So the exact thing I've always feared with 400 different modular power supplies all using the same keyed connectors for different voltages finally happened last night, I used the wrong cable and applied +12 (or at least +7 depending on where ground was/wasn't) and fried a 3tb HDD, a Vector 150 240gb, and my last BluRay/HD-DVD combo drive. All 3 are dead of course, but the SSD is the only thing that had a couple of non-backed up notepad files that I'd worked on that day.

http://www.richtek.com/assets/product_file/RT8070/DS8070-08.pdf

That's the datasheet for the chip that melted, I lifted it (badly) and accidentally pulled one of the pads, and I'm not sure if the copper square is what's left of the 9th pin/ground pad from U20 or part of what that soldered to on the board side.

Obviously I'm not too solid on soldering surface mount stuff, and the only tools I have are a Hakko 936 and a Hakko de-soldering gun. the 936 has one of the two smallest tips they make but still not really ideal for this.

Also I should mention I've got a couple 128gb versions of the Vector 150 and they've got that same chip in them, wouldn't mind donating parts but I'm not sure how to handle that 9th pin/grounding plate to get a clean removal. Also can order that chip for a few bucks and not really a time crunch since it's just a couple files.

So suggestions? Was it too late as soon as I pulled the pad? I've has some luck in the past sanding off the edge of the trace and re soldering to that, but idk if that kind of hax would work here.

Pin #7 is the feedback pin. If you leave that open, then the output of the chip will probably hit the upper limit, namely +5V. You definitely don't want that.

I would first perform some simple resistnace measurements to assess the feasibility of a repair.

Confirm that none of the coils (3 x 4R7 and 1 x 2R2) are open circuit. Then measure the resistance between each coil and ground. This will check whether any of the loads (SDRAM, NAND flash, flash controller) are shorted.

Note that U17 has a black spot near C317. It appears to be a triple-output buck regulator. Can you determine its part markings (TPSxxxxx)?

_________________
A backup a day keeps DR away.

All 4 of those coils show continuity, and all 4 measure a resistance between each terminal and ground.

U17 looks like a TPS 652510

See viewtopic.php?f=10&t=28767

The large regulator appears to be a TPS652510 which is rated for 16V on its input.

_________________
A backup a day keeps DR away.

@thanatos2k, can you show us the bottom of the PCB? Similar OCZ boards leave U20 unpopulated. I have an idea that does not require replacing the chip, but I need to find out where the supply goes.

_________________
A backup a day keeps DR away.

Here's the underside pic.

Hmm interesting, I think it didn't post my previous response. I tested the coils for open and resistance to ground, all are fine.

U17 is the 652510 ya.

Initially I thought that the RT8070 regulator may have been a secondary regulator that supplied Vcc to the NAND array on the other side of the board while the TPS652510 supplied the top side array. Now I believe that the former supplies the VccQ voltage to the NAND arrays on both sides of the PCB while the TPS652510 supplies the Vcc to both arrays. The reverse side of the board has 4 configuration links (zero-ohm resistors) which appear to select which of two voltages (V2 or V5) supplies the Vcc and VccQ.

On those PCBs where the RT8070 location is unpopulated, the 4 resistors select the V2 voltage (TPS652510). Your PCB has the resistors in opposing positions. AFAICT this means that V5 supplies VccQ while V2 supplies Vcc. You could confirm this with a continuity test.

My approach would be to power up the PCB as is and check for the presence of V1 - V4. This will confirm whether the TPS652510 and 2.5V LDO are working.

My next test would be to measure the values of the two feedback resistors for the RT8070. We could then calculate the voltage output at V5. To this end you would need to measure the resistance between pin 7 and ground, and between pin 7 and V5. I expect that the result might be 1.8V.

You could try to replace the RT8070 and repair the connection to pin 7, or you could patch an external 1.8V (?) supply into V5. Alternatively we could investigate whether it is possible to power the NAND VccQ rail from 3.3V rather than 1.8V. This would involve adding resistors R4 and R2.

This is what I have been able to find out about your NAND chips:

TH58TEG7DDJBA4C, Toshiba, 16GB 19nm, Toggle Mode MLC NAND flash

TH58 = multi chip
T = toggle mode
E = voltage ??? (Vcc/VccQ)
G7 = 128Gbits = 16GB
D = 4 level (2 bits per cell)
D = organisation (x8/x16, page size, block size)
J = design rule (19nm ??)
BA = BGA
4 = dual channel, 2 x CE
C = 52 lands, 11x14x0.9

Part Number Decoder for Toshiba NAND Flash:
http://www.triton-prog.ru/Dnl/Nand_Toshiba_NEW.pdf

Unfortunately the part number decoder doesn't tell us how the voltages are configured (E). If type E is similar to B or D, then we could power VccQ from 3.3V (V2).


We would need to be very careful if we were to choose the easy route.

_________________
A backup a day keeps DR away.

18k ohm between the remains of pin 7's pad and ground as well as pin 7 and V5 off the back (R1/3/5/7)
V1 - 5.01v
V2 - 1.80v
V3 - 1.50v
V4 - 1.00v
V5 - 1.53v

As far as replacing U20 or patching a new supply, I don't really have an opinion on it. Soldering-wise I think hacking it would be easier for me, I worry about that grounding pad with the equipment I have. But unknown about what voltage to provide could make it tricky.

I should also mention I've got several (fairly cheap: FleaBay) adjustable bench power supplies if that would help at all with a temporary repair.

A few surprises there.

Firstly, V1 should be +2.5V. Perhaps I confused you by placing the "V1" text adjacent to the centre pin. The voltages should be 0V - 5.0V - 2.5V.

V3 appears to be the SDRAM supply. AFAICT, the SDRAM is nominally a 1.35V part but it can operate in 1.5V mode.

