µF caps. We can presume that power consumption when the PLP circuit kicks in would be similar to average power consumption during sustained writes. The Samsung SSD EVO 2TB is rated at 4.7W max during sustained writes. Unless this thing is way more power hungry (or hotter) than other drives on the market, I'd say 5W is a safe upper bound.
So even half the caps here (7F) could deliver 1.8V, 2.8A for 4.5 seconds.
Keep in mind, the Intel drives are special in that they are the only SSD's out there that use a 12v rail, hence the need for 35v caps. The face that the Renice drive uses <5v caps is seriously interesting because it means the SMC, controller and memory/NAND are all running somewhere between 1.8v and 3.3v. This means it is probably high draw, and will drain those caps very quickly. They appear astronomically large on paper but physically are only 1.8mm larger diameter and 0.35mm longer. The Intel cap is mostly wasted because it is capable of charging 35v but handles only 12v. Still, the run time for the Renice drive is going to be seconds, however much longer than the Intel drive I am uncertain, but not significantly longer, maybe 5x?
@Einy0: Solid caps are vastly bigger and more costly per F: they can't do the job. Milspec electrolytics are rated for 105C and some I think to 125C, plenty of heat tolerance.
I agree. To get solid ceramic caps in this capacity would be prohibitively expensive, if they are even available...they are quite large for their charge capacity as well because the bulk of their size is insulation (which is why they tolerate temp extremes) but electrolytic caps are pretty stable for just about any application on Earth. The other benefit of solid caps is ripple suppression which is irrelevant for this application because we are going to assume the power supply is equally as rugged as the SSD.
The caps might double as some sort of emergency destructive erase. The placement of the button and the application for which the drive seems to be designed fits the narrative better.
Man, those connectors. I want to see them replace the official connectors as standard. ... If only the SATA-IO had considered the slightest possibility of abuse when designing the current connectors. Flimsiest parts in my entire PC are SATA connectors.
Those are really nice connectors. I don't see why everything has gotten so fragile over the years. USB connectors are pretty robust but the sockets are not. SATA is a joke, I see motherboards with broken connections (often just cracked, still working) in the shop quite often, especially on Dell's and HP's because they use locking SATA cables and people never notice the spring latch. I've seen people rip DisplayPort connectors out the side of their laptops because they didn't even notice the spring latch! But the alternative is HDMI, which literally just falls out over time on connections that face down (which is why most displays, particularly TV's, have them facing sideways.
It's a start contract to the 80's and 90's with ribbon cables that were sometimes SO jammed in you'd rip the cable or crack the plastic compressor strip removing them. Starting with ZIF sockets there was this trend toward more user friendliness at the cost of component integrity/lifespan. And the annoying this is the manufactures/standards agencies know the initial specs are weak, which is why almost every standard since, and including, USB, has had a soft recall of connectors, sockets, or cables, shortly after their initial public adoption. In the case of SATA, it started with SATA (1.5Gbps) cables that inherently couldn't communicate at 3Gbps when SATA2 was introduced, leading to a lot of data corruption. With the SATA3 (6Gbps) spec there was a integrity algorithm implemented during init that would downgrade the connection speed if the cable wasn't up to snuff (this init detection wasn't implemented for SATA2)
No need to go into detail on Thunderbolt, we all know it's vast shortcomings, which are now finally being sorted with USB-C. Ironically, IEEE1394 that it spiritually succeeds was one of the few connection standards that was very robust from the get-go.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
16 Comments
Back to Article
woggs - Friday, March 25, 2016 - link
14F of capacitance. Wow... That's some serious brute-force power loss protection.extide - Friday, March 25, 2016 - link
Yeah, that is pretty serious, can probably run that drive for a few seconds at least. I'd imagine they would be using a series parallel setup with two groups of two in series and then the two groups in parallel, unless they are running them on like a 1.8v bus. I dunno.repoman27 - Friday, March 25, 2016 - link
My back of the envelope calculations put the run-time closer to 2 minutes at peak power consumption or 6 minutes at idle, which is cray-cray. I might have gotten the math wrong though.Samus - Saturday, March 26, 2016 - link
Generally these PLP circuits, as in the Intel 730, are in parallel series, in case one circuit is faulty as judged by the SMC. The SSD 320 had a simple series circuit with a bunch of small non-electrolyte caps on the PCB, foregoing redundancy because of its consumer roots. (The SSD 730, although consumer, is basically identical to the 3500/3700 PCB)So actual runtime would only include two capacitors, which is twice as many as the two TOTAL capacitors on the Intel SSD 730/3500/3700, only one of which operates the drive long enough to flush the indirection table from cache to NAND, or about 1200ms, assuming total power consumption is at peak (and it usually is when entering flush-shutdown state) or about 5w.
