How many uf do i need to cope with the HDD inrush current

[evilkitty]

In my computer i have 2 hot swap bays, i made my power cable and i used some "18 awg copper" wire, i think it is actually 20awg (got it on ebay...)

anyway i have some legit 18awg copper wire i was going to remake it and stick a couple caps at the end
my issue is when i am using one bay and i go to use the second the other bay looses power for a moment (if i use that drive to boot it as annoying)
i have some 470 low esr caps and i have some 10uf general purpose caps i could use as well as some 10uf low esr ones
my PSU is a Seasonic M12II 520W

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

In my computer i have 2 hot swap bays, i made my power cable and i used some "18 awg copper" wire, i think it is actually 20awg (got it on ebay...)

anyway i have some legit 18awg copper wire i was going to remake it and stick a couple caps at the end
my issue is when i am using one bay and i go to use the second the other bay looses power for a moment (if i use that drive to boot it as annoying)
i have some 470 low esr caps and i have some 10uf general purpose caps i could use as well as some 10uf low esr ones
my PSU is a Seasonic M12II 520W

You shouldn't be seeing a "dropout" like this -- unless your drives have some hellacious spinup currents and/or you have high IR drops in your cabling/connectors.

But, to answer your original question...

The voltage on a capacitor is a function of current and time (and capacitance). Specifically, CV=It.

"I" coming INTO the capacitor (remember, there's current flowing in while it is also being drawn out) will be limitted by the peak capabilities of your power supply as seen behind the impedance of your interconnect to the powersupply. "I" exitting the capacitor will be defined by the spinup requirements of your disk drive. So, you have a "net" I flowing out of the capacitor that reflects the difference in these two.

Similarly, "t" will depend on how long this "transient" lasts (presumably, your power supply can meet the steady-state operating demands of the disk drive).

"V" should be treated as "dV" -- a change of voltage over a particular change in time (dt -- time interval).

So, if your drive specs an input range of [Vmin,Vmax] and your power supply puts out Vout... and, the drive actually "sees" Vin (Vout behind the IR losses of your cables) then:
- Vin must be > Vmin (else you won't be able to run the drive)
- Vin must be < Vmax (d'uh)
- Vout must be < Vmax (i.e., when the load is minimized and IR ~= 0)
- Vin - Vmin is the dV that you have to work with; you can afford for the voltage at the drive to drop from Vin to Vmin and still be assured the drive will function.

[The capacitor starts out at Vmax before you plug in the drive -- cuz there are no IR losses due to the absence of a load!]

This is really just a crude approximation of the situation. In reality, your drive will probably work with a lower Vin on the 12V supply (motor) during spinup. OTOH, the 5V supply probably is a lot more fussy about Vmin as "it can lose its train of thought"

[One "simple" fix is to run a second power pigtail from the power supply so the load from "disk #1" isn't being handled by the cable that is ALSO powering disk #2 (see your photo)]

[evilkitty]

Re: How many uf do i need to cope with the HDD inrush current

high IR would not surprise me with the way the plugs are made and the sire binging thinner than it should
a stupid high spin up current would not be hard to imagine as one of the drives i have is old, really old and had a annoying wine (sounds similar to coil wine)

when i have this be an issue is when i boot off one and plug in a second drive, it crashed windows with the drive loosing power for a moment

Here is a list of the amps my drives pull
HDD | 5V | 12V
2.5A | 1.0A |
2.5B | 0.6A |
3.5A | 0.5A | 0.7A
3.5B | 0.65A | 0.5A
3.5C | 0.72 | 0.35A
3.5D | 0.68A | 0.55A
3.5E | 0.5A | 0.7A
3.5F | 0.9A | 0.9A
3.5G | 0.62A | 0.6A

I figure if i remake this cable i can slap a cap at the end of each rail to handle a spike in current draw

[redwire]

Re: How many uf do i need to cope with the HDD inrush current

Are these hot-swap bays any good, they are supposed to connect GND first on the connector and delayed switch-on power to the drive. So the 5V, 12V pins are recessed, no zaps to the Sata lines.

Otherwise, it's probably the inrush current of the plugged in HDD.
Drives are around 2A peak during spin-up. HDD controller boards have pretty much nothing for capacitance, so any glitch causes the drive to reboot.
Check the wiring/lead/connector resistance, -10% or 10.9V is the spec.