MT41K256M8DA-125:M, Micron, 1.35V, 32M x 8 x 8 banks, marking D9PFJ:
http://www.micron.com/~/media/documents ... _ddr3l.pdf

V4 looks to be the Vcore for the Indilinx controller. A value of 1.0V seems typical.

I wasn't expecting V2 to be 1.8V, but the "D" version of NAND flash does in fact allow for a 1.8V Vcc, so perhaps this is correct. Otherwise, it may be that the Vcc and VccQ rails are reversed. That is, it could be that the rail I have identified as Vcc is in fact VccQ, and VccQ is in fact Vcc.

V5 is the real surprise, though. The RT8070 generates this supply, but the IC is not present, so where is V5 coming from? Is this normally what happens when VccQ is not connected, ie are the NAND flash chips making some internal connection between Vcc and VccQ, and is this to be expected or does it indicate a failure in one or more of them?

The resistances at pin 7 of the RT8070 appear to be impossible to measure in-circuit as they appear to be loaded by the surrounding circuitry. You would need to remove one of them, but that would require a little microsurgery.

In short, assuming that V1 is in fact +2.5V, we need to confirm the voltage specs for the "E" version of NAND flash and we need to confirm whether I have correctly identified the Vcc and VccQ rails, assuming that is what they are. I'm not sure how we can proceed until we have this information. A datasheet would be ideal. Try contacting Toshiba.

_________________
A backup a day keeps DR away.

Oh it does appear that I had measured the middle pin for V1, the correct point does read 2.5v.

I popped the cover on a 120gb Vector 150 and the PCB is very similar except for the side with U20 is not populated with memory chips. It does use the same "E" toshiba chips, and V2 I measured at 1.8v. Is that enough data to assume V2 @ 1.8 is normal on the 240gb as well?

The V5 on that board measured 3.3 volts if that helps at all.

V1, 3, and 4 all match also on the two.

I had originally thought that the V2 and V5 regulators supplied the NAND arrays on separate sides of the PCB, but then I found a similar PCB where both sides were populated with flash but U20 was missing. That's what led me to discount my idea and look for an alternative explanation, namely the split Vcc and VccQ hypothesis. Your most recent finding would now suggest that Vcc and VccQ are not split, but that V2 and V5 each power separate NAND arrays.

Could you test for continuity between the V2 and V5 supplies and the various test points around the configuration resistors on both boards? Could you also measure the voltages in the area of the configuration resistors on your working single-sided PCB?

I don't understand why V5 measures 3.3V on the unpopulated PCB. That's a real puzzle. Are there any other chips that could be responsible for this voltage?

One other thing that could help to sort out the Vcc/VccQ confusion would be to perform continuity checks between V2, V5 and each of the bypass capacitors adjacent to one of the NAND flash chips. Some caps would be bypassing Vcc while others would be bypassing the VccQ rail.

_________________
A backup a day keeps DR away.

I just noticed that the TPS652510 has a PowerGood pin that monitors the state of all 3 of its regulators.

_________________
A backup a day keeps DR away.

Here's the front and back of the good board.

On the measurements, I'm not sure which ones to measure. I understand the basics of what each component does and how to measure it, but I'm not able to see which component is doing what in the circuit.

That said, I tested for continuity on the bad board between V2/V5 and some of the bigger caps near the nand chips, and some have continuity to V2 and some to V5, C56 and C57 are examples. Not sure if that helps at all.

Also ordered one of those RT8070s from digikey for $1, should be here in a couple days if we end up needing it.

A picture is worth a thousand words. I had misunderstood that U20 was unpopulated on your 120GB board. It is now clear that U20 provides 3.3V to the NAND array on both sides of the board, and that the configuration resistors do in fact select the Vcc and VccQ sources.

You could confirm the function of the configuration resistors by checking for +3.3V and +1.8V at R1-R8, or you could simply check the resistance between R1-R8 and each of V2 and V5.

I had wondered whether it was possible to power the NAND array from a single regulator, as was the case in that other HDD Guru thread. However, it now appears that the OCZ flash in that thread uses a single 3.3V supply whereas your Toshiba flash requires 1.8V and 3.3V.

AISI you now have two options. Either replace the RT8070 and run a tiny wire link from pin 7 to the adjacent through-hole, or patch in an external 3.3V regulator. You may even be able to use the +3.3V from your PC PSU, although this would be relatively "dirty".

_________________
A backup a day keeps DR away.

Could this be a viable solution?

_________________
A backup a day keeps DR away.

Oops, brain fart. That's the +12V end of the SATA connector. The other end is the +3.3V end.

_________________
A backup a day keeps DR away.

I soldered a jumper between the +3.3 and V5 like in the bottom pic, but the drive wasn't detected. It's still on there so if there's any troubleshooting to do beyond that.

I do have a RT8070 coming on Thursday but I'm concerned about being able to verify whatever connection I make to the little through-hole. If that ends up being open you mentioned the RT8070 is probably going to output +5v which would be bad for NAND that expects 3.3, no?

I'm willing to try the external 3.3v regulator option if you can give me some guidance parts/connection-wise. But I'm fine waiting till Thursday and giving that a shot too.

It could be that the RT8070 goes into protection mode when the feedback pin is open, but most chips would hit the upper stops under such circumstances.

Did you confirm that 3.3V was present at the configuration resistors on your 120GB board? I asked you to confirm continuity between each of the V2 and V5 regulators and the respective resistors as I was only guessing as to their function. Did you do that for your damaged board?

Did you check that 3.3V was present at the SATA connector?

Assuming all the voltages are where they should be, it could be that the RT8070 failed in such a way that it took out one or more of the flash chips downstream.

_________________
A backup a day keeps DR away.