GekkoScience BM1384 Project Development Discussion

[toptek]

So it looks like the fix is consistent. I reseated chips on a couple failed previous stick assemblies, applied the fixes and they both work now. It's almost cheeseburger time (and other errands, mostly spending money on awesome stuff...) but tomorrow I'll be in and build some more sticks. We're running low on sample chips after all the prototype testing of various things, so I need to save the rest for the 18-board test setup which means I'll be scavenging chips off dead S5 boards for the sticks. Should be fun? And then I have to build heatsinks for them all, which means cutting and drilling and tapping.

But maybe tomorrow end of the day I'll have a hub full of sticks all blasting away I can post a picture of. Be pretty sexy.

Guess you didn't see my PM yet i hope that's it just need a address I'll send off one dead S5 board in the next few days . and the other one once i get it .

[sidehack]

Not a bad nine-hour day today. Actually, no it was a pretty lousy nine-hour day today. End of yesterday I had three working sticks. End of today I have four working sticks. I started building five and got two of them going, but my original stick decided to throw power issues when I plugged it back in at the end of the day. The three other non-working sticks are also giving power issues. Probably nothing too substantial, maybe pull and reseat the buck driver. I wasted a lot of time with pulled chips; I think the whole board I started pulling from must be dead because not a single stick lit up with those chips but they started behaving a lot better when I pulled chips off a previously tested (mostly) functional S5 board. One of the primary problems I was seeing (besides "whoops there's no Vcore power) is UART stuff, and after the ASIC itself one of the most finicky chips is this stupid little SC70 level shifter that's never really worked right and it's really hard to put on proper and the thing's just stupid and I hate it. But the good news is that won't be on the next version of the PCB because after learning a bit more about chip functions and such, I figured out I can actually replace that part with a couple passives and some clever wiring and it's cheaper and better and also way less annoying. The last stick I got working today was giving me worse fits than usual with this part, so I chucked it and deadbugged the new circuit onto the board and it's currently hashing away right now.

Maybe tomorrow I'll diagnose the problems with the power on the other sticks and get some more working. I need to have a few to show off, you understand. And maybe one for Novak to gank so he can work on the code.

Sample heatsinks are on the way which I'll use on the next-rev PCBs. They're not the right color but should be mechanically as-designed so I'll need to test fit those and make sure they're right when they come in next week.

I also need to figure out what's the holdup compiling cgminer on my workbench machine. The OS has never worked satisfactorily and I need to reload the thing anyway, but right now I'm stuck testing with a dodgy XP laptop that doesn't like more than one stick at a time running without errors. So y'all biscuitheads will have to wait for the latest and greatest in stickminer orgy porn until I can get some more heatsinks worked up and shift my test hub over to something that'll support five or six sticks at once. Which actually won't be that hard since the test hub already has a molex wired onto it and the fileserver has cgminer running with one of our two-chip test setups so I'll probably just use that. On another day. Right now it's about time to pass out.

[ManeBjorn]

Just let me know when and I will buy a couple from you and give you feedback.
I am still happy you are doing this.

Not a bad nine-hour day today. Actually, no it was a pretty lousy nine-hour day today. End of yesterday I had three working sticks. End of today I have four working sticks. I started building five and got two of them going, but my original stick decided to throw power issues when I plugged it back in at the end of the day. The three other non-working sticks are also giving power issues. Probably nothing too substantial, maybe pull and reseat the buck driver. I wasted a lot of time with pulled chips; I think the whole board I started pulling from must be dead because not a single stick lit up with those chips but they started behaving a lot better when I pulled chips off a previously tested (mostly) functional S5 board. One of the primary problems I was seeing (besides "whoops there's no Vcore power) is UART stuff, and after the ASIC itself one of the most finicky chips is this stupid little SC70 level shifter that's never really worked right and it's really hard to put on proper and the thing's just stupid and I hate it. But the good news is that won't be on the next version of the PCB because after learning a bit more about chip functions and such, I figured out I can actually replace that part with a couple passives and some clever wiring and it's cheaper and better and also way less annoying. The last stick I got working today was giving me worse fits than usual with this part, so I chucked it and deadbugged the new circuit onto the board and it's currently hashing away right now.

Maybe tomorrow I'll diagnose the problems with the power on the other sticks and get some more working. I need to have a few to show off, you understand. And maybe one for Novak to gank so he can work on the code.

Sample heatsinks are on the way which I'll use on the next-rev PCBs. They're not the right color but should be mechanically as-designed so I'll need to test fit those and make sure they're right when they come in next week.

