Re: Hacking my old eSATA Dock; and why will i (maybe not) fail to bypass the IC?

[faekjarz]

Hey there!

I have an ole Sharkoon SATA QUICKPORT DUO II which is in perfect working order, but its limits are USB 2.0 and SATA I / 1.5 Gbps.



Now i want to hack it, like this 



…the idea is to bypass (cut traces) the slow IC and directly connect the SATA connectors to the eSATA sockets.

Here are some money shots of its nice Rubycon caps




My plan A is to implement this bypass as a "Turbo" switch which energizes some relays that directly connect the storage device through to the eSATA port and let the computer main board negotiate transfer rates. Well, those negotiations will hopefully result in SATA III / 6 Gbps. 

I'm a little worried about the signal integrity, because I’ve found some craycray expensive high frequency relays (those) on Mouser. Do i really need those? The English Wikipedia article on SATA doesn't state it, the German one does; apparently the frequency of the SATA bus correlates with transfer rates, 1.5 Gbps = 1.5 GHz.

How about a trial and error approach? Just use cheap relays and see what happens - if the "Turbo" doesn't work, well than just don't switch it on.

Will i fail?
What about solid state relays?
Are there other types of switching mechanisms i haven't thought of?

[MosherIV]

Will i fail?

Probably.

SATA, eSATA and USB are all SERDES interfaces. To achieve the high serial data rates, they use transmission line networks - that is the impedance and line terminations are carefully controlled.

Randomly adding bits of wire and solid state relays will change the line impedance and termination way out of specification.

[wraper]

Almost guaranteed that relays will screw up signal integrity enough to cause SMART UltraDMA CRC error count parameter growing very fast, as minimum.

[faekjarz]

OK, those are interesting points to consider in order to avoid failure.

Nevertheless i want to try, and well, it's a hack, i know that. So please no insults, i also already know that i'm stubborn.
In case of success, it'll be an upgrade: The temptation is strong with this one. 
In case of failure, there's something to be learned. I'll document in this thread.

Here's my Plan B:

• I'll only hack one of the two channels - mainly to retain a pristine backup, a control group, one to compare kernel logs and SMART parameters against
• I'll not use relays, solid state relays or other switching mechanisms for now
• I will not "randomly add bits of wire", carefully keep an eye on the wire length, and hope those juicy differential pairs like what i do to them
• I already learned that the degradation of high frequency signals is worse over multi stranded wire - so i'll use solid core single strand magnet wire, like on the board, there's only one conductor too

Will my Plan B reduce the probability of failure?
What else should i consider to avoid failure?

[TheBay]

It's gonna end up in data corruption and a lot of triggered SMART events, so don't use it with any drive you care about or with any data.

[faekjarz]

Hmm, that's still not a sufficiently encouraging answer. But thanks, @TheBay!

A'ight here's my Plan C:

• I'll sever the traces on the board, as close as possible to the connectors (termination / reflection)
• I'll use a standard SATA cable - which is compliant to SATA III / 6Gbps spec, and shielded and stuff - to directly bridge the SATA connector to the eSATA socket

This should be death proof!
(Veto, anyone? Come right at me!)

_{*loves the juicy evolution of this idea*}

[TheBay]

Hmm, that's still not a sufficiently encouraging answer. But thanks, @TheBay!

A'ight here's my Plan C:

• I'll sever the traces on the board, as close as possible to the connectors (termination / reflection)
• I'll use a standard SATA cable - which is compliant to SATA III / 6Gbps spec, and shielded and stuff - to directly bridge the SATA connector to the eSATA socket

This should be death proof!
(Veto, anyone? Come right at me!)

_{*loves the juicy evolution of this idea*}

SATA can be a nightmare, I have seen data corruption with cheap/dodgy cables. And "X" amount of certain flags on SMART can render a drive useless or will constantly nag on boot up that it's failed.
So many variables with it and keeping cables the same length, capacitance, resistance etc. Will be a nightmare if you did decide to use relays. You may get away with removing the IC and putting some small flat bridge wires on the board as if both pass straight through directly to the eSATA ports, just bypassing the IC but using the same tracks if that makes sense?

[faekjarz]

Here's my log
…hacking in progress…

…the pin-out…


…soldering directly to the SATA connector that plugs into the storage device…


…marking pin 1…


If you do this twitter thing, there i am, follow me, or block me, do what you want ? https://twitter.com/faekjarz

[wraper]

Did you cut tracks very close to the connector? If you cut them at IC side, it will ruin signal integrity.

[faekjarz]

@wraper: nope, i cut as close to the connectors (SATA / eSATA) as possible
(let's hope those ~600mV electrons don't "jump" that ~1.5mm gap )



…getting ready to solder


…looking forward to benchmarking this puppy

[faekjarz]

…soldering done, re-assembling and benchmarking next…

I'll try a faulty old spinning HDD, that already throws SMART errors, first
…to see if any counter increase rapidly, and stuff
…to see if it's safe to run real benchmarks with an almost new SSD

[Monkeh]

Nice work. Actually got a pretty good chance of working okay.

[faekjarz]

Nice work. Actually got a pretty good chance of working okay.

Haha, that Kepton tape makes quite a professional look, doesn't it.

