This card is pretty good, when it works. Transfer-speeds were fine, close to 100MB/s with an external USB3 disk. It works nicely with Linux – plug and play.
Then the problems started…
…the card wouldn’t be recognized by the kernel anymore (apparently at random), strange hissing / whistling noises came from the card, much like an inductor being tortured or a ceramic capacitor dealing with higher-frequency current. I switched PCI-E slots, wiggled connectors here and there, tested the power supply voltages, cleaned connectors, all sorts of stuff. All fine!
It appeared to be working afterwards for a while, but that might have been chance, maybe a thermal issue, who knows. Time passed…
THEN… my monitor went dark!
Power cycling the computer didn’t work, it only made half-arsed attempts of booting up. Then I pulled the plug and pulled the card.
Fortunately it didn’t take down the computer with it, that would’ve been bitter. I still had warranty, so I claimed it. 2 working days later I got a replacement card, which works nicely, for now… fingers crossed.
Now to the fun part:
There are 2 different DC-DC (buck) regulators on this card. One either takes 5V or 12V and might be a pre-regulator (ACT4060A).
The 2nd one (TD6810) seems to be creating the core voltage (1.05V) for the USB3 controller chip, a Renesas µPD720202.
In step #1 I’ll wire up a 3.3V and a 5V regulator to create the supply rails for this card. The 12V will come straight from my power supply. The PCI-E part only needs 3.3V and 12V, but as the power-plug also supplies 5V, better play it safe. I suspect it goes straight to the USB connectors, but who knows. The ACT4060A might use it.
In step #2 I’ll probe around on the board and see which regulator still works, what the output voltage looks like and whatever else I might find.
…some time later…
Well, it seems the 3.3V rail is shorted to GND (Diode/beep test-voltate: ~1.4V, GND-lead on GND)! That would explain the sudden denial of operation:
A4 (GND) beeps with: A9 (3.3V), A10 (3.3V), A12 (GND), A15 (GND), A18 (GND)
Let’s use the mOhm-Meter. Well, this is odd… all seems fine now (Test-voltage: 1.2V).
Let’s use a different DMM (Diode test-voltage 3.1V). Very odd. There is a 0.7V diode drop from A9/A10(3.3V) to GND. Reversing polarity gives 0.2V.
This is seriously fucked up it seems. I might have expected some sort of signal clamping on the differential-pairs, but nothing of that sort on the 3.3V power rail (which doesn’t really make sense with the low voltage that triggers it).
You know what, I’ll just pull the µPD720202 off the board and test again. I can’t do anything with that one anyway!
Testing with DMM1 (Diode/beep 1.4V) A9/A10 still stand out a tiny bit, but that may be capacitance or some sort of leakage. The B-side looks good too. Definitely an improvement. One may come to the conclusion that the Renesas chip was fried.
…you know the drill…
The 12V on the auxiliary power connector is not directly connected with the 12V on the PCI-E bus.
12V from the PCI-E bus connects to D2, 12V from the aux. connector goes to D3. They are diode-OR-ed. Then 12V goes to the input of the ACT4060A (pin2).
Therefore neither 3.3V nor 5V go there ;-)
The output out the ACT (pin3) goes to the “220” inductor and then to the 2 green poly-switch fuses (about 100mR when cold). The 5V power for the USB ports is therefore derived from 12V. Why? Probably so it also kinda works without the aux. connector, but that is only good for low-power devices. Not sure how much current D2 can handle!
3.3V (A9,A10,B8) go to the input (pin 4) of the TDA6810. The core voltage for the µPD720202 is derived from the PCI-E bus.
The 5V on the aux. connector appears not to be used at all! At least I can’t find it.
For testing the ACT4060A it should therefore be sufficient to inject power via the aux. connector’s 12V/GND lines. The diode-OR-ing tested fine, so I won’t need to solder a 12V wire to the PCI-E connector.
The board takes just 10mA @12V, no short. Good. The USB-port supply voltage measures as 5.03V (no load). Good so far. Connected an USB stick, input current goes up to 60mA, then down to 20mA. 5.03V. Good. A few tiny sparks when charging the 12V input caps. USB stick still works ;-)
Now testing with my dummy load… ignore my feet, please!
Looks good! Overall efficiency of about 86% @1A.
The ACT4060A goes into some sort of low-power mode when the current exceeds 1.1A, it recovers once the current falls below about 450mA. The 1.1A fuse didn’t “blow”.
Marking unused pins on the PCI-E board-edge-connector.
Soldering a 3.3V bearing wire to the PCI-E board-edge-connector.
Applied 3.3V to B8, current < 5mA. TD6810 Vin: 3.3V, OK. Vout: 0 ;-( ‘RUN’ (pin 1) must be above 1.5V to enable it, let’s see… 3.3V. It should work! Did I measure properly? Let’s push down harder on the probes: VFB: 0.64V, Vout: 1.072V. BINGO, it works too!
Let’s put a load on, say 150mA (datasheet: up to 800mA). Result: 1.054V. GOOD!
Efficiency 82% @ 150mA. Not bad.
Verdict: The USB3 controller IC was the culprit. Why?
P.S.: Yes, I probably should’ve fired up the old oscilloscope (Tektronix 314), but that wasn’t worth the effort. The controller card is mostly scrap anyway, it went into a box for further salvaging, sometime later.