Without knowing more about this drives controller, number of NAND and other power parameters, it's impossible to say how long it will operate on 7F @ 1.8v but even assuming 100% efficiency from the capacitor circuit, it won't be more than a few seconds, way more than it needs unless the cache is in the gigabytes and the drive is incredibly power hungry. My guess is the PLP circuit is over engineered because it covers all configurations of the drive up to 2TB loaded with NAND and using a 1GB buffer/indirection cache.
repoman27 - Saturday, March 26, 2016 - link
I had voltage way too high in my calculation, but even still. The SSD DC S3700 only has a pair of 35V 47µF caps. We can presume that power consumption when the PLP circuit kicks in would be similar to average power consumption during sustained writes. The Samsung SSD EVO 2TB is rated at 4.7W max during sustained writes. Unless this thing is way more power hungry (or hotter) than other drives on the market, I'd say 5W is a safe upper bound.So even half the caps here (7F) could deliver 1.8V, 2.8A for 4.5 seconds.
Samus - Sunday, March 27, 2016 - link
Keep in mind, the Intel drives are special in that they are the only SSD's out there that use a 12v rail, hence the need for 35v caps. The face that the Renice drive uses <5v caps is seriously interesting because it means the SMC, controller and memory/NAND are all running somewhere between 1.8v and 3.3v. This means it is probably high draw, and will drain those caps very quickly. They appear astronomically large on paper but physically are only 1.8mm larger diameter and 0.35mm longer. The Intel cap is mostly wasted because it is capable of charging 35v but handles only 12v. Still, the run time for the Renice drive is going to be seconds, however much longer than the Intel drive I am uncertain, but not significantly longer, maybe 5x?lilmoe - Friday, March 25, 2016 - link
The do have random speed info (crystaldisk) on their industrial SSD. I'd imagine it delivers the same performance as the military grade option.http://www.renice-tech.com/html/2014/06/18/2014061...
lilmoe - Friday, March 25, 2016 - link
**They...Einy0 - Friday, March 25, 2016 - link
Interesting electrolytic caps in extended temperature range military class product. I would opt for some sort of solid cap for a product like that.Samuel Lord - Sunday, March 27, 2016 - link
@Einy0: Solid caps are vastly bigger and more costly per F: they can't do the job. Milspec electrolytics are rated for 105C and some I think to 125C, plenty of heat tolerance.Samus - Sunday, March 27, 2016 - link
I agree. To get solid ceramic caps in this capacity would be prohibitively expensive, if they are even available...they are quite large for their charge capacity as well because the bulk of their size is insulation (which is why they tolerate temp extremes) but electrolytic caps are pretty stable for just about any application on Earth. The other benefit of solid caps is ripple suppression which is irrelevant for this application because we are going to assume the power supply is equally as rugged as the SSD.thope - Saturday, March 26, 2016 - link
The caps might double as some sort of emergency destructive erase. The placement of the button and the application for which the drive seems to be designed fits the narrative better.Anato - Saturday, March 26, 2016 - link
Even cased I doubt those caps will stick there in "16.4G (10 – 2000 Hz) sustained vibration".Lord of the Bored - Saturday, March 26, 2016 - link
Man, those connectors. I want to see them replace the official connectors as standard....
If only the SATA-IO had considered the slightest possibility of abuse when designing the current connectors. Flimsiest parts in my entire PC are SATA connectors.
bigboxes - Saturday, March 26, 2016 - link
YesSamus - Monday, March 28, 2016 - link
Those are really nice connectors. I don't see why everything has gotten so fragile over the years. USB connectors are pretty robust but the sockets are not. SATA is a joke, I see motherboards with broken connections (often just cracked, still working) in the shop quite often, especially on Dell's and HP's because they use locking SATA cables and people never notice the spring latch. I've seen people rip DisplayPort connectors out the side of their laptops because they didn't even notice the spring latch! But the alternative is HDMI, which literally just falls out over time on connections that face down (which is why most displays, particularly TV's, have them facing sideways.It's a start contract to the 80's and 90's with ribbon cables that were sometimes SO jammed in you'd rip the cable or crack the plastic compressor strip removing them. Starting with ZIF sockets there was this trend toward more user friendliness at the cost of component integrity/lifespan. And the annoying this is the manufactures/standards agencies know the initial specs are weak, which is why almost every standard since, and including, USB, has had a soft recall of connectors, sockets, or cables, shortly after their initial public adoption. In the case of SATA, it started with SATA (1.5Gbps) cables that inherently couldn't communicate at 3Gbps when SATA2 was introduced, leading to a lot of data corruption. With the SATA3 (6Gbps) spec there was a integrity algorithm implemented during init that would downgrade the connection speed if the cable wasn't up to snuff (this init detection wasn't implemented for SATA2)
No need to go into detail on Thunderbolt, we all know it's vast shortcomings, which are now finally being sorted with USB-C. Ironically, IEEE1394 that it spiritually succeeds was one of the few connection standards that was very robust from the get-go.