Try add a large 100-470uF capacitor on 12V power, and maybe another cap for the 5V rail. A few hundred uF is good.

If your problem is due to a crappy hot-swap connector, it could be zapping the Sata lines etc. during plug in and might damage the HDD?

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

high IR would not surprise me with the way the plugs are made and the sire binging thinner than it should
a stupid high spin up current would not be hard to imagine as one of the drives i have is old, really old and had a annoying wine (sounds similar to coil wine)

when i have this be an issue is when i boot off one and plug in a second drive, it crashed windows with the drive loosing power for a moment

I figure if i remake this cable i can slap a cap at the end of each rail to handle a spike in current draw

Are you sure the first drive is having a problem and not the CPU??

If your power distribution is such that this sort of sag can be present in the power leads (esp if you've moved to a single drive per pigtail), then I'd be suspicious of the suitability of your power supply for your particular SET of loads. I see ~50W of DC load on the 12V supply and ~30W on the 5V supply. That's not counting any power you're soaking up on the motherboard or video card.

In general, you don't want to go around changing the impedance of your load (adding C) without understanding its effect on the power distribution "system" (you can send a power supply into oscillation, etc.).

If your drives stagger their spin-ups AND if your system is within the normal operating margins for the power supply (which, hopefully, doesn't have shitty caps inside), then you shouldn't see problems like this. E.g., many of my systems have 4 or more 10-15K spindles (not counting the external 15 spindle enclosures) and nothing sags during power up.

Put a DMM on the power supplies (you can do one at a time as the problem is repeatable) to see if you're really seeing a notable sag, and not "something else".

[Or, shed some load and see when the problem goes away]

[evilkitty]

Re: How many uf do i need to cope with the HDD inrush current

Are these hot-swap bays any good, they are supposed to connect GND first on the connector and delayed switch-on power to the drive. So the 5V, 12V pins are recessed, no zaps to the Sata lines.

Otherwise, it's probably the inrush current of the plugged in HDD.
Drives are around 2A peak during spin-up. HDD controller boards have pretty much nothing for capacitance, so any glitch causes the drive to reboot.
Check the wiring/lead/connector resistance, -10% or 10.9V is the spec.

Try add a large 100-470uF capacitor on 12V power, and maybe another cap for the 5V rail. A few hundred uF is good.

If your problem is due to a crappy hot-swap connector, it could be zapping the Sata lines etc. during plug in and might damage the HDD?

they are just cheap bays, they are just well placed sata connectors, there is nothing fancy at all as far as i can see on them

@Curious.George Not the CPU never had a issue using the bays when running a OS off the internal bays (different power lead)

I am using a quality PSU and i am not even stressing it, it is either a bad connection and/or inrush current

[goontron]

Re: How many uf do i need to cope with the HDD inrush current

^ i have this issue as well. On my poweredge and my homebrew server. the SAS disk wait for the controller to bring them up, but SATA just spins right up.

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

they are just cheap bays, they are just well placed sata connectors, there is nothing fancy at all as far as i can see on them

Note that connectors not designed for hot-plugging (or eSATA) are usually only designed for a limited number of cycles (like 50!).

@Curious.George Not the CPU never had a issue using the bays when running a OS off the internal bays (different power lead)

I am using a quality PSU and i am not even stressing it, it is either a bad connection and/or inrush current

You misunderstood. The CPU never saw its power supplies "tweeked" the way your abruptly adding this load is doing. If the CPU isn't accessing ANY disk and "the system" crashes, then its possible the 5V (or any other supply that is applicable to whatever the COU is "thinking about") is glitching and causing the crash. I.e., that the drive insertion isn't interfering with the "other" drive(s) but, rather, the CPU/mobo itself.

Try booting a more quiescent OS (even DOS or a Linux CD) so you can have the disk interface largely idle when you choose to insert your "extra" disk. If the system survives (doesn't crash/reboot) this, then try starting a "lengthy" disk operation (e.g., dd if=/dev/firstdisk of=/dev/null bs=1024k count=1024) while you try to insert your "new" disk. If the disk operation proceeds as expected, then the act of inserting the new disk isn't interfering with the other disk's operation...

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

^ i have this issue as well. On my poweredge and my homebrew server. the SAS disk wait for the controller to bring them up, but SATA just spins right up.

I think AHCI support is required to gain some of the more "advanced" features (hot plugging).

