Open Source Bitcoin ASIC miner project that uses 2x BM1387 (Antminer S9)

[Skot]

Only working with 4 x BM1397 in Series

This is where being clear on terminology used is crucial: When you say 'series' I assume you are talking about daisy-chaining the data chain and the Vcore is feeding power to the 4x chips in parallel right? I recall seeing that the s17 chip can pull up to 15A(?) so x4 in parallel that means the regulator will be supplying up to 60 amps. Those better be some pretty hefty power planes...

The supposed efficiency of the BM1397 is 0.03 J/GH. If you're pulling 15A at 1.5V that's 22.5W. At the rated efficiency of 0.03 J/GH, 22.5W should be 750 GH/s. I don't think the BM1397 can go that high. I think the highest practical hashrate is more like 350 GH/s, which would be 7A @ 1.5V

Of course they don't say what hashrate gets you the stated efficiency.

Anyways, I agree with NotFuzzyWarm, we should parallel the power and connect the data lines in series

[1714110326]

1714110326

[1714110326]

1714110326

[1714110326]

1714110326

[1714110326]

1714110326

[1714110326]

1714110326

[developeralgo]

The supposed efficiency of the BM1397 is 0.03 J/GH. If you're pulling 15A at 1.5V that's 22.5W. At the rated efficiency of 0.03 J/GH, 22.5W should be 750 GH/s. I don't think the BM1397 can go that high. I think the highest practical hashrate is more like 350 GH/s, which would be 7A @ 1.5V

I might be mixing my Terminology. From the little info floating around, here is what one of the repair S17 manual  says

Voltage domain: The whole S17 Hash board has 15 voltage domains, and each voltage has 9 chips. The total voltage domain voltage is 18V. The voltage of each voltage domain is 1.2V. The 9 chips in the same voltage domain are powered in parallel, and then connected in series with other voltage domains after being connected in parallel.

and each voltage domain has only 1A 1.8V and 0.8V LDO powers 9 chips in a voltage domain.

I agree i don't think it would suck that much power for 4 x BM1397 chips in a single voltage domain connected in Parallel , if that's the case then i am getting a frying pan and some eggs, its going to get hot.

[NotFuzzyWarm]

Going by the Compac-F which can pull >15w (overclocked) with Vcore of say 1.5v that gives the one chip pulling 10A. If Vcore is 1.4v then it's 10.7A per chip. at 1.2v, 12.5A. Using 9 chips in parallel for each block (domain) in the string that means each block is pulling between 90 to 112.5a which fits well with the s17 PSU can put out per-board. Even using 10w power @1.2v it's still 8.333A/chip or 75A per block of 9.

Those currents explain why there is extensive use of bus bars soldered on top of the power planes in miners like the s17. Canaan does the same in their Avalons.

Somewhere on Bitfury's github they have the full open source schematics and PCB gerbers/BOMs for their older model miners that used either 24 or 48 vdc per-string so those could be used as a reference to using strings.

[n0nce]

The supposed efficiency of the BM1397 is 0.03 J/GH. If you're pulling 15A at 1.5V that's 22.5W. At the rated efficiency of 0.03 J/GH, 22.5W should be 750 GH/s. I don't think the BM1397 can go that high. I think the highest practical hashrate is more like 350 GH/s, which would be 7A @ 1.5V

In my experience, you can run one BM1397 (Compac F) at 500+GH/s, pulling around 15W at the USB port. That would result in exactly 0.03J/GH if I'm not wrong, however I don't know about giving it more than 15W. That's 10A @ 1.5V through the core for 0.5TH/s per chip.

[sidehack]

The BM1397 will go to 750MHz. The Compac F will go to 750MHz, know of a guy that has a few running on that setting for multiple months. You're more likely around 1.6V core, and closer to 40W/TH efficiency than 30. Stock setting on S17 models is around 35W/TH at the chip, T17 around 45W/TH. 15W into a Compac F will get you about 350GH.

I've yet to see an S17 with 9 chips in parallel. Remember the total chip count is spread across three boards. What I'm most familiar with are the 195-chip version, with 65 chips per board in 13 sets of 5 and a domain around 1.5V

Bitfury's last couple generations of chip used a very different level shifting and data multiplexing setup than the daisy-chain of their original chips and what most other vendors use now (including Bitmain).

Fuzzy, the ASICMiner Tube used TPS53355 regulators for every I think three ASICs. The Prisma, using the same heatsink profile, took in unregulated 12V into a series string. I believe Bitfury was already capable of series but I don't know much about early generation BF gear, since it was expensive (boasting 2W/GH at a time when most gear was 6+W/GH). Bitmain's second gen S7 (45 chip models, not the 54 chip models) used a regulated string. Novak and I were experimenting with the idea of a regulated string about 9 months before Bitmain brought it to market, which was fun.

[Mcpepe]

Hello, I am fascinated by the project. It's amazing how much progress has been made in such a short time. I would like to help by doing some beta testing. since I do not have the necessary knowledge for the design.

