The Treadboard: ATX power control for JBODs and Expanders

T

treadstone

Gawd
Joined
Jan 25, 2010
Messages
567
I'd like to ask those that have, use or intend on buying a Norco 4220, 4222 or 4224 case or any other type of external storage enclosure for some feedback. What I am looking for is some sort of 'wish-list' of features you would like to see in a power control adapter board. I am designing this adapter board for use with the HP SAS Expander, however the functionality will not be limited to to this expander alone.

The functions of this adapter so far:
  1. Single PCIe x8 slot to power the HP SAS Expander
  2. A 24-pin ATX12V 2.x compliant power supply connector
  3. A 2 pin connector/header along with a circuitry for the chassis power push button to control the ATX power on/off signal (PS_ON#)
  4. A 2 pin connector/header for the chassis power LED as status feedback
  5. Circuitry to buffer the ATX power supply PWR_OK signal for use as the PWRGD signal on the PCIe connector

Suggestions I've received so far:
  1. External +5V trigger input for remote power on/off control (requested by PigLover)

Since I don't own one of these Norco cases (yet), I would like to know how or where you guys would prefer to mount this adapter. Pictures of an empty case (top down view) would also be highly appreciated so that I can clearly see the location of any standoffs that could potentially be used for mounting this adapter.

Once the 'wish-list' is pretty much done, I'll design a PCB for it and get a few boards done...
I would also appreciate feedback on who would be interested in such an adapter.

So fire away guys...

P.S.: I would like to thank odditory for his awesome thread title suggestion!
 
Last edited:
If your target market is inside a Norco case, chances are there is already a MB in there, but I bet the PS has an unused video card PCIe 12v plug with lots of switched power. A few rectifiers would yeald +- 3.3v but I bet the HP SAS Expander doesn't even need that. (tape will tell, a job for odditory :)

keep it simple.
 
I think it would be ideal if you could mount it using the ATX standoffs that are in line wiht expansion slot 7.
This way it could also be used independently even if there is also a mATX board installed.

So you could use the drives for storage expander from another case while still utilizing this case to house another server simultaneously.
 
@picker: I think you are missing the point here. The idea is to use one of the above mentioned Norco cases as a drive expander by utilizing one of the HP SAS Expander cards and hook it up to a separate server without having to put a motherboard inside of this drive expander. So this adapter would require the necessary circuitry to convert the power on/off push button into a signal that can control the ATX power supply's PS_ON# line.

@nitrobass24: The last slot (slot 7) is what I had in mind too, hence my request for some decent top down pictures of the Norco enclosures.
 
I'd be very interested in a few of these.

But..

Aren't Norco already working on a cheap SAS expander based on this chassis series?

Edit: Also, the HP SAS Expander is PCI-E x8, not x4.
 
Last edited:
newegg has good pictures from the top down on all the models.
 
Treadstone, Here are pictures of the PCMIG version next to the Supermicro X8SIL-F (mATX) in a RPC-4220.

Let me know if you want me to take more pictures with the mATX board in there.

The big question is whether you want power-on to be dependent on another enclosure because if you lose the PSU in the JBOD enclosure, and you span RAID arrays across enclosures, you can be in bad shape. I just don't do this, but I could imagine someone spanning two enclosures because they want to try a 48 drive RAID Z3 config.

Another thought I had (may be ancillary) is can we make custom stand-offs to add another 2-4 drives, more 2.5" drives, and etc. I've probably spent a few hours doing cardboard mock-ups because there is a ton of space in the 4U's not being occupied.
 
@pjkenned: Thanks for the link, I did see those pictures when you first posted them :)
How did you mount that PCMIG board to the enclosure?

The remote trigger input that PigLover suggested could be used to control the power supply across multiple enclosures. I personally would not span a single RAID across multiple enclosures either. Just too many things that could go wrong and bring your entire RAID system down!

