Community Miner Design Discussion

[navigatrix]

This project is awesome to hear/read about.

If they really can repurpose old gear and give it new life that would be awesome.

Not sure what the final design will be or how it will be sold but something to think about would be how to sell the units. I know in the automotive world they do "core" charges which means you are charged extra for the part until you bring your old part back in so that it can then be sent off to be reused/refurbished. Perhaps something like that could be done here. charge the customer an extra $xx.xx on top of the miner to cover buying new shells/fans incase they don't return the "core". This would allow those that have these older units laying around but may not have the skill/ability to take apart/rebuild a miner an option to purchase a plug and play miner. The other option would be to just sell the blades for each model and let the owner do the retro fit.

Cant wait to see what comes of this venture.

Great thread! I am feeling optimistic that I may be able to extend my mining efforts far longer than anticipated.

Are you thinking about upgrading folks' existing S1/S3/S5 boards as well as selling new PnP boards?
Either way, I'm eagerly anticipating more news.

[kilo17]

This project is awesome to hear/read about.

If they really can repurpose old gear and give it new life that would be awesome.

Not sure what the final design will be or how it will be sold but something to think about would be how to sell the units. I know in the automotive world they do "core" charges which means you are charged extra for the part until you bring your old part back in so that it can then be sent off to be reused/refurbished. Perhaps something like that could be done here. charge the customer an extra $xx.xx on top of the miner to cover buying new shells/fans incase they don't return the "core". This would allow those that have these older units laying around but may not have the skill/ability to take apart/rebuild a miner an option to purchase a plug and play miner. The other option would be to just sell the blades for each model and let the owner do the retro fit.

Cant wait to see what comes of this venture.

Great thread! I am feeling optimistic that I may be able to extend my mining efforts far longer than anticipated.

Are you thinking about upgrading folks' existing S1/S3/S5 boards as well as selling new PnP boards?
Either way, I'm eagerly anticipating more news.

I know that exchanging or "core" charges may complicate a project that aims to be very simplistic and straight forward.  It may be possible to have a "complete" miner and a blade for the DIY guy but that will depend a lot on the final design from Sidehack. 

[spiccioli]

I know that exchanging or "core" charges may complicate a project that aims to be very simplistic and straight forward.  It may be possible to have a "complete" miner and a blade for the DIY guy but that will depend a lot on the final design from Sidehack. 

Having stackable boards requiring just a raspi to run or boards you can put in a wooden frame like in the good old days of GPUs would be a plus, IMHO.

Or something like the m/h boards of the first bitfury rigs; maybe they can run with just a fan blowing on them.

spiccioli

[hurricandave]

   I think BTCGarden had the right idea of chaining the boards together too a single point on a Pi. But that cable to the Pi was to short and the jumper cables aren't exactly off the shelf common. For some reason, Bitmain seems married to the BeagleBone as the custom base if its controllers, they even have them posted for sale . That shows you, its at least possible to mainstream the controls. EXCEPT. both of those company's as well as many others have continued to use the GPIO/SPI whatever headers.
   I'd love to see a board that can interface with a common controller(PC,SBC) with common cable or chain to the next board using the same cable. But, I hate the headless/browser based GUI's Bitmain and the like have deployed.
   Can we get the Power safely separated so that it DOES NOT matter if the controller is powered first or the hashboard is.
   I;m using whatever power is left over/available, in my bedroom, so a little less than 20A/110v draw is a ceiling I won't overcome anytime soon. As well as a $1k entry point, let us have the boards individually as we can afford them, I'd assuredly spend more than a $1k over several months time. Win Win for us all.

[Witrebel]

As I understand it, the goal is to use USB as the interface, that way you can use whatever controller option suits you provided it can run cgminer.  Quite flexible.

[2112]

As I understand it, the goal is to use USB as the interface, that way you can use whatever controller option suits you provided it can run cgminer.  Quite flexible.

In the ideal world some multipoint interface would be better than USB, which is point-to-point only. So less cable mess.

https://en.wikipedia.org/wiki/CAN_bus
https://en.wikipedia.org/wiki/Local_Interconnect_Network
https://en.wikipedia.org/wiki/Vehicle_Area_Network

https://en.wikipedia.org/wiki/RS-485

The problem with all of them is that they aren't common peripheral interfaces for household computers. They are used in automotive or general industrial control applications. So very few people would understand them and many people would be simply afraid.

https://en.wikipedia.org/wiki/List_of_automation_protocols

USB wins because although it is a devil, it is the devil everyone is familiar with.

[hurricandave]

Maybe it would make sense too consider using a Cat6 connection when chaining boards and the final output would be sent thru USB to the controller or hub but if some one had a dual NIC motherboard, then that could be used instead.

[philipma1957]

Maybe it would make sense too consider using a Cat6 connection when chaining boards and the final output would be sent thru USB to the controller or hub but if some one had a dual NIC motherboard, then that could be used instead.

Okay am I missing something.

a simple cheap usb 2 hub like so

http://www.amazon.com/Sabrent-Adapter-Control-Switches-HB-U14P/dp/B00HL7Z46K/ref=pd_sim_sbs_147_1?

and a simple cable tie

http://www.amazon.com/Case-Logic-Attaching-Assorted-Colors/dp/B00004TZF9/ref=sr_1_1?

and you have 13 boards to 1 pc.

If that proves problematic   due to too many board on 1 hub then two of these

http://www.amazon.com/AmazonBasics-Port-USB-Power-Adapter/dp/B00DQFGJR4/ref=sr_1_3?

with the same ties and you have 14 boards.

