Making SATA Drives Work With A SCSI Backplane

The problem of persistent and reliable storage plagues us all. There are a myriad of solutions, some more expensive than others, but a dedicated and redundant network attached storage solution is hands down the best choice for all problems except natural disaster (ie: fire, flood, locusts) and physical theft. That being said, the issue of price-tag rears its ugly head if you try to traverse this route.

[Phil’s] had his mind stuck on a very large NAS solution for the last ten years and finally found an economical option. He picked up a powerful motherboard being sold as surplus and a server enclosure that would play nicely with it. It came with a backplane for multiple hard drives that utilized SCSI connections. The cost and availability of these drives can’t compare to the SATA drives that are on the market. Realizing this, [Phil] completely reworked the backplane to make SATA connections possible. It’s an intense amount of work, but there’s also an intense amount of documentation of the process (thank you!). If doing this again his number one tip would be to buy a rework station to make it easier to depopulate the connectors and extraneous parts from the PCB. Since he needs to keep using the board, the old blow-torch trick is out of the question.

24 thoughts on “Making SATA Drives Work With A SCSI Backplane

  1. I misunderstood the HAD title and description. This guy is not making SATA drives work with a SCSI controller, just cutting holes in the SCSI board to route the SATA cables to a real SATA controller.

    The SCSI backplane is just used to hold the SATA connectors he glues to them.

    Still nice work though.

    1. Agreed. I think a better title would have been along the lines of “modifying a SCSI backplate for hot-swapping of newer SATA drives”.

      I thought the same thing, that he was somehow making a scsi controller communicate with SATA drives… I knew it was too good to be true!

  2. Is it me, or would etching a new board be much simpler ? Cut out a board of roughly those dimensions and very roughly trace the leads on that.
    Glue the SATA connectors to the other side and simply bridge solder them.
    That would make the layout much more tidy and uniform. It would probably be a lot quicker also.

    1. Off the top of my head, I get the impression that it may be difficult to put something like that together without getting cross-talk and other interference issues at the rates SATA runs at.

      Though I could be completely wrong on that.

    2. Even cheaper, would be to use a thin piece of wood or plastic and just cut holes in them to stick the relevant connectors.

      Making a pcb for data traces requires some care. The traces need to be a certain distance from each other and the length needs to be roughly the same. Sata uses LVDS (and I don’t think sata cables are shielded), so the signal should survive alright but one still needs to be careful.

      1. Actually, a differential signal at 3MBit/S is using super fast edge rates, any impedance mismatch will have a very bad effect on signal EYE quality. Making a PCB for this would that require the single ended and differential impedances are maintained within SATA specs to ensure that such a project worked right.

        Anyway, HAD needs to clarify that this guy did NOTHING but make SATA drives fit in an old SCSI housing – And at the end of the project, he shows no results but a PCB with connectors glued to it.

        So dissapointing :(

    3. Event simpler, buy protoboard, the One with the holes in it and copper on both sides. And buy 2 SATA connectors per hdd. Print Version. Put One on each side of the board and solder them directly to each other. Like a sandwitch. Connector, board, connector. Dont forget ti turn 1 connector 180 deg to maintain 1:1 Pin mapping.

  3. Isn’t JB Weld a steel reinforced epoxy, and therefore um….. electrically conductive?

    Also – I don’t exactly understand the point of this. He wanted a large NAS. Aside from just buying a few SATA cards and doing software raid (trivially easy) he instead bought a XEON workstation to do hardware RAID…. seems overkill for storing a few seldom accessed files on a LAN but ok whatever floats your boat…..

    But why didn’t he just buy a hardware PCI-E RAID card and plug his drives in and call it a day? Why go to all this extra effort for no gains and a lot of extra work?

    1. I thought so, too.

      But try though I might, I don’t find any meaningful conductivity in JB Weld.

      I also don’t find any meaningful conductivity in Arctic Silver thermal paste (which is just pure silver suspended in goo).