Mounting additional HDD inside an otherwise empty enclosure (since there is no motherboard in a drive expander), is definitely doable. There are quite a few different ways you could go about it, one and probably the easiest is to make a flat sheet of metal with mounting holes that correspond to the spare motherboard standoffs. This piece of metal would then have the mounting holes for the drives... This however takes us off topic here :)
 
Mounting ideas:

  • Perhaps similar to the Chenbro AEK mount, place it flat in MB space and use the IO shield space for the external inputs. Pretty cool because you can use standard mount hole spacing (mITX spacing) and it would work in any ATX-standard case, even the ones that don't include slot 7, which is the only one that lines up with standard mounting holes. Slot 7 might also be a tricky place for cooling.
  • A nitrobase suggests, and pjkenned demonstrates, mounting in line with slot 7 does let you use the MB space for a small server. As pjkenned did, you can use something like a pico-PSU for that extra server, which is genius (well done).

I predict you'd have buyers for either arrangement.
 
@PigLover: I think Slots 1, 3 and 7 are the most likely candidates for mounting this adapter. I still have to confirm that with the standards. The I/O shield area would only work for the adapter if you don't plan on using an expander (such as the HP SAS expander) in the PCIe connector. This might come in handy for other implementations such as those that use SATA port multipliers.
 
I'm thinking I might also throw in a fan controller into this design that can control fan speeds based on internal temperature of the enclosure. This would work with 4-wire (PWM controllable) fans. I may go as far as making it USB controllable so you can get some feedback and or control the fans from the server...

Have to think about this :)
 
Great idea... fan controllers in the Norco RPC-4220 and RPC-4020 are awesome if you dial the fans back to like 80-85% because they lose an annoying tone or two.

@ PigLover: #1 anyone that has me an genius in the same sentence gets a :) #2 I tried slot 8 but the cooling there sucks. With slot 8 being one of those semi-vented expansion slot covers, it works fine from a cooling perspective having the HP SAS Expander in slot 7.

I also mocked up a mount for the upper left portion of the case, a 90 degree mount bracket that would mount it and allow a full motherboard like a X8ST3-F underneath with LP cards with a full height bracket. The problem was that the rear SFF-8088 port is a pain unless you put it in an expansion slot so that made me move to slot 7 or 8 still allowing mATX or smaller motherboards.

@ treadstone: the mounting is a PITA because of all the backside components. Ended up using mounting holes but ripping off a bunch of backside garbage on the PCMIG due to the motherboard standoff tray in the Norco cases (my RPC-470 does not have this however).

At first I didn't like the idea of the ATX connector going to the HP SAS Expander PCMIG board, but doing so (plus the on-PCB header for the power connector) allows me to control power via the front on/ off switch. If you did not use the ATX connector, I bet a lot of people would put a mATX board in there an accidentally power off the mATX server, and the rest of the enclosure with it. Tried this, and realized that would do bad things to have 20+ drives go offline if the other server power cycled.
 
treadstone said:
I'm thinking I might also throw in a fan controller into this design that can control fan speeds based on internal temperature of the enclosure. This would work with 4-wire (PWM controllable) fans. I may go as far as making it USB controllable so you can get some feedback and or control the fans from the server...

Have to think about this :)
Click to expand...

I think you could sell something like this by itself :)
 
Since I am not particularly technical I'll put on my BA hat and provide my "gold" requirements in business terms and let you guys figure out if its technically possible of a PICMG (ish!) board based solution :) I realise that a lot are probably not feasilble and more so probably stupid but might give some ideas!

"Mandatory"
Provide sufficient power for 1 x SAS Expander (prime purpose...)
Power on / off via chassis button

"Nice to Have"

Ability to power multiple enclosures up simultaneously (its a pain in the arse if I sit down to watch a movie and for example I forget to switch on one of my slave chassis. It can really confuse media browser!)

Ability to power up "slave" enclosures on power up of chassis holding the actual computer - Getting lazy now...

Ability to delay power up of actual computer powering up to allow other chassis to come online - I do like to make sure the chassis are up with all disks spinning before the RAID card kicks in, though I know the chances of this not being true are probably zero...

Only power on actual computer if all connected chassis come online - absolute nirvana, absolute laziness. :)

Switch off all enclosures when host computer powers off - I don't leave my disks spinning, would cost the Earth, but then my RAID card doesn't have MAID!
 
treadstone said:
I'm thinking I might also throw in a fan controller into this design that can control fan speeds based on internal temperature of the enclosure. This would work with 4-wire (PWM controllable) fans. I may go as far as making it USB controllable so you can get some feedback and or control the fans from the server...

