OWC Mercury Helios 3S Thunderbolt PCIe enclosure – teardown, screws and a bit of camera comparison (WHL #123)
Back to regular tech – and what better product to start with than the OWC Helios (3S in that case, but the 3 and 5 should be pretty similar). These are single-PCIe x4 electrical / x16 mechanical, double-slot to Thunderbolt 3 (or 5 in the Helios 5…) enclosures, primarily aimed at the more tech-y end of the Apple crowd, but universally usable for all TB-capable devices.
I already owned one of these, but I also wanted to get rid of an entire server enclosure. Putting a SAS card into them (one of the smallest cases on the market) is the most straightforward way to do just that. And since my old case is already occupied with an external SAS card…I needed another one. Thankfully, despite them being pretty rare and therefore not often sold on the used market, one was available for pickup in Hamburg for a few weeks on a classified ads page. Hamburg is a 1000km round trip away from me, but *checks map*, lies DIRECTLY on my way from home to my first ferry port on the island of Fehmarn. Well – I asked the guy a couple days before, he put it aside, and I picked it up, no problems. (Sony ZV-1 sellers however, UGH…). I brought a simple, small serial/parallel card with me for testing that I didn’t check beforehand…and the case only initialized half of the TB structures and the card didn’t show up. A small 5000km detour later I checked it with my unit, and it also didn’t initialize, so that was a dumb idea. But it also didn’t work with a known-good SAS card, which of course caused me to tear the thing open. After all, I own an identical unit that is known to work – why not swap boards and wiring and find out what’s wrong with it? I have to add here that the seller claimed the enclosure didn’t work with his intended Blackmagic audio woo-woo card either – but that one is also known to not work with certain TB cases, Apple BS. So no major red flag here either.
(press photo, couldn’t be arsed to photograph my unit…)
In the end, I tested every piece and isolated the main board as the faulty piece. PCIe x16 board works, internal power cable works, TB cable works, PSU and AC cable works. So I put it back together for one last test before throwing it into an ESD-safe bag for an inspection at work…aaand suddenly everything fired right up. Contact issue, likely on one of the PCIe sockets. Just a simple re-seat did the job. What a boring result.
We do have some photos, though – these are comparison shots of my old Sony RX100M2 (which is currently up for auction on eBay) and the Sony Alpha 6100 that I bought as a photo (as in “not-video”) camera replacement, previously discussed in WHL #120. It’s equipped with the kit glass, a SELP1650 (E PZ 16-50mm f3.5-5.6 OSS), so this is probably the baseline of what the camera can do. After all, everything is bigger and more betterer™ in an APS-C camera compared to a compact one, right? The lens isn’t really all that great for macro shots as I soon found out…
All internal boards and wiring (except for status LED and fan) connected – RX100M2 first, α6100 second
Super simple construction, and although the PCIe connector should be able to deliver 75W of power, they opted for a separate power connector. The power brick is rated 90W, but that also supplies power to the USB-C downstream port and all the TB components on the main boards. Yeah, technically a x4 connector can only do 25W and the x16 can do 75W, but the power pins are identical on both, it’s the part before the notch.
Same thing again, just oriented like in the enclosure. In there, a metal plate sits over the main board for both mechanical protection and likely shielding. The x16 board is fixed with three out of the four screws, as one cannot be reached with the lower board installed. (RX100M2 / α6100 again)
Since the x16 PCB is purely mechanical, here’s two more shots of the Thunderbolt board – top side, main ICs being an Intel X215TD11 JHL7440 (TB3 controller), TI TPS65983BA 28CL6QW (2x, USB-C PD controller) MPS AGQP700 (4x, five-port 2.8-5.5V PMIC)
Back side – not that much more than a Winbond 25Q80DVNIG 8Mbit SPI flash chip and a couple of weirdos in tiny strange cases:
I have to say that I generally prefer the photos of the RX100, but the Alpha 6100 is on an important disadvantage here: It was set to f/10 (or f/11), while the RX was in the f/3.5 to f/4 region. I typically operated the RX100 much higher (f/8-ish) in my light box, but this was recently reset and I didn’t notice, and the difference between those numbers is like a 6-fold reduction in photon count or 2.5 stops. That’s likely much more than the bigger physical lens of the Alpha can compensate, so we got longer exposure times instead, resulting in more blurry images (hand-held shooting!). The theoretical better depth of field only shines in one of the images, and it doesn’t really matter on the flat PCBs. So not a fair apples-to-apples comparison, but just what silly me as a photo amateur had hoped would be somewhat comparable. Young padawan me still got a lot to learn about using a more professional camera to his advantage.
All back together by the way looks like this – I clearly have no addiction problems with OWC cases, as I now own more than I can daisy-chain on my single TB port… (not all of these are in use, though – it’s 2 pieces Helios 3, plus 4 units 4M2)
While disassembling the units, I had a hard time with screws. The outer ones, black anodized Torx, are easy to deal with, these aren’t even security Torx with the center pin. But the shiny ones inside..WTF are these?
(Super difficult to get right with the Alpha6100, I should probably look for a macro lens right away!)
Close-up shot from my TOMLOV DM9 microscope / WHL #100 – background grid is millimeters.
Let me repeat myself: WHAT THE FUCK IS THAT? Bad quality Torx Plus? None of my Torx drivers of similar sizes fit, some T6 does finally sit inside but has minor friction points. Not because the shape is that much different, but because the visible “large” part of the drive is very shallow, and the T6-sized one, well, is T6-sized. T6 has 1.7mm of point distance and a torque rating of like 0.8Nm, that’s not exactly the most manly driver bit out there. Thankfully, none of these screws were absurdly tight or used any kind of thread lock.
I ended up removing all of them with a slotted screwdriver that happened to be the correct size across the lobes, and I threw them away and replaced with regular Torx heads (I think TX10 size). Does selling to Apple customer rub off some of the Pentalobe insanity that Apple has in terms of screws? Jesus…
Feeding measurements from the image into some AI yields Torx Plus 9 size for the outer part, which allegedly has 2.48mm of diameter and my measurements being 2.55m (without parallax correction and stuff). I’m still weirded out by that two-step profile (regardless if intentional or QA distaster), that’s so fucked up man.
Anyway, all in all a nice catch, an important addition to my collection of Thunderbolt devices, a worthy opponent on disassembly, and also a good target for the camera comparison. Would I buy it again? Absolutely. Would I bring a working PCIe card and get a much better price for the faulty equipment? Also absolutely. But I never thought it was defective in the first place, so that worked out just fine, and I’m always happy to pay a fair price for good hardware (which it undoubtedly is, once reseated…)
Now is probably the time to build a Y cable for the silly PCIe-style power connector, because even two cards combined with all the TB gear in place draw nothing even close to 90W of power – and the Y splitter cable for my two combined 4M2 units also worked out fine. Or should I also put both assemblies in a custom 3D printed chassis that isn’t this heavy and bulky? TB² (Thunderbolt-Terabyte) SAS arrays, here I come!
