Topic: Hacking KNC Titan / Jupiter / Neptune miners back to life. Why not? - page 33. (Read 76775 times)

Felt the same way when I troubleshot my SP31.  Even emailed Spondoolies and asked about schematics.  That when exactly as expected, nowhere.

Most of what I now know about it I found out thorough experimentation.  Managed to get a non-working loop back, minus one noisy ASIC.

I need to keep reminding myself that I am doing high level logic and circuit analysis on boards that I have no documentation on, no schematics, no board layouts, no prior art to compare to, no source code, no logic trees, hell I don't even know what some of the PARTS are, and to be quite honest a very limited supply of units to even try to reverse engineer most of which are screwed up in some fashion or another.

Not to mention a company who doesn't say a peep.

Given all that I think I'm doing pretty well here. Just need to remind myself of that from time to time....

Now that I think about this it could make sense: The logic levels on the FPGAs are pretty low, there are only a few 3.3 volt capable lines for stuff like scl, so using these as relay amplifiers could make sense. They're not switching the power supplies, those are generic sda/self powered stuff.

Which leads the question: Just what *are* they powering?

Just popped in out of curiosity but level translators are use to translate I/O to-from core logic levels typically these days less than 1v, back in those days maybe as high as 2.2v, to-from data bus coms levels, usually 3.3v or up to 5v.

Well that's weird. Is the second part similar, this could explain why they have different parts. Still, what the heck are they and how did you find this much?

C

For your unidentified part, try this:

http://www.ti.com.cn/cn/lit/ds/sces613i/sces613i.pdf

Looks like it might match.

Interesting. Pin 4 also goes to the elusive and stupid U10, whatever that is. Big enough component to sustain a short, I wonder if it also has pin 6 going to it...

Hip deep in Titan thoughts here, a few things to note.

Today I took a look at what's going on with these things since no one can figure out what the fuck they are.

U17 is pretty easy to get to with a small heat sink, so I put a small sink on the unit and did some probing. Found an odd 1.0 volt signal on pin 1 only when the unit was hashing. Further screwing around found that when die 1 is on, there is a 1v signal there. When die 1 is off and any others are on, it's a 0 volt. Which means these signals are coming from the FPGA.

U19 is a bit different, it powers itself off pin 4 of the 10 pin yet U17 and the others power off pin *6*. That is interesting, I now know that pin 4 powers U17 and the temp/eeprom and pin 6 powers U17,U18,and U9. Why two power supplies? Fuck if I know.

So here is the latest map of the 10 pin interface.
Pin 1--U18 pin 1 (and 3?)
Pin 2--SDA signal
Pin 3--U19, pin 4
Pin 4--VCC, U19 pin 5
Pin 5--U17 pin 3
Pin 6--VCC 2, U17 pin 5, u18p5,u9p5
Pin 7--SCL signal
Pin 8--One side of those lines on each of the the CPU dies.
Pin 9--U17 pin 1, 1.0v only when die 1 is on.
Pin 10--Ground

I wonder what pin 1 looks like when running. Hm.

Searing:

Not sure, lot to read there. However someone posted that there is a crystal on the Titan bridge board, I haven't taken one apart but that might be needed. Someone with a Bridge would know a bit more.

And head deep in a Titan right now. Hm.

   




Hey figured I'd better ask 'supposedly' I have all the parts below to see if the old KNC 550gh Jupiter 4 port old style ...stacked LCD board works as a Titan controller or not
...but have a questions

Here are the parts that I was told (pi by knc back in the day) would work (its knc who knows)

1) New Raspberry PI B+ 512mb (seems to match the ones I have)

2) (1) Titan Clone Board from the 10 card run from Qberty.

3) KNC Jupiter 550gh 4 port Board (Oct 18th 2013 arrival) (and of course BBB shall be removed)

4) Left over from my NOW 6 port Titan Hashing the Titan LCD plug and play on my orig Titan 6 port board as such. I can now use on this card.

5) Ribbon cabless and y adapters will come with the cubes I have coming (1 full cube and 1 One  die working only Titan cube to 'sacrifice" to the ASIC God's on this project ..if needed)

6) To start an SanDisk SD class 10 with the default KNC last Firmware IMG then later if this all 'floats' Glen Tarkin's Firmware.



To my questions:

MOST IMPORTANT QUESTION BELOW IS (1)!!!! (no clue!)

1) IS there some pins/settings/dips I have to do on the PI BEFORE I snap it into the 4 port board via the titan clone bridge?

2) I don't need these at the present time..BUT are the KNC Jupiter 550gh cables compatible with the Titan ribbon cables?