Have to think about this :)
Click to expand...

Sounds useful. Will 2 wire SM bus travel any distance? (Quite a few boards have an SMBus header and I'm pretty sure most monitoring / fan control chips support it)


Does anyone know what kind of power on control tends to work with external SAS multilane? Is there a universal way that the raid card on the host could start the chassis?
 
Sorry for the delay, I am extremely busy at work this week (New customer, BIG project, trial installations, etc...)

@pjkenned: I think I will definitely include a fan controller. Sort of makes sense :)
I figured you had to mess around with that PCMIG board due to the double sided component issue and hence was wondering how you managed to mount that to the standard motherboard standoffs.

@MrD1234: I think I will implement the fan controller based on my design idea for a distributed fan controller for my server (see my My 100TB Media Server Build Log for more details...)

@haileris: The control circuitry on the adapter should be able to handle most if not all of those scenarios you listed, except maybe the scenario where you want the expander enclosure powered up first and then have the main computer slaved to it. The idea here is to have the expander as a slave to the main computer/server. I might even go as far as make the delay times, etc programmable...

@Krobar: SMBus is a form of I2C (at least it's mostly based on I2C with a few differences). I use I2C extensively in my designs and was planning on using it in this design as well. The distances within a chassis are no problem for I2C but it was never really designed or intended to be used as an interface between enclosures. Although there are a few examples where I2C is in use in such cases (e.g. between the graphics card and the monitor via a VGA or DVI cable, between HDMI capable devices, etc.)
I've looked at a LOT of fan controller chips and have yet to find one that fits my needs or would handle the fan control exactly the way I want. Some of the work I am involved in at the moment actually has to do with fan control and hence it is along the lines of what I am working on right now anyway (except on a far larger scale than small computer enclosures!). I can take some of those algorithms and stick them into a smaller controller and put it on this adapter board :)
There are a few different ways power (or environmental control for that matter) can or is handled via the SAS type interfaces. On SFF-8087 type interconnects (internal to enclosures), there are dedicated pins that are or can be used for this kind of communication (SGPIO and/or I2C depending on the controller manufacturer, however SGPIO is the more common standard). On SFF-8088 type interconnects (used to interconnect enclosures), no such pins exist and hence all of the communication for environmental or enclosure management is handled in-band and requires a processor to get access to this data. The processor on most expanders will usually do this.
 
treadstone said:
Sorry for the delay, I am extremely busy at work this week (New customer, BIG project, trial installations, etc...)

@haileris: The control circuitry on the adapter should be able to handle most if not all of those scenarios you listed, except maybe the scenario where you want the expander enclosure powered up first and then have the main computer slaved to it. The idea here is to have the expander as a slave to the main computer/server. I might even go as far as make the delay times, etc programmable...

@Krobar: SMBus is a form of I2C (at least it's mostly based on I2C with a few differences). I use I2C extensively in my designs and was planning on using it in this design as well. The distances within a chassis are no problem for I2C but it was never really designed or intended to be used as an interface between enclosures. Although there are a few examples where I2C is in use in such cases (e.g. between the graphics card and the monitor via a VGA or DVI cable, between HDMI capable devices, etc.)
I've looked at a LOT of fan controller chips and have yet to find one that fits my needs or would handle the fan control exactly the way I want. Some of the work I am involved in at the moment actually has to do with fan control and hence it is along the lines of what I am working on right now anyway (except on a far larger scale than small computer enclosures!). I can take some of those algorithms and stick them into a smaller controller and put it on this adapter board :)
There are a few different ways power (or environmental control for that matter) can or is handled via the SAS type interfaces. On SFF-8087 type interconnects (internal to enclosures), there are dedicated pins that are or can be used for this kind of communication (SGPIO and/or I2C depending on the controller manufacturer, however SGPIO is the more common standard). On SFF-8088 type interconnects (used to interconnect enclosures), no such pins exist and hence all of the communication for environmental or enclosure management is handled in-band and requires a processor to get access to this data. The processor on most expanders will usually do this.
Click to expand...