I also need to figure out what's the holdup compiling cgminer on my workbench machine. The OS has never worked satisfactorily and I need to reload the thing anyway, but right now I'm stuck testing with a dodgy XP laptop that doesn't like more than one stick at a time running without errors. So y'all biscuitheads will have to wait for the latest and greatest in stickminer orgy porn until I can get some more heatsinks worked up and shift my test hub over to something that'll support five or six sticks at once. Which actually won't be that hard since the test hub already has a molex wired onto it and the fileserver has cgminer running with one of our two-chip test setups so I'll probably just use that. On another day. Right now it's about time to pass out.

[sidehack]

Bitmain just let me know they're checking stock on chips, so hopefully they have enough that we can move forward on the build. Novak's talking to a new US-based proto-etch place we'd like to have run out the next Compac PCB, but in the meantime I reckon I'll be working on getting the last batch assembled and modified, if only to have something to play with. That and all the other stuff Novak and I have to do will be keeping us pretty busy.

[aarons6]

Bitmain just let me know they're checking stock on chips, so hopefully they have enough that we can move forward on the build. Novak's talking to a new US-based proto-etch place we'd like to have run out the next Compac PCB, but in the meantime I reckon I'll be working on getting the last batch assembled and modified, if only to have something to play with. That and all the other stuff Novak and I have to do will be keeping us pretty busy.

this is the news i been waiting for.. i cant wait.

i been watching the solar powered Raspberry Pi thread where the guy is running a couple u2s from a camping solar panel and a cell phone battery pack...

i think if i get a couple of these i will try the same with them.

[philipma1957]

Bitmain just let me know they're checking stock on chips, so hopefully they have enough that we can move forward on the build. Novak's talking to a new US-based proto-etch place we'd like to have run out the next Compac PCB, but in the meantime I reckon I'll be working on getting the last batch assembled and modified, if only to have something to play with. That and all the other stuff Novak and I have to do will be keeping us pretty busy.

 Nice looking forward to this.

[ManeBjorn]

That is very good news indeed.

Bitmain just let me know they're checking stock on chips, so hopefully they have enough that we can move forward on the build. Novak's talking to a new US-based proto-etch place we'd like to have run out the next Compac PCB, but in the meantime I reckon I'll be working on getting the last batch assembled and modified, if only to have something to play with. That and all the other stuff Novak and I have to do will be keeping us pretty busy.

[Ikron]

Can't wait to see what the finished boards are. More greatness pls!

[cavaliersrus]

im enjoying the updates there side and novak cant wait to invest some into this

[PlanetCrypto]

Bitmain just let me know they're checking stock on chips, so hopefully they have enough that we can move forward on the build. Novak's talking to a new US-based proto-etch place we'd like to have run out the next Compac PCB, but in the meantime I reckon I'll be working on getting the last batch assembled and modified, if only to have something to play with. That and all the other stuff Novak and I have to do will be keeping us pretty busy.

We'll take a couple, either proto or production.
Please indicate how much and where to send the BTC.
Thanks.

[sidehack]

So I ironed out the issues with two of the three other sticks I put together on Saturday. One still has no power, which is annoying but Novak reminded me the boards weren't electrical-tested so it could be a bad trace somewhere.

Anyway I have five of the six working sticks all running off a Pi right now, pointed at the Burger address. The sixth was throwing hardware errors so I need to figure out what's up; could be as simple as a poor heatsink fit. Which I sorta believe since, as you can see, most of the heatsinks were cut from something else. These five are putting up 41GH off about 3.5A (it was right around 3A with just four sticks and the Pi).

[Mikestang]

Beautiful!  I can't wait to convert my S1 over, keep it up!

The sticks do OK with just the passive heat sink cooling, no additional airflow needed?

[sidehack]

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

[valkir]

  so beautiful!

Congratz Sidehack and Novak!  

You guys did what big company didnt do! You are awesome.

[sidehack]

No, "awesome" would have been if it was working a month ago. That said, the whole R&D budget so far is on par with the cost of a pair of S5, so that's probably alright.

[philipma1957]

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

If can image the sdcard and send a copy to me so

I can run it on a rasp pi.  I would copy sdcards for rasp pi's. offering a sdcard /hub/stick package for a new buyer may be a good selling idea.

A real plug and play for the masses.