3) Is it hard to put this titan bridge port with a PI on and also the corresponding whatever 4/6 port board? (ie any tricks to this)

4)  Anything else that comes to mind before I commit to this!

Thanks


I sent a copy of the above to Lightfoot but figured people might also have these questions so feel free to chime in also if you've
had any expience with say:

Adding a new PI to your Titan Board (any pin settings/dips etc Again is my most important question in all this)

The corsair 1200i replacement (they RMA'd it surpisingly) comes on Friday Feb 19th...just trying to get my ducks in a row

(should probably make video just to watch 'fireworks' and 'it go ka boom' ) I'm always available for my 'tragedies' to amuse others.

thanks

It simply plugs in to the first port and also connects the the 10 pin on the left side.

I have some if you want to see a picture.

thanks


boomin

You won't have to twist my arm!  I have access to a Cirrus Sr22.  My favorite plane to fly!

Thanks again man!  You are a gentleman and a scholar! 

Not to mention a great teacher and inspiration!

Carry on.

Boomin

If my controllers don't have a lcd, is it possible to retrofit one? It looks to be a 16x2 but I dont know if its serial or parallel. By looks of the controller board on google images it looks to be serial (jp 13?), that would mean 3 wires. Power, Ground and RX

No problem! I might give another talk at DefCon about hacking hardware, when I get out there this summer let's meet for a beer. And maybe rent a plane for a quick jaunt around, just not in that order :-)

C

I have to say - Lightfoot is the MAN!  Not only did he fix a few of my controllers for me - he also advised me about the tools I should buy to tinker!  I just successfully fixed one of my own controllers!  

Lightfoot is a different kind of cool.  The kind that are in rare form these days!

I just told him, this is so cool!  I only had to spend a little over $500 in tools, tons of hours learning and practicing and now I fixed my own controller! HAHA

Las Vegas now has a little more hash! LOL

Thanks again Lightfoot!

Boomin

Well if you run them more slowly then one connector should be fine.

In the meantime I unboxed a shorted Neptune that came in today. Unlike many of the shorted units this one had no capacitor damage, everything looked fine. Hm. So I checked continuity on the 10 pin connector, all good there, then checked the resistance of the cube power lines.

One pair of supplies had 0 ohms, which usually points to a short in the high side FETs. The reason is that those FETs kick power from the 12 volt rail to the .7 volt rail. The trick is they are on for a very short time, so the voltage rises a bit then goes down.

In this case one of the FETs had shorted (probably overheat) which put the full 12v on the chip die. Down goes the unit, but more interestingly bye bye goes the die. Removing the FETs and restarting shows a die that will suck power, but will not come up to hash. Still this is another avenue for fixing a shorted Neptune or Titan and can probably get at least 2-3 dies back online.

Never dull. Got another controller to fix, and am packing up the latest one to go back to the client. I still haven't figured out the Titan issue, if I can't by tomorrow I will send it back and back-burner this for a bit.

Looking to buy a Neptune in hopes of underclocking it for maximum power saving.  What is the lowest voltage you have used and had all the dies come

up and hash.  Do these things to .5w/GH or better?

Well I should say they are Neptunes, I wish they were Titans. I am just getting bare boards and cables only to save on shipping. But I could eventually buy a Titan cube here and there. By far the best $/Ghs deal I have ever gotten. Hopefully if all is well I will throw on some heat sinks and solder another 6 pin on them, hopefully underclock/undervolt to replace my noisy s5s in my house.

So I got another controller working here. This was the third of three fried controllers, this one was the worst with lifted PCB traces on the TPS power supply side. To be honest it took *days* to troubleshoot and clear the three shorts and one open connection. But it's all cleaned up, new solder to bridge the connections.

I did have to use a .5mm solder ball to bridge one of the connections, but that's life, it's a sense lead not a power lead (thus resistance is negligible) and it's securely holding without any shorts anymore.

Put down the FPGA (getting very good at this), fired it up, nice white then green light and it's hashing on the test Neptunes. I'll box it up and send it out tomorrow to another hopefully happy customer.

Time for lunch.

Neat! I highly recommend taking things one slow step at a time in setting them up. For example I would:

Start up each controller without anything attached. Look for the boot, the IP address, and of course the light show. Upgrade each controller to the latest software version and make sure it's doing what you would expect.

Fire up each Titan on a power supply. Smell it. Smell for burned stuff, see if it blows up the power supply, etc. Then (if you know how to check resistance look for shorts on the 10 pin plug to ground. Only one pin should do this). After doing that I would plug in a Titan to one controller, make sure it works, hashes, whatever then move it to the "production controller" along with peers.

Good luck and mine like hell!