[evilkitty]

Re: How many uf do i need to cope with the HDD inrush current

I think AHCI support is required to gain some of the more "advanced" features (hot plugging).

In addition to this on my board you have to enable hot swap on a per port basis

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

In addition to this on my board you have to enable hot swap on a per port basis

I avoid hot swapping bare drives (ditto power supplies)! There are just too many places where I think the system software can screw up -- and too much to lose if it does!

I have one server configured to support multiple OS's -- by swapping out the boot drive (instead of multiple partitions/slices). Of course, that means a reboot is required so that's a convenient/safe time to swap out the drive!

Any other time that I want to spin a drive up or down without disturbing the system its an "external" drive -- deliberately -- connected by USB, FW, SCSI, SAS, etc. (I keep "docks" and dismantled USB enclosures on hand that I can slap drives in and out of with ease).

Sure, I am still relying on software in the system to unmount the drive properly (at least in the Windows world; I can ensure that in the UN*X systems by actually sync'ing and unmounting the volumes myself before disconnecting). But, I think "removable (external) drives" probably see more use and would probably have more deliberate (or consequential) testing of the process, as a result.

Consider the times when Windows won't eject a device ("something is STILL using that device") and you're trying to hot-swap it... (?)

[eccerr0r]

Re: How many uf do i need to cope with the HDD inrush current

SATA (and of course SAS) are indeed capable of hot plugging. However the OS and the protocol may not. AHCI is required for hot plug as it has events that can be used by the OS to inform it that disks were removed, etc. If you're using legacy emulation, all bets are off as legacy does not support hot swap. You may be able to swap data plugs around so that the disks do not emulate a "master" and a "slave" to the same port - ensure both disks are masters to different "ports" - and your luck may be better, but it's best to use AHCI.

I would think that if you have a good/beefy PSU, hot plug inrush of one disk should not be a problem. It's tougher for the PSU on boot up *IF* it needs to deal with inrush of all of the disks at the same time. If there's a voltage drop during hotplug I'd look for bad connectors or bad PSU. Adding capacitors won't work, basically you'd need to supply 2A for 3 seconds or whatever spinup time is and keep it from dropping below min voltage - meaning you're looking at fairly large capacitors to make any difference. Without knowing the impedance of the whole path from PSU, or any calculations at all, I'd guess 10,000uF or so, at least, to handle most situations(!!!) - But this is not the right way to fix things. The PSU switching/low resistance wires/connectors are supposed to handle that droop, NOT the caps.

I have a 4-bay bare drive "dumb" SATA hotswap bay and the power connectors are a weak point, they need to be strengthened. In any case, I've been able to swap disks without the machine going down in Linux. It works well enough to be able to remove a defective disk, replace with a good disk, and rebuild my RAID array without stopping most processes. It's almost invisible to the users though I haven't automated it - I still need to partition the new disk and initiate rebuild. Pretty neat, though currently I'm only using 3 of the 4 bays. The bays are connected to 4 of the 5 onboard AHCI SATA ports (the fifth port is connected to another, empty hotswap bay ) and I'm using Linux software RAID5.

Yes, I need my caffeine, or rather need to make sure I'm not asleep doing this hot swap surgery; one false move minimally means a reboot and of course worst case is data loss if I pull the wrong drive or down the wrong disk. However at this point since I've done this dance a few times now, the larger risk is not tending to RAID disk loss in a timely manner - ignoring failure warnings is bad. Very bad.

[goontron]

Re: How many uf do i need to cope with the HDD inrush current

I think AHCI support is required to gain some of the more "advanced" features (hot plugging).

Which my homebrew server has, and hotswap enabled in the EFI. As for the poweredge 2900, IDK. Its a SAS controller. Its SATA compatibility is unkown ATM.

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

SATA (and of course SAS) are indeed capable of hot plugging. However the OS and the protocol may not.

Yes. And I am not confident in the sorts of testing that goes into the various implementations that you might come across "in the market".

I (almost!) had a complete data loss experience exactly once. I updated my OS (FreeBSD, at the time, IIRC) and mounted an external 4G SCSI drive (back when 4G drives were $1K -- if you bought *10*!). Jibberish!

<frown> "Well, that was a crappy $1K investment!"

Replaced drive with the cold spare that has identical contents (like half of a cold RAID1). ALSO jibberrish!