I have some chips from old equipment
BM1397 BM1387

so I could test both the new bitaxe max version and the first bitaxe v2

You have a list of components to purchase. any recommendations to request pcb?

[Skot]

Just spotted this topic fantastic project looking forward to seeing the final product on this.

Also fantastic to see open source project like this!

Keep up the good work!

Hello, I am fascinated by the project. It's amazing how much progress has been made in such a short time. I would like to help by doing some beta testing. since I do not have the necessary knowledge for the design.

I have some chips from old equipment
BM1397 BM1387

so I could test both the new bitaxe max version and the first bitaxe v2

You have a list of components to purchase. any recommendations to request pcb?

Once the design settles down a bit, we'll do an at cost group buy of assembled PCBs so people can start hacking on them. Stay tuned!

Finding a reliable (and affordable) supplier of BM1397AG for this is on the to-do list!

[Mcpepe]

Just spotted this topic fantastic project looking forward to seeing the final product on this.

Also fantastic to see open source project like this!

Keep up the good work!

Hello, I am fascinated by the project. It's amazing how much progress has been made in such a short time. I would like to help by doing some beta testing. since I do not have the necessary knowledge for the design.

I have some chips from old equipment
BM1397 BM1387

so I could test both the new bitaxe max version and the first bitaxe v2

You have a list of components to purchase. any recommendations to request pcb?

Once the design settles down a bit, we'll do an at cost group buy of assembled PCBs so people can start hacking on them. Stay tuned!

Finding a reliable (and affordable) supplier of BM1397AG for this is on the to-do list!

I have some chips from old equipment for test
BM1397 BM1387

[NebulaMiner]

I did some progress on the BM13xx serial protocol understanding.

I just published the Saleae Logic2 High Level Analyzer extension : https://github.com/GPTechinno/bm13xx-hla.
It should be also available in the Saleae Marketplace

More need to be done....

[xraid]

RE:
"Once the design settles down a bit, we'll do an at cost group buy of assembled PCBs so people can start hacking on them. Stay tuned!
Finding a reliable (and affordable) supplier of BM1397AG for this is on the to-do list!"

hi,

regarding recycling BM1397 chips, there is a "lot" for sale of "need repair" s17+ (65 chips per hashboard * 3 hashboards per machine) in Dallas.

If somebody is local and can inspect corrosion of the hashboards maybe we could do a group buy . i can take 10 machines myself (600 in lot).

when sourcing BM1397 AD AG AH AI now they are on offer for 15-27$ each as new and around 6.5$ each as used, need ship from China.

the chips are sorted in production into bin numbers, a hashboard has a label of version(AG AH AI) and bin of chip + a Line number of hashboard, bin numbers are allocated on errors in frequency testing and Line number after assembly testing of hashboard.

so it is best have matching "bin" of chips in a assembled Bitaxe, (bin is not labeled on chip but after in sorting, but can in case of recycle from hashboards be deducted from label.)

i am thinking bid 20USD per machine for all with pickup on site in Dallas, today's price for a s17+ would be around 150-250usd depending connection for a working machine, these are in need of repair and for them its scrap and if somebody take all We could probably bid at 20 ok say 21.

guesstimate 33% of chips not working, would leave about 130 chips to be saved per machine.

if we can not buy the lot of 600 maybe we could give an offer of x machines at price y, count me in for 10.

but first need understand if dampness in farm has corroded the machines so we need a inspection, and after sort out logistics.

[Skot]

hi,

regarding recycling BM1397 chips, there is a "lot" for sale of "need repair" s17+ (65 chips per hashboard * 3 hashboards per machine) in Dallas.

If somebody is local and can inspect corrosion of the hashboards maybe we could do a group buy . i can take 10 machines myself (600 in lot).

when sourcing BM1397 AD AG AH AI now they are on offer for 15-27$ each as new and around 6.5$ each as used, need ship from China.

the chips are sorted in production into bin numbers, a hashboard has a label of version(AG AH AI) and bin of chip + a Line number of hashboard, bin numbers are allocated on errors in frequency testing and Line number after assembly testing of hashboard.

so it is best have matching "bin" of chips in a assembled Bitaxe, (bin is not labeled on chip but after in sorting, but can in case of recycle from hashboards be deducted from label.)

i am thinking bid 20USD per machine for all with pickup on site in Dallas, today's price for a s17+ would be around 150-250usd depending connection for a working machine, these are in need of repair and for them its scrap and if somebody take all We could probably bid at 20 ok say 21.

guesstimate 33% of chips not working, would leave about 130 chips to be saved per machine.

if we can not buy the lot of 600 maybe we could give an offer of x machines at price y, count me in for 10.

but first need understand if dampness in farm has corroded the machines so we need a inspection, and after sort out logistics.

Wow, $20 for a S17+ is a great deal considering a single BM1397 is almost $20.
$12k is pretty reasonable if they're not totally trashed. So, anyone live near Dallas want to go check this out?

