RST, RXD, and TXD make up the serial connection to the ASIC chips.

SCL and SDA is the I2C interface to the PIC microcontroller, eeprom, and temperature sensors. It looks like A0, A1, and A2 are used to set the address of the eeprom and PIC on the I2C bus.

First, check that the correct voltage is present at the output of the voltage regulator for that voltage domain, then trace it through each chip to find out where it goes bad. Most likely, a connection in or out of an ASIC chip is not good.

The capacitors are just convenient places to measure, if the voltage is not correct they would not be the components you would need to replace. If voltages are bad at the capacitors, then normally either the connections are not being made through the asics, or the voltage regulator has failed.

Pretty sure it's identical to 17 series since I think the control boards are compatible, which is this:

Are you checking the IO supply voltages? Once you find where either RO or CO is bad, the next thing you should do is check the IO supply voltages (1.8V and 0.8V) for that chip. The 1.8V voltage is generated by a regulator at each voltage domain, and the 0.8V supply voltage is generated by two different regulators per voltage domain. They are fed through each chip. So the output of the regulator could be fine, but the input to a chip may not be if there is a bad connection to a chip before.

If the IO supplies aren't good, then a perfectly good chip will not have good RO, CO, and CLK signals.



You can measure the IO voltages for each chip at the capacitors shown below:

Unfortunately, your best option is probably to just mine with 2 hashboards. This is a very common issue with these miners.

If you can find a shop nearby that fixes them, it looks like the going rate is ~$200 to repair a hashboard.

Not finding all the chips on chain 0. It is likely that either excess solder laying on the hasboard moved during shipment, or marginal solder connections to a chip opened up. You can try tilting the miner or putting it on it's side like someone suggested in this thread. Or just shaking it around a bit. When I'm testing hashboards, many times tapping the board on a table and flexing it will produce or solve issues. So you might be able to get it running at least for some period of time that way.

Thanks, let me know what you find. I have a bunch of APW9s, so one option is to just remove the PSU which is ~40% of the weight.

---

I've seen a lot of excess solder that dripped down from heatsinks on hashboards I've repaired. The solder forms small solder balls that can sit right between the exposed pads for the chip connections and the bottom of the heatsink. It can sometimes be very close to shorting out signals, and just a bit of pressure on the heatsink can cause it to touch the signals and short them out. I'm sure the opposite is true as well, so a bit of pressure in the right spot could pull the solder away enough to open a short.

---

I find that very frequently heating and removing a heatsink will melt excess solder on nearby chips causing other issues. Also just handling and flexing the board can expose other issues. Especially when starting out, it is a good idea to test in stages as you put chips and heatsinks back on. I set the test fixture to only do the asic count test (this can be set in the config.ini file on the sd card, Only_find_ASIC=1), and will test the board before applying either heatsink, then after applying the back heatsink, and again after applying the top heatsink. Many times the board will find all chips fine, but after applying the back heatsink for the chip I'm working on I get 0 asics.

I use a soldering iron set at 315 deg C to remove solder from chips. As long as you don't let the chip get too hot for too long, it won't get damaged. The key is time + temperature. Most components and the PCB have materials that will start to break down above 150 to 200 degC. Both have specs for how long it can safely be above a certain temperature so they can be reflow soldered in a reflow oven.

As for removing solder from the board before installing a new chip, I do that because I find it very difficult to place the chip when there is solder on both the board and the chip. Since the solder on the board and the chip create rounded bumps, the bumps on the chip nearly always will fall off the bumps on the board.

I'm in the northeast USA.

I'd consider buying them, but if they are partially working it may be better for you to limp along with them while profitability is as high as it is now. Guess that depends on how often the 5 sometimes working hashboards go down.

If you want to sell any of your dead gear instead of watching it do nothing, shoot me a PM. Shipping might be too expensive to make it worth it, but maybe not.

Not sure how commercial places are making money repairing these. Hmtech charges $40 diagnosis + $150/hashboard for "basic" repair, $300/hashboard for "advanced" repair. I'm sure their techs are way more skilled than me, but I wonder if they would repair the board from my last post for $300 or if they'd just say it's not repairable. It's possible it could need $100 worth of replacement chips.

So far, every board I've fixed has taken more than 8 hours of work. The S17+ that I repaired took ~24 hours of work, although that was the first S17+ for me so there was some learning curve on that one.

I also found on the S17+ that chips will come up with the heatsink when trying to just remove the heatsink. So just pulling heatsinks to get access to repair other chips can cause other issues.

I use a heatgun set at 315 degrees C set at low airflow to remove/attach heatsinks and chips.

Attaching heatsinks on S17 boards is pretty tricky. For S17 and S17 pro, I've been removing most residual solder from the heatsink and adding what I think is the correct amount of solder to the top of the ASIC. I put some tacky flux on the top of the chip, then heat the heatsink with my heatgun for ~20 seconds and set it down on the chip. There is then a small window of time where you can adjust the placement of the heatsink. If you are not quick enough, the solder holding the chip to the board can melt and then you can shift the chip.

On the S17+, that technique didn't work for chips with the smaller heatsinks. So on those, I prepare the chip and heatsink the same way, but then just place the heatsink on the chip and heat it directly until the solder flows.

