Register for the iXsystems Community to get an ad-free experience and exclusive discounts in our eBay Store.

How to measure the drive spin-up peak current

Status
Not open for further replies.

MindBender

Member
Joined
Oct 12, 2015
Messages
61
Yeah, but I don't like coffee :p
I don't understand that phrase. I read the words, but there's just no meaning...
NB: it's not the 12 V power but the 12 V + 5 V current, I know you like to be picky on these kind of things :)
So some of those amps aren't at 12V, but at 5V. That makes it slightly better, but it's safer to calculate with all of them being at 12V, because they load the 5V rail only lightly.
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Yep, but I think it's better to also keep an eye on the current on the 5 V rail too when you chose a PSU, just to be sure you're not over the max ;)
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
I don't understand that phrase. I read the words, but there's just no meaning...
Pretty sure he's just not had his morning coffee yet. I know sometimes I don't make sense until I've had a bunch of coffee.
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
The Seagate NAS 4TB with a 3.25 A peak and more than 3 A during 2 seconds\0/
Wow, yeah.

I think you can update the capture in the sticky as it's more representative of the kind of drive we use in our NAS.

NB: it's not the 12 V power but the 12 V + 5 V current, I know you like to be picky on these kind of things :)
If I wanted to be picky, I'd scold you and say we just wanted the 12V. I am *guessing* that the 5V fluctuates much less, and that you've primarily introduced an offset of a slightly unpredictable sort. Still, I'm not going to criticize the work because you spent the time to do it, when I've mainly been blathering on about it.

Edit: didn't think of it at first but now it's confirmed that when the datasheet says 2 A on the 12 V rail it's true (not just a max value not attained IRL) and that it's in addition to the 4.8 W during normal operation.
anigif_enhanced-12000-1406595908-3.gif
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
Yep, but I think it's better to also keep an eye on the current on the 5 V rail too when you chose a PSU, just to be sure you're not over the max ;)
I don't disagree with this, but in practice I haven't seen this be a problem in ... at least a decade? ... for the sorts of stuff done with a NAS. Look at the amps available on a Seasonic G 550W:

  • +3.3V@20A, +5V@20A, +12V@45A, -12V@0.3A, +5VSB@2.5A
And your typical mainboard isn't consuming much on the 5V anymore, much of a mainboard's draw is on the 12V for the CPU, and I'm guessing some of it ends up on the standby rail too.

TL;DR: don't worry about it.
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
Wow, yeah.
Yep, I double checked that I didn't mess-up the measure setup when I saw that but I guess the number of platters makes a big difference.

I am *guessing* that the 5V fluctuates much less, and that you've primarily introduced an offset of a slightly unpredictable sort.
Not totally unpredictable actually, the datasheet says 0.3 A average on the +5 V during operation. And the label on the drive says 0.55 A. So I guess the average is at 0.3 A with peaks at 0.55 A but during spin-up the +5 V current seems to be about 0.5 A (by making educated guesses on the waveform) so this means the +12 V current is about 2.75 A during spin-up but from the datasheet it should be 0.25 + 2 so there's an additional 0.5 A.

Well, I guess that's why it's a good idea to take a 20 % margin when we chose the PSU :)
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
Well, I guess that's why it's a good idea to take a 20 % margin when we chose the PSU :)
I still think that in many cases it'll average out, but, boy, that's actually a little scarier than I thought it was, AND I'M SUPPOSEDLY THE PARANOID ONE.

I'd love to know what kind of damage actually happens to a PSU / hard drive / etc in the case where you've undersized it, but truthfully I'm just not rich enough to bankroll an experiment of any meaningful size. The problem is I've seen arrays of thousands of hard drives for many years now, and I've seen both modular PSU failure issues and hard drive failures in apparent batches, but since they were customer machines that I was only seeing every few months, and I don't have sufficiently detailed service records on what was done, all I can say for sure is that I knew there were loose correlations between machines with failed supplies and additional failures.

But I seriously appreciate your efforts in looking into this, because I was getting a little OCD about it and was possibly going to go and figure out some answers myself. You've done a gorgeous job, much better than I would have, and the answers seem clear.

By the way, could I talk you into putting a link to the PSU sizing sticky in the first post in this thread? I've been trying to encourage making it easier for people to find information on the forums through frequent and extra links. Also I strongly suggest you change your forum label from "FreeNAS Guru" to "Server Sorcerer" or something like that ;-) You've done a ton of stuff that merits some extra recognition.
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
@jgreco: Bah, you're not the paranoid one if you don't have a RAID-Z3 with only 8 drives... :D

Yep, I'd love to know too. If I had the money I'd do a ton of experiments (starting with the good old "whats is the limit (how many times per day) to spinning up/down the drives for saving some drive life") but yeah, all I can do for now is doing a fair comparison between the WD Red and Seagate NAS (that's why I have 4 of each).

Yep, I'll add a link, no problem. I'll also complete the missing parts while I'm at it.

Thanks a lot ;) I'm maybe very tired but I can't find an option to change this label in my profile, can you tell me where it is exactly?