      Either one is easy enough to test: Just plop your meter probes into a glob of it and see.

      It seems that 2-part epoxy and/or silicone oil are both very good insulators, and that any conductive metallic “filling” is not so conductive at all once it forms a compound with the rest of the stuff that makes up these products.

      YMMV.

      1. What voltage were you testing them at though? At higher voltages they may show some conductivity, we have a glue in the UK called UHU which i have found to be conductive even at low voltages, luckily i checked this before using it in my project.

  4. Not to rain on your parade (it’s a cool project), but for real homebrew applications it might be more sensible to do this.

    http://blog.backblaze.com/2011/07/20/petabytes-on-a-budget-v2-0revealing-more-secrets/

    I wonder if someone (myself??) could be bothered to port the case design from Protocase to Ponoko, fitting a straightforward steel box.

    Or perhaps there is a company which can do the same job as Protocase, but with the assembly left to the builder? It’s just a waterjet/laserjet, after all.

  5. I have to agree that this article is very misleading. I thought this guy had worked out the solution to my woes, as I have 4 PE2650’s that I’d love to put some IDE HDDs in rather than SCSI. He did however do an awesome job!

    1. new dell poweredge servers will nearly all be bought perc raid controllers.. these generally all are SAS and will take SATA drive but for production machine i will always use SCSI.
      You can buy cheap second hand SAS RAID controllers and then use SATA drive.. basically forget about IDE as these are too slow compared too newer SATA (hey 80-100 bucks for 2TB drives)

  6. Wouldn’t it have been cheaper to pick up motherboard and CPU and throw together a proper server without all the JB weld?

    My home server is a Pentium 4 with 5TB worth of hard drives (2.5TB usable, mirrored RAID). Aside from the hard drives the system cost me next to nothing to put together and it is capable of transferring files onto the network from hard drive at 45 MB/s. I lost a hard drive about a year ago and replaced it. The RAID synced and all was well.

    I really wouldn’t want this mess of glue in something that I trust my data to.

    1. This definitely falls in the category “because one can”. That being said cabinets that sport these types of breakout boards tend to be very well build iron with tons of options to add and expand.

      Hacking a proper server housing to do you bidding in that sense probably is worth it.

      As for trusting ones data with whatever hardware, I trust mine (and his) soldering ( and glue ) skills to the extent that I prefer that over a Pentium 4 for a server any day.

  7. A great solution to a problem that I gave up on.

    I have often wondered whether some hardware vendors (HP, Dell etc) had replacement backplanes.

    Sadly, mose servers don’e allow their front pannel to operate without the backplane being installed :-(.

  8. Heh,

    I take it the reason for this is to get the satisfying feel of hot-plugging your drives into the front of a nice controller.

    For me though, having experience of failures in 2 different consumer RAID set-ups, I really would only consider 2 controller set-ups:

    1. Software RAID. We have ample processing power to spare for this today.

    2. Dual redundant controllers using SAS drives or similar (these having 2 ports each).

    The complexity and single point of failure that fakeRAID or even fully fledged but singular RAID controllers bring to the party simply aren’t worth it.

  9. I totally appreciate the “look at this amazing thing I did” and am in awe of the work put into this project. I’m interested to know how that 3ware card will work out. I’ve been down this same road, but never comitted myself to a project quite like this. Unfortunately I take a very business-orientated view on these sorts of projects, since I’m trying to build a cost-effective SAN for work. ROI, ongoing maintenance, expandability, etc. all factor into my calculations. While good business, not fun at all. :-)
    It was unfortunate he didn’t get himself an Intel SR2400 chassis. Effectively the same chassis, but can use a SATA backplane. I’m working on an upgrade project for an existing server I have, which runs the Intel SR1400, the 1U version of the SR2400, and amazingly got the same RAID card as this person did. I’d like to see how it goes in the future.

Leave a Reply to PouncerCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.