For attaching chips, I remove the solder from the board with solder wick, apply tacky flux to the board and place the chip in position, then apply heat for ~45 seconds. For new chips, I use a stencil to apply solder paste to the chip and melt the solder before placing it on the board. Nearly every time I do this, all connections are not made, so I add solder paste with a small syringe to the edge of the chip, reflow, add more flux, reflow again, and clean up any excess solder with an exacto knife while the solder is still liquid. Pretty tedious process.

Flux I use: https://www.chipquik.com/store/product_info.php?products_id=310003
solder paste I use: https://www.chipquik.com/store/product_info.php?products_id=470006
solder wick: https://www.chipquik.com/store/product_info.php?products_id=640002

Heat gun I use: https://www.masterappliance.com/proheat-1500-lcd-programmable-professional-heat-gun-kit/

---

Started working on another S17+, this one is a real mess. One hashboard had 9 heatsinks fall off, some with chips still attached.

The heatsinks that fell off look like they came off with the film from the top of the chip. Without the film on the chip, it is impossible to get solder to flow evenly on the top of the chip, so impossible to get a good contact with the heatsink. I saw the same issue on some chips on another S17+. Looks like a manufacturing issue with the chip packages. Or maybe it just breaks down if it gets too hot for too long.

To save the chips, I'd have to use a thermally conductive adhesive instead of solder to attach the heatsinks. Not too excited about that idea. A lot of the chips got hot enough to melt the solder holding them to the board and slid down towards the bottom of the miner. Pretty sure some/most of them would be toast anyway so might just replace them all.

Pretty sure this board is going in the "maybe later" pile.



These chips would normally be covered in solder, or would be an amber color if a heatsink was never attached.



Most of the film from the chip is still stuck to the heatsink.

Ok, I'll get a quote and PM you.

If anyone else wants one let me know soon, I'm going to try to get an order placed within the next few days.

Yes, so far I've repaired 3 S17 hashboards and one S17+ hashboard. Most were reporting 0 Asics at some point. I'd say more issues cause 0 asics than some number fewer than the total asics.

To solve it, you need to trace the signals through the board to try to locate the spot where they go bad. The repair manuals on Zuesbtc are a good place to start. It's not easy though, both the locating of issues and then the techniques to remove/replace heatskins and chips. I'm an Electrical Engineer with >25 years experience and it took me a month of nights and weekends to decode the manuals/reverse engineer the hashboards to get to the point where I could repair one. I'm sure it would have been a lot easier to just take the Bitmain repair course, but I guess I'm just a glutton for punishment...

Is anyone still looking for 1047 or 1066 PSUs? I have been looking for a 1047 psu and finally got through to someone at Canaan that let me know they have some used 1047 and 1066 PSUs.  They quoted me $110 for the 1047 psu, but $150 for dhl shipping to northeast USA. If anyone is interested shoot me a PM and I'll give you the contact email.

Also, I'm trying to find out if shipping per unit will be significantly cheaper if multiple PSUs are ordered. If so it might make sense to coordinate a group order.

Yeah, the tester is only a small part of being able to repair hashboards. The output of the test is only sometimes useful, most of the time you need the fixture just to operate the board while you use an oscilloscope and/or voltmeter to measure signals at the test points.

Just FYI, with the COVID situation, Bitmain now offers an online repair course for $500. I haven't taken it so I'm not sure how useful it is, but once you complete it you get access to all of their official repair manuals and downloads, and you can order Bitmain's test fixture and spare parts direct from them. Previously it was $1000 and you had to travel to one of their sites to take the 1-week long course.

All of the testers of that type are built similarly ... I agree that it is not the best. Just kind of hacked together. I've ordered a different tester from here https://tester.asic.repair/en. It was much more expensive and ships from Russia but looks like a much better product. Slow shipping from Russia to USA though ... ordered a month ago and the tracking estimated arrival is still 3 to 5 more weeks...

I think the manufacturing capacity is already booked ... so I don't think there's anything left to buy or bid on.

I suppose if the price goes high enough they'll just start manufacturing old tech again ... If the price goes up to 100k and the diff can't go up that far, they could just start selling S9s again at a profit.... Maybe even take the S9,S11, or S15 chip and put it in the S19 form factor.

So that would be about 100,000 wafers. A few articles in early 2020 I found indicated that TSMCs 7nm capacity is about 140,000 wafers per month.


If that capacity is accurate, then Bitmain would have needed to secure 6% of TSMCs total production capacity for the whole year to get their 100,000 wafers worth of chips. So possible I guess, but with the Apple and AMD fighting over the 7nm capacity, don't know what scraps are left.

So, for the s17 chip (bm1397AD), package dimensions are 8x8 mm, die looks like about 5mm x 5mm. Still seems like a lot of wafer real estate for a chip you can buy for $6. S19 chips are going for about $25.

I'd be surprised if the die is actually that big ... on a 7nm node that's a lot of transistors. A 14nm node 6 core i7 is about 12x12mm.

I have a bunch of dead S17 chips, I'll see if I can grind one down to expose the die.