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

[samborambo]

-- biggest problem that I see, is that it needs to be passively cooled, I don't expect you would want the extra power drain and bloated size from running a fan or whatever. not sure if this is going to be able to run passively.


If you could build a device like that, everyone I know would be getting them for christmas!

Long answer:
A typical junction-ambient thermal resistance for a 48QFN EP part is about 23K/W. So, at 1700mW, the Avalon ASIC will be 39K above ambient. That's a worst case of 80deg C die temperature in uncontrolled environments which may be of concern, depending on the temperature range desired (doesn't meet consumer grade, etc). A small aluminium heatsink or aluminium case would guarantee temperature stability in the most extreme of situations, but is probably not necessary.

Short answer:
Passive cooling is no problem. Heat dissipated = power input. Dissipating <1.7W of heat is a doddle.

Sam.

[loshia]

Nope. The core runs at 1.2V, not 12V. That means a 12V - 1.2V DC-DC converter will need to added to the BOM, significantly increasing the cost of a 'stock' device. If you're after processing power density for your serious mining operation, go buy an Avalon mining rig.

Maybe the direction of this project has changed but I saw this as a way of getting hashing power and Bitcoins into EVERYONE's hands, not just greedy profit driven serious miners (like myself, too, I admit). We're aiming to have only slightly worse ROI than a full rig to encourage long term use.

My personal view is that USB stick miners will be incredibly popular and will eventually account for more hashing power than mining rigs/farms. The main reason is risk, or the the lack of it. The average person has no problem spending $20 on a punt; lotteries, betting and gambling are clear evidence of this. But what if this $20 punt was several orders of magnitude better odds than gambling? Make these units accessable - put them in a big dump stack at every 7-11 right next to the lottery poster. You'll sell millions.

Sam.

Sam,
I am completely aware at what voltage core runs and what is needed. Do you still think that whole PCB can be fit in single USb stick  and mine just with plug and play?
However what i suggest is just to have two version of the boards or at least not throwing away 12V option. I personally would like to be able to build more dense boards with real power supply. And of course i will pay additional price for DC/DC and other components. USB is just a crap and it can not be reliable to power anything which requires more that -.2A even though standard says 0.5A just my two cents.

Best

PS: just some calculations
http://en.qi-hardware.com/wiki/Avalon - Power

Chip power efficiency: 6.6W/GHs @ 1.15 V
Single ASIC board is making between 20-22GH depending on clock and has 80 chips

so 21 GH at average/80 = 0.2625 GH per chip
6.6 x   0.2625 = 1.7325 W per chip at 1.15 V (i am not taking into account losses when converting 5V USB power down to 1.15V )

Max output of USB by specs is 0.5A so 5x0.5=2.5W - when you take in account power loss (5V down to 1.15V) and assuming that we will have at least 15% Spare of USB specs (0.42A max) it is turning out that even a single ASIC chip will not be powered.

Am i wrong ?
allten,
I am not in hardware part at all, but can you comment that pls?

10X

[Jay_Pal]

Nope. The core runs at 1.2V, not 12V. That means a 12V - 1.2V DC-DC converter will need to added to the BOM, significantly increasing the cost of a 'stock' device. If you're after processing power density for your serious mining operation, go buy an Avalon mining rig.

Maybe the direction of this project has changed but I saw this as a way of getting hashing power and Bitcoins into EVERYONE's hands, not just greedy profit driven serious miners (like myself, too, I admit). We're aiming to have only slightly worse ROI than a full rig to encourage long term use.

My personal view is that USB stick miners will be incredibly popular and will eventually account for more hashing power than mining rigs/farms. The main reason is risk, or the the lack of it. The average person has no problem spending $20 on a punt; lotteries, betting and gambling are clear evidence of this. But what if this $20 punt was several orders of magnitude better odds than gambling? Make these units accessable - put them in a big dump stack at every 7-11 right next to the lottery poster. You'll sell millions.

Sam.