[sidehack]

But then you'd be building an ethernet controller or dual USB hub onto each board. Or some other setup requiring either an additional adapter or adaption built onto each board. The point is to do exactly not that. I know USB isn't great, but it's flexible and it's standard and it's freakin' everywhere. So if I want to use this framework to build a rack machine or a stickminer, I can.

Trust me on one thing - you're not going to change my mind. I've heard all the reasons for the last year.

[2112]

Maybe it would make sense too consider using a Cat6 connection when chaining boards and the final output would be sent thru USB to the controller or hub but if some one had a dual NIC motherboard, then that could be used instead.

Ethernet is too complex and too expensive for this. Both on the hardware side (PHY) as well as on the complexity and requirements of the software stack.

The other problem with Ethernet is that it has minimum cable length requirements to work reliably. If you violate this constraint you'll incur too much errors.

Finally, it is just tremendous waste of bandwidth: Ethernet is designed for gigabit per second transmission rates.

All in all, Ethernet is such a bad choice that even the plastic TOSLINK fiberoptics would be better.

[2112]

Okay am I missing something.

a simple cheap usb 2 hub like so

http://www.amazon.com/Sabrent-Adapter-Control-Switches-HB-U14P/dp/B00HL7Z46K/ref=pd_sim_sbs_147_1?

and a simple cable tie

http://www.amazon.com/Case-Logic-Attaching-Assorted-Colors/dp/B00004TZF9/ref=sr_1_1?

and you have 13 boards to 1 pc.

If that proves problematic   due to too many board on 1 hub then two of these

http://www.amazon.com/AmazonBasics-Port-USB-Power-Adapter/dp/B00DQFGJR4/ref=sr_1_3?

with the same ties and you have 14 boards.

You missing the number of separate connectors.

Point-to-point connection with 12 devices through a hub:

13 cables + 1 hub + 26 connector pairs

Multipoint connection with 12 devices through a bus:

1 cable + 13 connector pairs + 2 terminators (at each end of the cable, but frequently devices have built-in terminators)

The problem is that you wouldn't be able to simply buy such a 13-tap cable at Amazon or in a supermarket. You would have to shop in e.g. car parts store, buy separately wire, crimp-on connectors, crimping tool, etc. It isn't rocket science, any car electrician could help you, but this would scare most of the potential buyers.

Edit: I know that sidehack & friends are fans of Atmel megaAVR. So I just did a quick price comparison for the top-of-the-line USB and CAN parts from Atmel:

AT90CAN128-16AU $6.44
AT90USB1287-AU $7.51

The price difference doesn't matter. It is the familiarity that counts.

By the way: Atmel was acquired by Microchip this weekend.

[sidehack]

I think the part Novak spec'd is an NXP chip.

[2112]

I think the part Novak spec'd is an NXP chip.

Could you guys disclose the shortlist? I'm just curious, I hope this isn't some important trade secret. Could you just list say a few top parts that you've considered? If it is a trade secret then include some misleading parts and sort the list randomly.

[sidehack]

I don't know offhand but I'll look it up this afternoon. If it was a trade secret that'd make the "open source" part pretty difficult.

[philipma1957]

But then you'd be building an ethernet controller or dual USB hub onto each board. Or some other setup requiring either an additional adapter or adaption built onto each board. The point is to do exactly not that. I know USB isn't great, but it's flexible and it's standard and it's freakin' everywhere. So if I want to use this framework to build a rack machine or a stickminer, I can.

Trust me on one thing - you're not going to change my mind. I've heard all the reasons for the last year.

So usb2 will be the standard correct?

Btw I would prefer it. Since I am pretty sure I can mange ten to thirty five boards via a pc

[cavaliersrus]

what about the ability to run them with a nic on each miner

or run 1 nic to a switch then the others would connect via usb to miner 1 and make a usb hub style where they could daisy chain down the line with less clutter of going mass into a switch with alot of cat 5 cables ? just a idea i had kinda like the old firewire days where you could keep chaining devices down the line till u needed another hub to boost the power

[Mitak]

But then you'd be building an ethernet controller or dual USB hub onto each board. Or some other setup requiring either an additional adapter or adaption built onto each board. The point is to do exactly not that. I know USB isn't great, but it's flexible and it's standard and it's freakin' everywhere. So if I want to use this framework to build a rack machine or a stickminer, I can.

Trust me on one thing - you're not going to change my mind. I've heard all the reasons for the last year.

totally agree.
As much I hate USB downsides, this is the cheapest and flexible solution

[Mitak]

But then you'd be building an ethernet controller or dual USB hub onto each board. Or some other setup requiring either an additional adapter or adaption built onto each board. The point is to do exactly not that. I know USB isn't great, but it's flexible and it's standard and it's freakin' everywhere. So if I want to use this framework to build a rack machine or a stickminer, I can.

Trust me on one thing - you're not going to change my mind. I've heard all the reasons for the last year.

So usb2 will be the standard correct?

Btw I would prefer it. Since I am pretty sure I can mange ten to thirty five boards via a pc

Depends
stack overflow , sound familiar

[HagssFIN]

So with USB it is going to be handy to control one or several units for example with Raspberry Pi?

[2112]

what about the ability to run them with a nic on each miner

or run 1 nic to a switch then the others would connect via usb to miner 1 and make a usb hub style where they could daisy chain down the line with less clutter of going mass into a switch with alot of cat 5 cables ? just a idea i had kinda like the old firewire days where you could keep chaining devices down the line till u needed another hub to boost the power

I would like to see the actual error statistics from those daisy-chained Firewires. I saw some stats from high-end MacIntoshes driving stacks of 5-6 external disks each through Firewire 400. This wasn't anything good and they did run in very clean offices not garages, like the mining farms.