DIY PCB with AVALON: "The Quarter Stick" - Needs Help!

[Gomeler]

Core voltage is 1.2V, not 120V.

Right but I did not state the core voltage was 120V. The stated power consumption of an Avalon is 620w @ 120v which is what I used in my estimate. If each chip does indeed use 1.5A each then an Avalon would use more than 43kW so that can't be right. If the core voltage is 1.2V then each chip uses ~24mW by my estimates.

Power consumption = amps * volts. Avalon's power consumption at the wall, through the PSU, is ~620w @ 120v, so 5.16 amps @ 120v. The PSU converts that down to +12vdc to feed each of the three modules, figuring ~80% efficiency let's just call that 500w of +12vdc. This is 41.6 amps of +12vdc. Now each module has groups of chips fed by a +12vdc to +1.2vdc converter to feed the cores. ~500w @ +12vdc divided by 240 chips = ~2w per chip BEFORE the losses of the +12vdc to +1.2vdc converter. Those are usually incredibly efficient, let's call it 95%, so 475w of +1.2vdc, 395 amps.

I'm sure I'm butchering all the math here with calculating PSU efficiencies and such but this shows 1.979 watts @ +1.2vdc or 1.64 amps per chip.

I hope this makes sense to you. With these numbers, 3 of their chips, which provides (63,000/240)*3=786MH/s could be powered by a USB port providing 0.5 amps @ +12vdc. Call it 2 chips to give a nice buffer under the 0.5 amp limit and you have a neat little device. It won't generate crap for BTC though.. better off targeting something in the 5-10 GH/s range.

The information i found stated 450W for 3 modules not including PSU. USB is 5VDC nominal +/- 5% limiting the device to 2.5W total. We'd need around 300 - 500mW for micro and clock generation. The rest of your math looks spot on.

So far I'm up to to around US$7 in parts, not including the Avalon chip, with a few technical assumptions. Assuming we'd get the avalon for $2 - $5, a unit price for a usb miner under $15 is realistic. For the next generation, hopefully they include a PLL on die!

Sam.

*doh* for some reason I thought usb was +12vdc. Oops.

[samborambo]

So once there is a ready made plan for this how will this made going? Groupbuys for the chips and everyone buys the remaining hardware and gets the plans and software for free or how can this work/is planned?

Kickstarter.com?

[SebastianJu]

So once there is a ready made plan for this how will this made going? Groupbuys for the chips and everyone buys the remaining hardware and gets the plans and software for free or how can this work/is planned?

Kickstarter.com?

That would be for collecting money. Is that needed here? Buying specialist maybe?

[samborambo]

So once there is a ready made plan for this how will this made going? Groupbuys for the chips and everyone buys the remaining hardware and gets the plans and software for free or how can this work/is planned?

Kickstarter.com?

That would be for collecting money. Is that needed here? Buying specialist maybe?

We need to secure a lot of capital for the first batch of 10k units. KS is a good way to raise capital while protecting investors if things go pear-shaped before we've got all supply contracts in place.

[peterepeat]

So once there is a ready made plan for this how will this made going? Groupbuys for the chips and everyone buys the remaining hardware and gets the plans and software for free or how can this work/is planned?

Kickstarter.com?

That would be for collecting money. Is that needed here? Buying specialist maybe?

We need to secure a lot of capital for the first batch of 10k units. KS is a good way to raise capital while protecting investors if things go pear-shaped before we've got all supply contracts in place.

How about on bitfunder then? Keeps everything in BTC - Im sure you can raise the ~3000-3500 BTC (guessing) for the 10000 chips and then start the USB mini hashing board develoment   
perhaps an arduino BTC shield project

[rgzen]

interested

[SebastianJu]

How about on bitfunder then? Keeps everything in BTC - Im sure you can raise the ~3000-3500 BTC (guessing) for the 10000 chips and then start the USB mini hashing board develoment   
perhaps an arduino BTC shield project

I think the bitcoin community is able to provide enough money too. But will it be so few bitcoins only? I mean avalon is in a conflict here. I know the owner is an idealist that wants to help bitcoin but he wants to earn money too. So if he gives away so many chips for a low price he is killing his business. So it will be a fight in their heads.
It would be helpful to know a real number for the price. Otherwise it could become a bad surprise because the only thing the community cant build easily is the asic-design. Thats the thing avalon still must sell and would make the money with. So knowing the price is of the essence.

Another interesting thing is that the chips-selling is beside the selling of asic-miners, because they dont have work with it. They only need to order it and when its there they can forward it. No big waiting time and uninfluenced by the miners.

Who would get this huge batch of chips? I mean he must be very trustworthy. Its not only a big investment but it will be an even bigger moneymaker when built into a miner. So the one(s) that gets the chips sent to have to be real trustworthy.

[peterepeat]

How about on bitfunder then? Keeps everything in BTC - Im sure you can raise the ~3000-3500 BTC (guessing) for the 10000 chips and then start the USB mini hashing board develoment   
perhaps an arduino BTC shield project

I think the bitcoin community is able to provide enough money too. But will it be so few bitcoins only? I mean avalon is in a conflict here. I know the owner is an idealist that wants to help bitcoin but he wants to earn money too. So if he gives away so many chips for a low price he is killing his business.