Sam,
I am completely aware at what voltage core runs and what is needed. Do you still think that whole PCB can be fit in single USb stick  and mine just with plug and play?
However what i suggest is just to have two version of the boards or at least not throwing away 12V option. I personally would like to be able to build more dense boards with real power supply. And of course i will pay additional price for DC/DC and other components. USB is just a crap and it can not be reliable to power anything which requires more that -.2A even though standard says 0.5A just my two cents.

Best

PS: just some calculations
http://en.qi-hardware.com/wiki/Avalon - Power

Chip power efficiency: 6.6W/GHs @ 1.15 V
Single ASIC board is making between 20-22GH depending on clock and has 80 chips

so 21 GH at average/80 = 0.2625 GH per chip
6.6 x   0.2625 = 1.7325 W per chip at 1.15 V (i am not taking into account losses when converting 5V USB power down to 1.15V )

Max output of USB by specs is 0.5A so 5x0.5=2.5W - when you take in account power loss (5V down to 1.15V) and assuming that we will have at least 15% Spare of USB specs (0.42A max) it is turning out that even a single ASIC chip will not be powered.

Am i wrong ?
allten,
I am not in hardware part at all, but can you comment that pls?

10X

Well, theres just one little flaw - when you step down voltage, you step up current in the same proportion if using a transformer. It's the most intelligent and efficient way to reduce voltage without significant loss.
So, yes, more than a single chip could be powered.

[loshia]

Well, theres just one little flaw - when you step down voltage, you step up current in the same proportion if using a transformer. It's the most intelligent and efficient way to reduce voltage without significant loss.
So, yes, more than a single chip could be powered.

That makes sense dude as i said i am not hardware guru Thanks for clarification

But let us have 12V option also - whoever wants it to be able to use it

best
PS: i am still not getting it. When you setp-up current it can not exceed USb 0.5A (2.5W) max current itself..Anyway do not explain me i will hardly understand it after all My point is that we are limited at 2.5W buy usb itself. and no matter what we do and how fancy we convert it we have to stay in USb input limit correct?
Assuming we power two chips 1.7325 W each it makes 2.46W and there must be some loss and spare correct? I am not mentioning the fact that some of the PCB elements will need power also and they will "eat" some Wats
 

10X

[Jay_Pal]

Well, theres just one little flaw - when you step down voltage, you step up current in the same proportion if using a transformer. It's the most intelligent and efficient way to reduce voltage without significant loss.
So, yes, more than a single chip could be powered.

That makes sense dude as i said i am not hardware guru Thanks for clarification

But let us have 12V option also - whoever wants it to be able to use it

best
PS: i am still not getting it. When you setp-up current it can not exceed USb 0.5A (2.5W) max current itself..Anyway do not explain me i will hardly understand it after all My point is that we are limited at 2.5W buy usb itself. and no matter what we do and how fancy we convert it we have to stay in USb input limit correct?
Assuming we power two chips 1.7325 W each it makes 2.46W and there must be some loss and spare correct? I am not mentioning the fact that some of the PCB elements will need power also and they will "eat" some Wats
 

10X

No problem, I'm not criticizing, just reminding something we all tend to forget - we can step down voltage using many ways - voltage regulators or resistors lower voltages converting electricity into heat (totally undesirable...!!!) maintaining the current. While it is useful and the only option for some applications, this is highly undesirable to power other devices.
Let's see if I'm doing the right math here... :-D
We have kVA = (V)oltage*(I)Amperes/1000.
From there we also can say that V= kVA/I*1000.
We might then say V1*I1/1000 = V2*I2/1000, or simply V1*I1=V2*I2
Let's fulfill the variables:
V1=5v
I1=0.4 (a nice current draw, even so)
V2=1.15v (from your data)
I2=A2
As so we have A2=V1*A1/V2=1.739A
But never forget that (P)ower (watts) = V*I
P1=V1*I1=5*0.4=2 Watt
P2=V2*I2=1.15*1.739=2 Watt which is more than 1.7325W on one chip.