Hmmm... can't believe two drives that have been sitting on a shelf both decided to fail at the same time! Sure as hell am not going to try any of the other identical drives until I make sure there isn't something wonky with the cabling, enclosure, power supply, etc.

Nope. But, discover there's a problem with the updated driver (which doesn't affect the IDE boot drive)!

Roll back to previous version of OS. Drag out third copy of the data (on MO media) and manually restore first two drives using that.

I have a 4-bay bare drive "dumb" SATA hotswap bay and the power connectors are a weak point, they need to be strengthened.

Most folks think connectors "work forever". What they don't realize is how FEW insertion cycles many are actually rated! (Of course, a rating doesn't mean that the connector catastrophically FAILS when that limit is exceeded; but, all bets are off as to what you can reliably expect!).

For example, many DIMM connectors are rated for *6* cycles. SATA connectors for 50.

If the failing connector is part of a wiring harness, you're in luck: replace the wiring harness and you're starting with a fresh "quota" of insertion cycles! OTOH, if the connector is on the actual device...

Yes, I need my caffeine, or rather need to make sure I'm not asleep doing this hot swap surgery; one false move minimally means a reboot and of course worst case is data loss if I pull the wrong drive or down the wrong disk.

Ohnosecond: noun; the time between commanding your finger to strike the ENTER key and realizing you shouldn't have done that; usually a wee bit less than human reaction time...

[eccerr0r]

Re: How many uf do i need to cope with the HDD inrush current

Ohnosecond: noun; the time between commanding your finger to strike the ENTER key and realizing you shouldn't have done that; usually a wee bit less than human reaction time...

And I believe there's an article posted recently in some medical journal that this is not just a silly anecdote. It's a real phenomenon with human anatomy where once your mind is made up the first time to make a muscle movement, you can't stop it no matter how hard you try even if it hasn't quite been done yet - only recover if it's possible. Human anatomy "queuing delay" measured to be a bit less than 1 second or so

So keep away from that ENTER key before you double check!

[Curious.George]

Re: How many uf do i need to cope with the HDD inrush current

And I believe there's an article posted recently in some medical journal that this is not just a silly anecdote. It's a real phenomenon with human anatomy where once your mind is made up the first time to make a muscle movement, you can't stop it no matter how hard you try even if it hasn't quite been done yet - only recover if it's possible. Human anatomy "queuing delay" measured to be a bit less than 1 second or so

Wasn;t intended to be silly! Rather, an acknowledgement of a very real experience I suspect most of us have had -- we KNOW our finger is on the downstroke (or whatever), realize we REALLY don't want to do THAT -- but can't do anything to stop it from happening.

IF you could speak in that interval, you'd be saying "OhNo...", in much the same way as Mr Bill.

[Stefan Payne]

Re: How many uf do i need to cope with the HDD inrush current

my PSU is a Seasonic M12II 520W

That could be the problem.
That one is ancient (since 2010), has low capacitance on +12V and is group regulated.

That could cause an issue because the +12V Rail drops and possibly also the +5V because of the increased load of +12V...

Do you have an independently regulated PSU lying around you could try?

[evilkitty]

Re: How many uf do i need to cope with the HDD inrush current

I never had a issue when booting off the internal drives that are on a different wire, then again i do not use windows on those drives so it could be idle and not care

My only spare PSU is a the same unit except it is not modular (purchased when on sale couple months back)
my modular one was purchased on *checks records* 7/26/2012 aka 2012/07/26

I went to use one of my bays the other week and it did not work, gonna rebuild that cable this weekend (hopefully before my pi 3 b+ arrives)

[evilkitty]

Re: How many uf do i need to cope with the HDD inrush current

I remade my cable using better quality wire (without adding caps to the wire)
this alone did not fix the issue
added a drive during a benchmark:



After adding 470uf caps on +5 and +12 (Low ESR Panasonic, rated at 25v)



it did not help,
When i connected the caps it caused my main HDDs to spin down

* The large drop in the read speed is when i connect a second HDD in my second bay

[eccerr0r]

Re: How many uf do i need to cope with the HDD inrush current

What you should be doing is datalogging the voltage on the lines... or finding some utility that will generate a graph from the on-board motherboard sensors if you have them (though it will not take into account any drop caused by bad wire/connectors to the hdds...)

Either way you still would need to add the caps as close as possible to the hotplug port, and hope the connector isn't actually shorting something it shouldn't.