Well if they don't want do this, perhaps friedcat & asicminer could be interested to sell their chips. Their company is planning to sell devices also, and maybe this brings it forward for them.
I actually think selling chips as well as finished product can work together for Avalon and/or asicminer, as it's a faster, easier OEM sale which can be in parallel with their retail business. Their business is I imagine to sell as many of these chips as they can, for the highest profit.
(Edit, phone post, needed rework)

[SebastianJu]

Well if they don't want do this, perhaps friedcat & asicminer could be interested to sell their chips. Their company is planning to sell devices also, and maybe this brings it forward for them.

I highly doubt because asicminer was created to earn money for bitfountain and the shareholders of asicminer. Selling the chips would only be possible for a high price then. And asicminer only starts to sell asics because they have so many problems with the needed power supply and the network. So in order to not have the asics laying around and lose value they will sell it. And via auction to get out the maximum in price.
I think the best bet is to set on avalon... or wait... whats with bfl? They have the chips already, they claim they are hashing but they cant manage to put it together. Maybe the community can?
I think the best way would be to ask all three developers if they would sell chips and what their price will be. Judging from the base of thinking of their bosses i believe avalon holds the highest bids that they sell. Asicminer and BFL are more for earning money while the boss of avalon is more about network security.

[peterepeat]

Well if they don't want do this, perhaps friedcat & asicminer could be interested to sell their chips. Their company is planning to sell devices also, and maybe this brings it forward for them.

I highly doubt because asicminer was created to earn money for bitfountain and the shareholders of asicminer. Selling the chips would only be possible for a high price then. And asicminer only starts to sell asics because they have so many problems with the needed power supply and the network. So in order to not have the asics laying around and lose value they will sell it. And via auction to get out the maximum in price.
I think the best bet is to set on avalon... or wait... whats with bfl? They have the chips already, they claim they are hashing but they cant manage to put it together. Maybe the community can?
I think the best way would be to ask all three developers if they would sell chips and what their price will be. Judging from the base of thinking of their bosses i believe avalon holds the highest bids that they sell. Asicminer and BFL are more for earning money while the boss of avalon is more about network security.

I hear what you're saying, especially as ASICMINER have stated they want to scale and maintain network share. But when they are hashing @ 250TH, I don't think selling 10000 chips hurts them at all. If anything, it world be good PR for them.
BFL won't want someone else shipping their asics before they do.
Eventually, I can see all three wanting to sell to OEM, but perhaps there is too much profit on keeping supply tight at this phase.

[SebastianJu]

I now asked friedcat about selling chips and he answered like i thought he will. Creating the miners can be done cheap, mining is most profitable now and selling miners only is done because of the problems and remaining flexible. So chips alone wont be sold. As a shareholder i think its a logical decision.
I asked avalon for the price and bfl if they will sell but have to wait for an answer yet.

[papamoi]

I now asked friedcat about selling chips and he answered like i thought he will. Creating the miners can be done cheap, mining is most profitable now and selling miners only is done because of the problems and remaining flexible. So chips alone wont be sold. As a shareholder i think its a logical decision.
I asked avalon for the price and bfl if they will sell but have to wait for an answer yet.

i ve asked him the same he never replied me
nor avalon

[tbd]

A lot of great feedback so far.

I think this is best kept as a simple device.  It's not going to be a "super miner", but something for the hobbyist and to hopefully introduce others to the Bitcoin world.  The more people involved, the better for everyone.

Keeping this in mind, perhaps we can reach a consensus on the design parameters:
 - Powered via the USB port
 - Passive cooling
 - Price target <$20

This will, of course, limit the hash rate.  But, when compared to current FPGA or GPU options, you're getting great performance on both a $ and watt basis.

[SebastianJu]

A lot of great feedback so far.

I think this is best kept as a simple device.  It's not going to be a "super miner", but something for the hobbyist and to hopefully introduce others to the Bitcoin world.  The more people involved, the better for everyone.

Keeping this in mind, perhaps we can reach a consensus on the design parameters:
 - Powered via the USB port
 - Passive cooling
 - Price target <$20

This will, of course, limit the hash rate.  But, when compared to current FPGA or GPU options, you're getting great performance on both a $ and watt basis.

Is passive cooling reachable anyway? Plus i would like it more as a modular construction system so that it can be extended like everyone likes. In form of adding more chips on one pcb or adding more than one pcb. Extendable...

i ve asked him the same he never replied me
nor avalon

It happens sometimes with friedcat that he forgets a pm. Im not sure but i write with him a bit often, maybe thats why he answered fast.

[tbd]

Is passive cooling reachable anyway? Plus i would like it more as a modular construction system so that it can be extended like everyone likes. In form of adding more chips on one pcb or adding more than one pcb. Extendable...

That starts to sound remarkably like an Avalon.  It's modular: 10 chips on a board, 8 boards to a card, 3-4 cards to a machine.  They're already building and selling that, so no need to reinvent the wheel.