[xraid]

also if somebody here is handy with 3d-printing ? well even answer if it could be done with a 3d-print.

i am looking at a BM1397 PCB breakout
https://www.pcbway.com/project/shareproject/Socket_BM1397.html

and if a pressure mount 3d-printed holder for the bm1397-pcb could be made for testing chips without soldering.

[Skot]

also if somebody here is handy with 3d-printing ? well even answer if it could be done with a 3d-print.

i am looking at a BM1397 PCB breakout
https://www.pcbway.com/project/shareproject/Socket_BM1397.html

and if a pressure mount 3d-printed holder for the bm1397-pcb could be made for testing chips without soldering.

They're expensive, but BM1397 ZIF sockets exist; https://www.aliexpress.us/item/3256804282575324.html

there is also a mysterious "tester" https://www.aliexpress.us/item/3256804725472466.html
We could make something like that though.

[Skot]

The bitaxeMAX v2 PCB is coming along! I think I've got most of the placement and routing done.

[n0nce]

also if somebody here is handy with 3d-printing ? well even answer if it could be done with a 3d-print.

I do; could give that thing a try.. However only looks like a breakout board like mine; I don't understand where to get STL files for printing?
There should be some kind of 3D printed piece holding the chip on through pressure. Would be pretty sweet. If someone has a working model, I would print them for free for the sake of Bitaxe development (or other open-source miner evolving from this).

[xraid]

also if somebody here is handy with 3d-printing ? well even answer if it could be done with a 3d-print.

I do; could give that thing a try.. However only looks like a breakout board like mine; I don't understand where to get STL files for printing?
There should be some kind of 3D printed piece holding the chip on through pressure. Would be pretty sweet. If someone has a working model, I would print them for free for the sake of Bitaxe development (or other open-source miner evolving from this).

We would need make a design or 3 -and draw it / test print

-i see a bottom base plate with a 0.8 * 0.8 * depth(?) indentation.

-and a lid where PCB-breakout is mounted, the lid could have a overhang with sides when placed over base, the overhang-sides could have m3 machine screw adjustments ? if the 3d-print could not be made exact, the threads for the adjustment-screws would be made after print.

-no hinges for lid ? , but weight of lid could maybe be sufficient make contact ? or the overhang sides of lid could almost meet the base where magnets could pull.

well any design that works will float my boat ...

then maybe a esp32 with display to drive some test confirming chip is operational

[n0nce]

well any design that works will float my boat ...

If you come up with some 3D / CAD model, feel free to share it! I can definitely do test prints and check tolerances & fitment of the chip; BM1397 should arrive in a couple weeks so I can try all dimensions. Once the lid works, a PCB can be made 'around' the lid design.

[dj789]

That’s so cool you ordered some bitaxe boards! Open source HW for the win! Let me know if you have any issues building them up.

Thank you again for making this possible!  I have all the parts now, but i seem to be getting slightly too much paste with the stencil.  It's probably just my poor technique.  I'll get some practice this weekend, and if it doesn't work out, i have one of those BM1387/97 tin tools on the way.

That's cool! I believe you should be able to test ESP32 code with the bitaxe v2, as well; a big part of the code should be going to work across chips. The whole Stratum connection and serial set-up; not sure whether the message format changed. Skot has looked more into that so far, as far as I know.

I have some ESP32-S3's ordered, so i'm keen to help with the software side.

It's on LCSC though; had to order from 2 places myself, too - that's the current supply chain struggle, I guess... :/

Thanks, i forgot about that site.  I'll wait for a future max/pro board revision before ordering more parts.  The bitaxe v2 will probably keep me busy for a while. 

[Skot]

Thank you again for making this possible!  I have all the parts now, but i seem to be getting slightly too much paste with the stencil.  It's probably just my poor technique.  I'll get some practice this weekend, and if it doesn't work out, i have one of those BM1387/97 tin tools on the way.

It definitely takes some practice. Don’t be afraid to wipe the paste off with some 99% isopropyl alcohol and start again. Clean the stencil too. I use an old toothbrush (and IPA) to clean out the openings.

Are you using a desktop stencil? Use extra PCBs to get the stencil at the right height so it lays flat on the board you are pasting. Tape everything down with masking tape. Make sure your paste is cool. Try and spread it with one pass of your thin metal squeegee at about 45°. I’m sure there are a bunch of YouTube videos too. Check out Greg Davill on twitter for some inspiration. It’s worth taking your time to get the paste right to avoid bridges during reflow. You can fix a lot of problems with a soldering iron and a lot of flux.

[developeralgo]

is anyone having issues getting INA260AIPWR modules from distributors ?  i have a couple of modified BitaxeMax_v2 boards from skot design and a work in progress xBit (Call it BitaxeMax4 with some routing attitude) but getting those modules is hard due to chip shortages. There are some unknown distributors that have them seem shady to me.