Hi Treadstone,


Thank you for clearing things up. I suppose the hit and miss presence of a 2 Pin SMBus connector on motherboards is the main issue with SMBus connectivity. I only mentioned it because it would mean software like Speedfan could control the fans speeds for you without any need for a software update if a supported chip is used on the board. I haven't checked but is there a connector to management bus on the Norco backpane, Am I right in saying the Norco is SGPIO? (Our Tyan barebones at work use this and they seem to have more indications and are more informative than my 4220)

Looking at all this I think what is needed is a combined Expander, Power control and Fan control board in an ATX compatible format. A lot of work I'm sure but the market for this sort of thing extends well beyond Norco as there are plenty of data storage type cases where there is no expander option or the expander option is overpriced or incompatible.
 
treadstone said:
Sorry for the delay, I am extremely busy at work this week (New customer, BIG project, trial installations, etc...)

@haileris: The control circuitry on the adapter should be able to handle most if not all of those scenarios you listed, except maybe the scenario where you want the expander enclosure powered up first and then have the main computer slaved to it. The idea here is to have the expander as a slave to the main computer/server. I might even go as far as make the delay times, etc programmable...
Click to expand...

Thanks, sounds good. I was thinking that if you hit the power switch on the main computer you could essentially delay the startup of that computer until all the drive chassis are on. But you are correct of course the logic of that would be driven at the computer chassis as opposed to the expander chassis.
Cheers
 
haileris said:
Thanks, sounds good. I was thinking that if you hit the power switch on the main computer you could essentially delay the startup of that computer until all the drive chassis are on. But you are correct of course the logic of that would be driven at the computer chassis as opposed to the expander chassis.
Cheers
Click to expand...

Can't you can configure the drives to start up in the spin down condition, and use the RAID card to control the spin up of the array?

Assuming this function still works through a SAS expander.
 
MrD1234 said:
Can't you can configure the drives to start up in the spin down condition, and use the RAID card to control the spin up of the array?
Click to expand...

My current RAID controller doesn't have a spin down or staggered start options.... bought a Areca so I guess I will have soon.
 
@brian89gp:

There are tons of those types of adapters around. However none of them seem to control the PWRGD signal and/or have the right mounting holes or control circuitry to control the power on/off for an ATX type power supply. Hence this thread...
 
On another note. I did start the schematic for this adapter, but unfortunately I was way to busy with my work to be able to finish it. Hopefully I'll find some time this weekend to complete it.
 
Could we power this board using a 4-pin molex? Seeing we only have 1 ATX-connector on a PSU...
 
Jeroen1000 said:
Could we power this board using a 4-pin molex? Seeing we only have 1 ATX-connector on a PSU...
Click to expand...

One of the key points of this board, IMO, was to be able to use an expander in a chassis without a motherboard. Using the ATX connector is the only way to activate the PSU.

Though perhaps Treadstone could add a 4-pin molex such that the board could be used in a chassis with a motherboard and not have to split the ATX connector. This would of course bypass the power control functions of the Treadboard.
 
I think it would be a good idea. The HP-expander is said to be picky when it comes to motherboards. If this board can be powered as such, no one would need to worry about motherboard compatibility.
 
epimetheus said:
One of the key points of this board, IMO, was to be able to use an expander in a chassis without a motherboard. Using the ATX connector is the only way to activate the PSU.

Though perhaps Treadstone could add a 4-pin molex such that the board could be used in a chassis with a motherboard and not have to split the ATX connector. This would of course bypass the power control functions of the Treadboard.
Click to expand...

Having a motherboard in there means one of two things:
1. It is being used as a dumb power board. That was my build your own JBOD SAS/ SATA enclosure iteration 1. I did find some boards with some issues powering the HP SAS expander.
2. It is being used as a system board and there is another board like the PCMIG or treadboard powering the SAS expander. In this case, you wouldn't want a power cycle of the system board to power cycle your storage which may be in the process of having data written to it from the system writing to the drives. This was my build your own JBOD SAS/ SATA enclosure iteration 2. Solution: PicoPSU to power the mATX board.

Just some thoughts.
 
