To be honest i have not put much time or effort into the board design as i've been fairly busy and I'm waiting on the Avalon specs, but I was thinking features similar to what you find on a GPU for monitoring - measure 12V/3.3V/1.1V with an ADC and report, measure temperature in a few areas (buck converter, perhaps thermally connected to heatsink of a group of chips.

Also perhap slight changes if my local pcb assembly outsource-place would perfer/do it cheaper with say, all SMT components and no through hole, etc.



Awesome!

question though - I am not familar at all with kicad, but i downloaded a copy of your github and it seems like it only looks to open .pro project files and i didn't see any in there?

I may get the board fabbed with a U.S. company (thinking of advanced circuits but have a few places to throw quotes to)

I will look for a random 48-QFN (7x7mm) that would have close to the same thermal properties and power dissipation.  resistors are a good start, but the key part is the thermal conduction from the pad to the PCB and the PCB to whatever else

edit : small 'snag' in that plan - the only 48-QFN 7x7mm chip that burn more than 1W are.. you guessed it, high speed, fairly expensive ADCs.  same thing if you expand it to all *-QFN.

Instead you can grab a 0.50$ linear regulator in a TO-263/D2PAK and solder the tab to the QFN pad, then connect the output to a high wattage resistor and choose input DC so you can burn any wattage you want across the D2Pak (and easily calculate it based on Power=V*I)

something like one of these - and actually the higher the dropout voltage the better, the less current you have to pull through the regulator to burn a given wattage.

http://www.digikey.com/scripts/dksearch/dksus.dll?FV=fff40027%2Cfff80182%2C142c004d%2C142c0057%2C142c0091&vendor=0&mnonly=0&newproducts=0&ptm=0&fid=0&quantity=0&PV507=69&stock=1

Suggestion for testing - dedicate one (or a few) boards to thermal testing.  Populate with a QFN chip (doesn't matter what it does) that dissipates as much power as an avalon chip.  If possible, one that uses the same Vcc voltage as avalon (hence, same current draw and will also test the buck converter limits)

I will probably do this myself.. (but still waiting on the damn avalon specs!)

Are you saying he removed the ASICs from a batch #1 miner and built up his own PCB assembly and tested it with them?

That's quite forward thinking of him

Litecoin has been consistently ~1.3-1.5x more profitable than BTC for GPUs for months now.  I expect most GPUs have already moved.

(Indeed, a large % of LTC GPU miners were even switching between alt coins because they were more profitable still, but nowhere near as stable as LTC)

What is the range of estimates for BFL?

assuming

1. Order limited (i.e. fill all 10,000-20,000+ orders)
2. Wafter limited (i.e. run out of ASIC chips)

I have heard that the upper theoretical limit is something like 1,000 - 2,000 THash/sec ordered from BFL.

also - difficulty follows price.  not the other way around.  we may well see a decrease in bitcoin price

indeed it has stabilized heavily and ceased it's gradual increase on average / massively chaotic pricing.  difficulty can skyrocket and price can do as it pleases.

Why not just use GitHub?

Small quantities are always expensive because you're eating a fixed setup cost across a tiny # of boards.  Usually for digikey/mouser component i provide them a BOM and they order the parts and roll it into their quote.  For other parts (like the Avalon chips) you just provide them with them

I will get a tiered quote after avalon releases their specs and a PCB design is good to go, but assuming 16chips per board, I'll be getting a few dozen boards just myself, i expect the setup costs will not be a significant per-PCB-assembly cost.

For someone like me grouping up to get the boards assemblied isn't a significant cost savings, but it absolutely is for someone trying to get, say, 3 boards made for 48 chips.  You may end up paying more for the fixed costs than for the Avalon chips.

I don't plan on organizing anything on the forum or any shipping, but I will very likely having this PCB assembly (or one 99% similar but perhaps with a small design change to add parts for temp monitoring/etc)  pick & place / reflowed with a local manufacture in WI that i've worked with in the past with good results.  My goal is to get them up and running ASAP after receiving chips.

 Just throwing it out if anyone in the region would be interested in doing a local group order to save money.

It's a good idea in general, communicate and meet up locally and get the boards built.  You could easily spend more downtime on shipping everything there and back then on the actual assembly itself, and it always introduces more risk and middlemen.

And with a local place you can always ask them to build up 1-3 boards, test them, make sure it's all good, then do the rest.

Latest difficulty estimates show a jump from ~480 to ~630.

Look at litecoin diff.  It's gone from what, 300 to 600 within a month?

It is going to go from 470 to 614 by the weekend.

ASIC NRE is very expensive, and the entire undertaking is very risky.  Lots of sunk cost and assumed risk worth repaying.  

10$ for a custom made IC for SHA256 is not bad at all - You can pay hundreds for mass produced commercial ICs.

I don't think you know what 'vaporware' means.

Which board design are you referring to?

If you are going to get these pick & placed, most assemblers will not want cut tape components, only reels.  - like i mentioned earlier, if you don't want to buy a full T&R of 5000/10,000 components you can usually pay a fixed 7$ + unit cost.  but for things like resistors you might be able to get a 10k reel for 10$ anyway

Is it just a simple standard value standard package SMT resistor?  if so you should be able to find another manufacturer version that digikey or mouser have digi-reels or mouser reels (i forget what they're called) where they sell you a partial size reel of the part for 7$ fixed part + higher per unit cost - this makes it orders of magnitude easier to pick & place.

Or a coin based on solving CAPTCHAs - so 'mining' has to be done by people manually, computers can't do it.  Perhaps somehow you could exchange the coins for solutions to captchas, making it attractive to botnet owners and spammers.  Or use it as a distributed rainbow attack on CAPTCHA, so you have a lookup table with realtively high % chance of success if it matches a given CAPTCHA image.

Why not a coin which automatically starts a new alt coin every time a block is found?

forkc()in

If you're a real bastard, all the child coins also fork

https://www.litecoinpool.org/stats

hash rate was over 20 GH/s for a while there

Is 100 wafer's what BFL ordered?  How many chips do they get per wafer?

i've seen estimates up to 10-20k BFL single orders but I think nobody knows how much hashing power they will add to the network or if they will be wafer limited or not

Is the DC Input fused?