Feeding SSDs from +5VSB for outage data corruption safety

[goontron]

Re: Feeding SSDs from +5VSB for outage data corruption safety

or just not use a ssd

[momaka]

Re: Feeding SSDs from +5VSB for outage data corruption safety

No, he could be right. Most 5VSB uses a simple flyback, which can typically operate over a very wide range. The main switcher won't work beyond about 20% of rated voltage.

So it's sort of like an early warning ac detect because the main rails collapse before standby.

I was testing a PSU last week and noticed that this is indeed true.

I had one light bulb on the 5VSB line drawing about 0.5W to 1W, one light bulb on the 5V rail drawing about 5W, and another on the 12V rail drawing about 20W. Also had an HDD hooked up (rated for about 600 mA on the 5V rail and 350 mA on the 12V rail).
When I flipped the switch on the back of the PSU while it was on, the main power stayed on for about a second and the 5VSB for at least another second after main power supply turned off.
This was on a 250W LiteON PSU from a Dell PC.

So I think it is possible.

or just not use a ssd

That's why I hide in a cave under a heavy rock. Never have to deal with any such problems, you know .

[goontron]

Re: Feeding SSDs from +5VSB for outage data corruption safety

That's why I hide in a cave under a heavy rock. Never have to deal with any such problems, you know .

well actually me and another Linux tech by the name pete were talking about the life of an SSD and how if you do stuff like compile from code ( usually 200+ mb ) you get some stupid amounts of writhe cycles that could cut the life of the SSD in half or more...

[mariushm]

Re: Feeding SSDs from +5VSB for outage data corruption safety

That has "ram drive" written all over it.

[goontron]

Re: Feeding SSDs from +5VSB for outage data corruption safety

That has "ram drive" written all over it.

and what about VMs? that's usually 3+ OSes plus the host! wheres your answer for that?????

[TELVM]

Re: Feeding SSDs from +5VSB for outage data corruption safety

Or just not buy crappy SSDs.

or just not use a ssd

Well as a matter of fact they aren't precisely fragile:

It's just the occasional nuisance of data loss (which spinners can also suffer under outages).

[TELVM]

Re: Feeding SSDs from +5VSB for outage data corruption safety

Curiously Intel uses just a couple good ole lytics in its SSD DC S3500 Enterprise:


They're 35V, why such a high voltage rating?

[mariushm]

Re: Feeding SSDs from +5VSB for outage data corruption safety

ESR, peak currents, longer lifetime etc etc

[tom66]

Re: Feeding SSDs from +5VSB for outage data corruption safety

Possibly charge them to ~30V then buck down to 12V though seems wasteful a similar thing is done in airbag computers for cars... in an accident the battery might be destroyed so they need a temporary supply to fire the airbags.

[budm]

Re: Feeding SSDs from +5VSB for outage data corruption safety

47uf (even if both of them are connected in parallel) is not going to give much of the hold up time.

[tom66]

Re: Feeding SSDs from +5VSB for outage data corruption safety

~100uF at 30V is 90mJ of energy. That's enough to supply 1 watt for 90ms, etc... if these SSDs use about 2 watts it gives a tiny amount of time to flush data--they had better be fast!

[cheapie]

Re: Feeding SSDs from +5VSB for outage data corruption safety

The SSD I have says 0.071W while active (0.046W idle). No idea how accurate that is.

[TELVM]

Re: Feeding SSDs from +5VSB for outage data corruption safety

... Try a Schottky diode (low Vf), to isolate the drive's power, and large electro across the 5V feed to the drive.

A problem with this approach is than even the best Schottky diode will drop voltage at least ~0.3V. That would lower the +5V feeding for the SSD below ATX spec (5-0.3 = 4.7V).

[tom66]

Re: Feeding SSDs from +5VSB for outage data corruption safety

If you design that into your product, that is fine. But you could also look at what the 5V is used for. I'd take a guess that 90% of SSDs will immediately regulate it to 3.3V or 1.8V, or whatever internal voltage they use, so as long as that is in regulation everything will be OK.

[TELVM]

Re: Feeding SSDs from +5VSB for outage data corruption safety

^ You're right Tom:

... typically, in an SSD the NAND flash is powered at 1.8 V (or sometimes 3.3 V), the SDRAM is powered at 3.3 V, and the controller is at ~1.2 V (although sometimes this may also be at 3.3 or 1.8 V) ...

http://www.electronicproducts.com/Pa...kup_power.aspx

(Warning - potential stupidity ahead ): I have four 6.3V 3300uF Panny MBZs (13.2mF total) and a STPS1545CT Schottky rectifier, all bored and looking for trouble. Would a concoction like this make sense?:

[tom66]

Re: Feeding SSDs from +5VSB for outage data corruption safety

Should work fine. Can't see anything obviously wrong, at least.

[mariushm]

Re: Feeding SSDs from +5VSB for outage data corruption safety

Other than the inrush at the start (a few amps on 5v) no problem.

But also don't see the point of the diode there, unless the voltage in molex comes from 5v standby.

[goontron]

Re: Feeding SSDs from +5VSB for outage data corruption safety

ok.... so where will the command to flush the cash to the disk come from?

[tom66]

Re: Feeding SSDs from +5VSB for outage data corruption safety

What you could do is "OR" the standby voltage into to provide extra hold-up, just connect other leg of diode to 5VSTB, you could have a whole second extra hold-up time.
The diode will essentially select the highest of the two voltages and draw current from that.

[c_hegge]

Re: Feeding SSDs from +5VSB for outage data corruption safety

But also don't see the point of the diode there, unless the voltage in molex comes from 5v standby.

I think the idea was to make sure the caps feed only the SSD, and that they do not feed power back into the PSU and other components.