@anodos: Thanks ;)
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
Top right, click your login name. Personal Details -> Under Avatar, "Custom Title"
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
I don't have this option, the next thing under the avatar is the gender option. I'm starting to guess that only mods and admin have access to this option.
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
I'd love to know what kind of damage actually happens to a PSU / hard drive / etc in the case where you've undersized it, but truthfully I'm just not rich enough to bankroll an experiment of any meaningful size.
I can tell you what happens.

Your voltage regulators may not be able to maintain voltage in-spec. Anyone with electrical theory for motors under their belt knows that as voltage drops the current must increase to get the same power. So the motor now draws more current because there is less voltage. This generally isn't a big deal because the starting current is generally expected to be lower than what the disk is engineered to handle, but it is notworthy from the cabling standpoint. Many cables cannot handle all that starting current from all the disks plus the added bonus of having voltage low out of spec. If voltage dips enough your hard drives may think that a bad voltage condition has developed and trip offline to protect themselves. Or they may simply go into a spindown/spinup cycle to try again in 10 seconds or so.

Also on the load side you may have problems with noise and ripple because your PSU caps aren't designed to handle that much load. They simply don't have enough capacitance to deal with the added workload. This can shorten the lifespan of the capacitors as well as slowly damage them by deteriorating the insulator and electrolyte inside increasing the rate at which they fail.

Ultimately you'll start burning out components on the hard drives, the cabling powering the hard drives, or the PSU. The question is "which one is engineered with the least margin to failure" as that is the component most likely to fail first. So how do you deal with it? Overengineer the system so your "weakest link" is never so weak that you can't deal with the workload. AKA "buy a bigger PSU".

It's all very scary though because when/if the PSU fails, there's the possibility that it may overvolt (or undervolt) all of the components on a particular bus. That can burn out everything attached to that bus, which for the 12v bus is basically "the entire system".
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
I completed the other measurement methods sections ;)

It's all very scary though because when/if the PSU fails, there's the possibility that it may overvolt (or undervolt) all of the components on a particular bus.
I thought good quality PSUs have OVP and UVP, did I misread?
 

jgreco

Resident Grinch
Moderator
Joined
May 29, 2011
Messages
12,732
Also on the load side you may have problems with noise and ripple because your PSU caps aren't designed to handle that much load. They simply don't have enough capacitance to deal with the added workload. This can shorten the lifespan of the capacitors as well as slowly damage them by deteriorating the insulator and electrolyte inside increasing the rate at which they fail.
Well, I understand in *theory* what happens, after all I minored in EE, and I learned paranoia doing work for a medical electronics company, but the interesting question is whether or not all of this stuff is really engineered with sufficient margin that maybe damage doesn't happen as rapidly as we fear. That's where real world testing would be fun. :smile:
 

cyberjock

Inactive Account
Joined
Mar 25, 2012
Messages
19,526
I thought good quality PSUs have OVP and UVP, did I misread?
You didn't misread. There's a reason why I said "possibility". Many PSUs that we would define as "good quality" do have OVP and UVP. Not all though, and unless you go looking at a particular model's insides or product propaganda, you'll probably never know for 100% certainty.

Most "good" PSUs should be able to do something ranging from 125% to 150% of designed static load for 10 seconds when cold. But (and here's the big "but") you aren't supposed to design a system that relies on that extra margin because it basically means that if you power off the system, you're committed to not powering it back on for several hours so all of the components can be "cold". I believe that the ATX spec says that the PSU has to be able to withstand 110% for 10 seconds when cold.

I've seen systems that are powered off and on weekly and *did* operate above 100% for the first 10 seconds. They lasted several years for that setup (long enough to be decommisioned). That being said, and considering that you shouldn't be powering your system on and off regularly anyway, if you were to go over 100% a couple of times per year it probably wouldn't ever matter. The only problem with giving a blank statement that it's okay to go over 100% is that you are basically telling someone to take the risk without a full understanding of what they are accepting and we can't generally judge what is a "good" PSU as well as assuming they aren't powering their box off every night like a small percentage of users do despite all of the warnings.

How many people have said they own a "Cyberjock PSU-450" and claimed it was a good brand because it's popular in their country when we all know its total crap?

There is lots of engineered safety margin in these PSUs, but if you are having to rely on them, you've screwed up. As they say at NASA and nuclear power "if you are relying on safety margin to operate, then you have no true safety margin". If memory serves me right, after the Challenger disaster, the appendix that was written by Richard Feynman as part of the accident investigation team said something like that when he discussed all of the management issues with NASA and other contracted companies.

So I think the conservative values we've been giving people is still the way we should continue to recommend PSUs.
 

Bidule0hm

Server Electronics Sorcerer
Joined
Aug 5, 2013
Messages
3,710
You didn't misread. There's a reason why I said "possibility". Many PSUs that we would define as "good quality" do have OVP and UVP. Not all though, and unless you go looking at a particular model's insides or product propaganda, you'll probably never know for 100% certainty.
The only problem with giving a blank statement that it's okay to go over 100% is that you are basically telling someone to take the risk without a full understanding of what they are accepting and we can't generally judge what is a "good" PSU as well as assuming they aren't powering their box off every night like a small percentage of users do despite all of the warnings.
How many people have said they own a "Cyberjock PSU-450" and claimed it was a good brand because it's popular in their country when we all know its total crap?
"if you are relying on safety margin to operate, then you have no true safety margin"
I agree at 150 % with all of this :D
 
Status
Not open for further replies.
Top