[ManeBjorn]

I agree this would be a great idea.
+1

Right now they're drawing less than 3W (probably less than 2.7, really) at 8.25GH per stick. The heatsink gets toasty in still air (especially the already-warm air in the shop) but it's not danger-zone hot for sure. If you want to push them up past probably 10GH (estimated 3.1W or so) I'd recommend air. I've tested the sticks for starting up to 250MHz (~14GH, estimated 5.5W) but I didn't run them for endurance so I don't know how well the stock heatsink with airflow works up that high. Yet.

Proto PCBs have been ordered for a chips-only 18-board (basically, a passive TypeZero Spec1, does not have power or control on board) and for the revised (version 0.4) Compac PCB which implements the changes to the previous PCB that got these sticks working. Technically only the silver-heatsink one has all the changes, as it's the one with the deadbugged modified level shifter circuit. The rest are still using the old circuit, which has both more complexity and worse performance.

Also, to note, three of the five sticks there are using chips pulled from one of those donated dead S5 miners from a while back. The sixth stick (what with the hardware errors) also has a pulled chip, which might explain a bit.

I'll probably be sending some of these guys out to my first-wave testers, maybe tomorrow if they run overnight without issue. In a week or two when I have the sample heatsinks and fresh PCBs in hand I'll build what I hope is the final version and get them out. If we have good word from Bitmain by then about fetching chips, and the boards test out well (I'll be looking forward to a review thread, whether it's good news or bad news I like having honest feedback), maybe we can start taking orders. We'll have a pick-and-place and a nice new SMD oven up and going here soon so it shouldn't take forever+day to assemble these things.

If can image the sdcard and send a copy to me so

I can run it on a rasp pi.  I would copy sdcards for rasp pi's. offering a sdcard /hub/stick package for a new buyer may be a good selling idea.

A real plug and play for the masses.

[richardamullens]

Maybe if you push the raspberry pi aspect of the project it could gain traction among users of the Pi and Pi 2.

Certainly if you are able to start making the stick miners I would use a Pi.  the B+ and the Pi 2 use less power than the original machine.

I guess that is a stripped down USB hub you have there on the wooden block.

I hardly ever bother looking at any other threads on bitcointalk now

[philipma1957]

Maybe if you push the raspberry pi aspect of the project it could gain traction among users of the Pi and Pi 2.

Certainly if you are able to start making the stick miners I would use a Pi.  the B+ and the Pi 2 use less power than the original machine.

I guess that is a stripped down USB hub you have there on the wooden block.

I hardly ever bother looking at any other threads on bitcointalk now

I have a rasp pi b  older but not oldest model
I have a rasp pi 2 newest model.

The idea of 3-5 sticks on a hub to a rasp pi. power cost lets say 5 x 3.1 = 15.5 watts  the rasp pi + hub total watt amount = 25 watts
 5 sticks = 40gh  if software allows 5 pools  32gh can mine on slush or btcguld or f2pool  the other 8gh can solo mine.

gear is so  efficient it comes close to breaking even while giving you a shot at every block via the 1 stick on solo pool.

I love the idea of 100 usd to 125 usd out of pocket that pays itself off while allowing me a shot at each and every block.

or mine all your sticks at a 'real' and actually get to break even in 150 days to 180 days.

[sidehack]

Right now it's a stock Minera 0.5.0 image (or whatever's the version number for the current build) running stock cgminer 4.9.1 and enumerating the sticks as U3. That's suboptimal, of course, but it works well enough for a single chip. The only real annoyance is the stock U3 driver has a limited selection of frequencies to choose from, and not much granularity in the low end where the stick is likely to operate. Oh yeah and the device reset timeout is too low because it's expecting five times the hashrate, but that doesn't have much effect on performance.

For now I'm just gonna hand out Compacs and let people test with stock cgminer as U3. By the time we get to Amita work, hopefully Novak has some sort of driver figured out to get two chips working, but since I won't have an Amita PCB to test until after the final Compac PCB is tested we do have several weeks to twiddle the bits, and the two-chip board on the USB-UART adapter we've had running for a while now is a pretty accurate testbench to code with.

Yeah, that old hub was stripped down and beefed up for overclocked Block Erupters back in the day. I've got extra capacitors on it and starred power wires dedicated per socket. I trust it up to at least 5A, possibly more if I use the Molex instead of the pin header power.

Also, regarding price, I need to go back over my BOM and make some changes to reflect the new design. Odds are it won't change much, since some of the parts I'm knocking off should pretty well offset the parts I have to add.

Thing about doing that stuff you just said, Phil, is that for the price of five stick miners and a hub you could probably fetch a Spec1 board, basic heatsink and a fan and have something drawing four times the hashrate at a comparable efficiency. But it would require more power and probably be more loud.