Oh, now I see your point...
Just need to explain out loud the math to myself, lol!

I'd vote for an external power source also at least 12V@500mA - this would power 4 chips, if I'm correct.
12v@1000mA are quite common, also. Hard to find are those from 1000mA up and they're not cheap...
Quite interesting would be to power the circuit from a Lead/Acid 12V battery, solar charged!!! muhahahahah!

[loshia]

Well, theres just one little flaw - when you step down voltage, you step up current in the same proportion if using a transformer. It's the most intelligent and efficient way to reduce voltage without significant loss.
So, yes, more than a single chip could be powered.

That makes sense dude as i said i am not hardware guru Thanks for clarification

But let us have 12V option also - whoever wants it to be able to use it

best
PS: i am still not getting it. When you setp-up current it can not exceed USb 0.5A (2.5W) max current itself..Anyway do not explain me i will hardly understand it after all My point is that we are limited at 2.5W buy usb itself. and no matter what we do and how fancy we convert it we have to stay in USb input limit correct?
Assuming we power two chips 1.7325 W each it makes 2.46W and there must be some loss and spare correct? I am not mentioning the fact that some of the PCB elements will need power also and they will "eat" some Wats
 

10X

No problem, I'm not criticizing, just reminding something we all tend to forget - we can step down voltage using many ways - voltage regulators or resistors lower voltages converting electricity into heat (totally undesirable...!!!) maintaining the current. While it is useful and the only option for some applications, this is highly undesirable to power other devices.
Let's see if I'm doing the right math here... :-D
We have kVA = (V)oltage*(I)Amperes/1000.
From there we also can say that V= kVA/I*1000.
We might then say V1*I1/1000 = V2*I2/1000, or simply V1*I1=V2*I2
Let's fulfill the variables:
V1=5v
I1=0.4 (a nice current draw, even so)
V2=1.15v (from your data)
I2=A2
As so we have A2=V1*A1/V2=1.739A
But never forget that (P)ower (watts) = V*I
P1=V1*I1=5*0.4=2 Watt
P2=V2*I2=1.15*1.739=2 Watt which is more than 1.7325W on one chip.

Oh, now I see your point...
Just need to explain out loud the math to myself, lol!

I'd vote for an external power source also at least 12V@500mA - this would power 4 chips, if I'm correct.
12v@1000mA are quite common, also. Hard to find are those from 1000mA up and they're not cheap...
Quite interesting would be to power the circuit from a Lead/Acid 12V battery, solar charged!!! muhahahahah!

Nice.. what i was pointing out that no matter what we do we can not exceed 2.5W in total allowed by USB simple as that  

I am voting for something powerful for 4 chips

thanks dude. Let us wait for allten input
My vote is: 12V/1A - will be able to handle 4 chips at full speed with some spare I guess when they run clocked at 300 they go above 2W

[samborambo]

Am i wrong ?

Short answer: Yes.

Long answer: You were so close! You mentioned 0.42A @ 5V (2.1W) remaining power budget. The ASIC consumes 1.7W (1.48A @ 1.15V). Add 5% (0.085W) for online DC-DC power conversion losses and voila; 2.1W - 1.7W - 0.085W = 0.315W power margin. That 15% margin might be handy for some dynamic overclocking, given the right conditions.

I went to school for this stuff I'm an electrical engineer for a municipal utility company, working on several research and development projects. If I can't get simple DC power calcs right, I'd have bigger problems than an overloaded USB port.

Powering multiple devices; why not just buy a powered USB hub for $10?

[loshia]

Am i wrong ?

Short answer: Yes.

Long answer: You were so close! You mentioned 0.42A @ 5V (2.1W) remaining power budget. The ASIC consumes 1.7W (1.48A @ 1.15V). Add 5% (0.085W) for online DC-DC power conversion losses and voila; 2.1W - 1.7W - 0.085W = 0.315W power margin. That 15% margin might be handy for some dynamic overclocking, given the right conditions.