Well, however, my hack failed for now; the HDD doesn't spin up.

Although i don't think that it's the SATA link i modified. I guess, this now bypassed IC doesn't see a device and therefore has no reason to tell the rest of the circuitry to enable power to the HDD/SSD.

Apparently i'll have to find a datasheet for the INIC-1611, to see which pin says "Maximum warp, engage!"
Or maybe i can just reverse engineer (follow traces) the SATA power supply and bridge some pads.

[Monkeh]

U1, U2, U5, U12 are the power switches, I'd say.

[TheBay]

Have a look on Stationdrivers if there is a firmware for no power down on there, I have flashed a few "docks/enclosures" in the past to remove power saving.
Might be worth a look?

[faekjarz]

U1, U2, U5, U12 are the power switches, I'd say.

U1 and U5 appear to be responsible for port 2, so irrelevant for now.
Port 1 does business with U2 and U12, which are fairchild P95AG (both the same, same as U1 & U5)
I can't find a data sheet for those …maybe not needed, i can probe for continuity


here's the SATA power pin-out, pin 1 (notch) is near the data link pins


_{But… but Mr. judge justice, si…sir, MacGyver uses the Twitter CDN as a globally, without login, accessible image dump for forum posts too!}

[faekjarz]

Update: the DMM says 5V and 12V, where and when it's supposed to be present, it's not a power supply issue.
Maybe it _IS_ the SATA data link …i'm checking and probing.

Edit:
Aye, seems like i reversed the eSATA pin-out 
The intensity of my embarrassment is directly proportional to my re-soldering effort - in binary, that's ONE
…well, that's why i posted in the "Beginners" forum 
…no worries, no damage done, along the way i learned that it doesn't break things when the data pins are reversed, apparently (in my case)

[chicken]

U1 and U5 appear to be responsible for port 2, so irrelevant for now.
Port 1 does business with U2 and U12, which are fairchild P95AG (both the same, same as U1 & U5)
I can't find a data sheet for those …maybe not needed, i can probe for continuity

Fairchild FDS4435 P-channel MOSFET
https://www.fairchildsemi.com/products/discretes/fets/mosfets/FDS4435.html

[faekjarz]

Q'PLAH¹! SIX GIX!



…aaand one bad CRC 
I left ddrescue running till it ran out of bits, reading the whole 120 Gigger² Bite² produced only 2 bad CRCs.

I reworked my bypass on both connectors and the SATA link seems stable at 6 Gbps. The trick is to solder as straight and uniformly as possible, every wire cut to the same length, every wire soldered onto the pad (connector through hole pin) at the same distance to the PCB, none crooked or with weird bends.

That one HDD, that didn't spin up, did so because it is a little weirdo and liked to annoy me, maybe it needed a stable SATA link to spin up. All the other HDDs, although those couldn't establish a stable link either, spun up (without the eSATA cable even connected). That confirmed what the Multimeter already told about power supply.

¹: that's Klingon and means success
²: i'm tired, and that makes me stupid or funny, you choose (but don't tell me)

Fairchild FDS4435 P-channel MOSFET
https://www.fairchildsemi.com/products/discretes/fets/mosfets/FDS4435.html

OH! Thank you, Chicken. Next time i'll search for everything, that's written on the package.

[Monkeh]

Not too bad.

Get those centre conductors trimmed to ~1mm and keep the shield right up to the end and you should get better results. It'll get a lot trickier to solder, though!

Once you can get 10TB+ through it without an error, it's perfect.

[dmills]

Did you solder the screens down to the ground plane at each end of those cables, kind of important for impedance control....

Regards, Dan.

[faekjarz]

No, i removed the screen / shield, and the exposed ends are ~5mm long. There's still the 2nd channel to take care of, but i have to wait until the additionally ordered SATA cables come in. Then we'll see if i can do better.

edit: @dmills: those shielded pairs have two grounding conductors on either side, which are in contact to the shield / screen. (Next time i'll go with the ~1mm approach of @Monkeh)

Meanwhile maybe some Kepton tape and alumin(i)um foil helps to suppress those rebellious CRCs.

Thinking about soldering those ~1mm-stubbies …maybe a small conical tip isn't all that bad after all.  There's the right tip for every job, i guess.

[dmills]

You will want the screens (or their drain wires) tacked down to the ground plane at both ends or that wire will be nothing like the impedance it is supposed to be.

Regards, Dan.

[faekjarz]

You will want the screens (or their drain wires) tacked down to the ground plane at both ends or that wire will be nothing like the impedance it is supposed to be.

Aye, that's what i did. I thought that's obvious.

[faekjarz]

I reworked my soldering and shortened the exposed ends even more. Here's a before and after:







…that was probably not necessary because i found out, that it's the eSATA cable that's not rated for 6 Gbps. Although initially negotiated for 6 Gbps, after 4 CRC errors, something (i guess the Leenoox kernel) throttles down to 3 Gbps. That's what the system log says. I ordered 2 quite expensive (>10€ each) 6gig rated eSATA cables.

This, by the way, was my attempt to enhance the shielding. I wrapped AL foil all the way around and zip tied it to the original shielding, on both ends. …nope, it didn't became 6Gbps capable (not w/o CRC errors).