All valid points, especially the two points pjkenned pointed out!

The problem with the 4 pin Molex power connector is that it only carries +5.0V and +12.0V. As far as I know (I have to verify this tonight on one of my HP SAS expanders), the HP SAS expander requires the +3.3V. But even if the HP SAS expander does not require the +3.3V (they may have their own regulator on board), I want to make this board somewhat universal to power other boards not just the HP SAS expander, which means I have to supply the +3.3V to the PCIe slot. Putting a buck converter on the board is not a problem from a design standpoint of view, but it adds additional cost that I don't think can be justified in most cases!
Especially if you want a low cost solution, the parts for this regulator would just add unnecessary complexity and cost. The only connector that does supply all the necessary voltages is the ATX 20/24pin connector. There might be one other solution and that would be to use a SATA power connector that is supposed to deliver +3.3V, +5.0V and +12.0V but unfortunately lacks the power ok and power control signals...

On the other hand I could design the board to also have the option of being powered via a 4 pin molex power connector and include a buck converter, but not populate those parts for the standard version and have an alternate version that can be powered via the 4 pin molex connector. The drawback here is that I would have to get two different boards assembled and that also adds cost on my end :(

I have to think about that, also feedback from you guys would help to let me know who prefers what...
 
Personally I would buy one of these in a heartbeat if it was molex powered. I doubt I will ever have a case dedicated for nothing but storage. If this could make the HP expander work with any motherboard though, that would be very useful. Honestly I'm hesitant to buy one of the HP expanders because I don't know if it'll work with my motherboard. If it doesn't I'm stuck with a $250 paperweight.
 
I'd be useless as a paperweight, it's not that heavy ;)

I understand your concerns, as I said, I'll have a look and might add the molex connector as an option to the design. My original idea was to design the board just for use in standalone storage expansion systems due to the incompatibility of the HP SAS expander with so many motherboards. But seeing as there potentially other uses for this board, I think I might add some more options to it and just not populate them depending on what the target application is...
 
Perhaps add a SATA/molex power jack and provide a switch to enable power OK and the power control signals? Essentially, some way of switching a mode to turn the board to a dumb pcb that has a pci-e slot with power.

I'm assuming that the power ok and power control signals are simple on/off signals with no logic involved...
 
S-F said:
Personally I would buy one of these in a heartbeat if it was molex powered. I doubt I will ever have a case dedicated for nothing but storage. If this could make the HP expander work with any motherboard though, that would be very useful. Honestly I'm hesitant to buy one of the HP expanders because I don't know if it'll work with my motherboard. If it doesn't I'm stuck with a $250 paperweight.
Click to expand...

LGA 1156 and LGA 1366 server motherboards (Intel and Supermicro) I'm 5/5 with it working. Cheap AMD board based on NVIDIA chipset 0/1. Consumer class AM3/ LGA 1156 motherboards I think I'm something like 5/7 with.

Worst comes to worst you buy a PCMIG or treadboard and viola, you have a working $250 card.
 
I'll add a working board to the list soon, just got to do some led checking:). Too bad I'm still short on time:mad:. For those who do not use server (mother)boards this board could prove utmost useful.

BTW, it is very cool you can make this stuff!
 
On the other hand I could design the board to also have the option of being powered via a 4 pin molex power connector and include a buck converter, but not populate those parts for the standard version and have an alternate version that can be powered via the 4 pin molex connector. The drawback here is that I would have to get two different boards assembled and that also adds cost on my end

I have to think about that, also feedback from you guys would help to let me know who prefers what...
Click to expand...

20/24pin power connector gets my vote. I would definitely want one of these when I add another chassis to my current 20 bay.
 
@The_Frag_Man: I honestly don't know yet. I have to have a look at the BOM when it's done and see how large the final PCB will be and how much the board shop is going to ask for them... I am trying to keep the cost as low as possible but at the same time put as many features and functions into this design as possible.

@Krobar: The board currently has a 24pin ATX power connector on it which should be compatible with both the older 20pin and the newer 24pin power connectors.
 
Is there a timeframe when these are projected for availability?
 
Last edited:
You must log in or register to reply here.
Back
Top