This should address a different market segment -- simple, small, low cost, for the casual/curious end user.

[SebastianJu]

Is passive cooling reachable anyway? Plus i would like it more as a modular construction system so that it can be extended like everyone likes. In form of adding more chips on one pcb or adding more than one pcb. Extendable...

That starts to sound remarkably like an Avalon.  It's modular: 10 chips on a board, 8 boards to a card, 3-4 cards to a machine.  They're already building and selling that, so no need to reinvent the wheel.

This should address a different market segment -- simple, small, low cost, for the casual/curious end user.

But you know how it will turn out... people will buy hundreds of chips and build hundreds of machines to get a bigger hashrate. Or do you mean that this shouldnt happen and that it should be somehow restricted per user?

[allten]

Is passive cooling reachable anyway? Plus i would like it more as a modular construction system so that it can be extended like everyone likes. In form of adding more chips on one pcb or adding more than one pcb. Extendable...

That starts to sound remarkably like an Avalon.  It's modular: 10 chips on a board, 8 boards to a card, 3-4 cards to a machine.  They're already building and selling that, so no need to reinvent the wheel.

This should address a different market segment -- simple, small, low cost, for the casual/curious end user.

+1

That segment is currently open for any takers. I think 1000 people having 1 small unit is better than 1 guy having 1000-small-units-strong machine.

Thanks tbd and SebastianJu!

You are exactly right! I'm going to simplify the design further!

[allten]

From the various specs i found floating around, core is 256MHz, 1.15 - 1.2v.  I/O is 8 lines at 3.3v. I worked out the chip would consume around 1700mW. That'll be at the core voltage so 1.5A per chip should do it. Easily within USB power limits even with a typical DC-DC stage.

Unless we can find a micro that can output a 256MHz secondary oscillator, we need an external oscillator block (if the core - i/o is asynchronous) or a PLL to multiply the micro oscillator.

I'm not a fan of making a through-hole soldering kit. Way too geeky and more expensive than a fully built and tested SMT device.  While work with 250MHz, physical wiring tolerances are a concern. A large blob of solder in the wrong place may be OK at DC but at 250MHz, couples signals to other tracks. Happy to help with an SMT design.

I can't be of much help right now. On vacation with just a tablet!

Sam.

I was hoping that the Avalon chips already had a PLL on board (aka clock multiplier for others reading) so we would only need to input 8Mhz or 16Mhz; however, they have not got back to my email or posts to confirm.

I see what you are saying on the PCB. I was thinking about doing two versions. The geeky hole through for now and for development. and one that is SMT shortly after. Since PCB prototyping is my specialty, I had planned on doing this why programmers got the firmware and software ready to go.

[marra]

BitSyncom,

Have fun working on documentation, it sure is fun, fun, fun...

I'm willing to voluntarily contribute if you require additional help, ie, building a website with schematics and diy flow?

 

[allten]

Core voltage is 1.2V, not 120V.

Right but I did not state the core voltage was 120V. The stated power consumption of an Avalon is 620w @ 120v which is what I used in my estimate. If each chip does indeed use 1.5A each then an Avalon would use more than 43kW so that can't be right. If the core voltage is 1.2V then each chip uses ~24mW by my estimates.

Power consumption = amps * volts. Avalon's power consumption at the wall, through the PSU, is ~620w @ 120v, so 5.16 amps @ 120v. The PSU converts that down to +12vdc to feed each of the three modules, figuring ~80% efficiency let's just call that 500w of +12vdc. This is 41.6 amps of +12vdc. Now each module has groups of chips fed by a +12vdc to +1.2vdc converter to feed the cores. ~500w @ +12vdc divided by 240 chips = ~2w per chip BEFORE the losses of the +12vdc to +1.2vdc converter. Those are usually incredibly efficient, let's call it 95%, so 475w of +1.2vdc, 395 amps.

I'm sure I'm butchering all the math here with calculating PSU efficiencies and such but this shows 1.979 watts @ +1.2vdc or 1.64 amps per chip.

I hope this makes sense to you. With these numbers, 3 of their chips, which provides (63,000/240)*3=786MH/s could be powered by a USB port providing 0.5 amps @ +12vdc. Call it 2 chips to give a nice buffer under the 0.5 amp limit and you have a neat little device. It won't generate crap for BTC though.. better off targeting something in the 5-10 GH/s range.

Thanks! Been meaning to get around to figuring out the math on this.

I'll make the PCB with room for 3 chips (the third one can be optional). Either you can run the two chips to the max or settle for running all three at slightly slower speed, but with more overall hashing speed. Using 3 and running them slower might be a good option for passive cooling too (decided to go with passive cooling).

You are right, this does not produce crap for the serious miner. But IMO the project will do few things:

1) Help the new curious miner get into bitcoin at a very low cost yet still be somewhat competitive compared to the GPU or CPU.
2) It would be cheap enough and small enough that they could be given out to friends to get them "hooked" on the concept.
3) Set the ground work for a better atmosphere with regards to ASIC competition; make cheaper mining options quickly available to anyone.