I went to school for this stuff I'm an electrical engineer for a municipal utility company, working on several research and development projects. If I can't get simple DC power calcs right, I'd have bigger problems than an overloaded USB port.

Powering multiple devices; why not just buy a powered USB hub for $10?

That was actually what i was saying - just one chip not at full speed for usb

What about add 10$ on top of usb hub and build single PCB with 12V/1A (assuming 20 USD are enough for such converter) and save money for 3 USB cables +3 PCB's + with all components apart of the ASIC chip, space for hub+power supply + 3 PCB's  
Seems more reasonable for me
What i am saying is:
A board with usb power ony will doubtfully handle even single ASIC chip at full speed reliably. As many chips as we put on a single board we go down in terms of price for the board itself and all the components inside excluding asic chip
The killer board would be 4 chips 1-1.2GH with option for the chips to be added later who wants will solder just one, two or three or four.
I dream for a bigger board personaly but we have to start from somewhere and a start is a board with external power for two chips at least just my two cents

[Jay_Pal]

It is really hard to decide on anything before Avalon reveals some more info about the chips. Data we have so far are insufficient for deciding about
PCB design, cooling solution, ROI, etc.

I like the powered USB hub idea. It is really the most elegant solution for power-hungry setups.

To be honest, I don't quiet like the USB solution.
Unless it would be stackable...
I'd prefer something quite similar to arduino's shields stacked on top or side by side to a power source.
I'm a little allergic to cables hanging all over the place... :p
How much current could an USB connector handle?

Could it be something like this:

|USB F|           \cooler/
|RISER|           AVALON
*******BOARD**********
|RISER|
|USB M|

Or...

                              \cooler/
------                      AVALON        -------
USB F|*******BOARD**********|USB M
------                                         -------
The board it self could be the "hub", i think, since the connectors would be used to communicate and transmit power between them.
The first board would be a controller of some type.

[allten]

Update to OP: April 3rd 2013

we are not looking for investors, and unfortunately batch #3 is sold out at the moment, this means you can do a few things

1. wait till when we finish installing our SMT line and see our production capacity and decide if we were to do another batch or not. ( late April )
2. wait for us to finish our documentation and start selling the Avalon ASIC chips which you can built your own mining machines with. ( mid April )

Currently we don't want any money.

Thanks for querying though,
Cheers.

Can't Wait!!!

Thanks BitSyncom!!!

[SebastianJu]

Sounds good. A date when they start selling chips...

[burnin]

awesome news.
I 'd really like stack of those chips to play with.

Also love the idea of a completely usb powered mining-stick.
Maybe 2 Chip design that could be powered by both: 2x USB or 1x USB and 1x external psu.
But i agree with you guys that USB power is troublesome.

Will your design(s) be open sourced when done?

And a modular expandable solution would be attractive to many people.
Buying modules to increase power whenever you have money to spare for that. - nice
Really like the stacking idea, but some ppl will get overboard with the stacking
and have the connectors melt down. 

[tbd]

Great news about the chip documentation.

Perhaps we have two different designs:

1. Simple - USB powered, passively cooled, inexpensive.  As many chips as fit within the 500 mA limit.  (Note, USB 3.0 has a 900 mA limit, but not sure it is worth it).
2. Bigger - Externally powered, fan connector.  Some sort of modular connectivity if possible

[ProfMac]

1. 5V or 12V power
2. 1.2V DC-DC efficienct regulator module
3. avalon chips
4. ?
5. profit!

(Seriously, I don't think it's *that* easy... Who is doing the PCB? Who is writing the Firmware?)

I have done this PCB. 
http://www.kickstarter.com/projects/1378252804/681865516?token=bfb40102

[burnin]

Im going to start work on a modular design.

Thinking of towers of cards with 10 or less chips per card.
One is the main controller that interfaces to the PC and power supply.
With a maximum of 5 to 8 cards per stack.
Cooling is my biggest concern at the moment.  

Any suggestions which micro we should use?
So we have a common base for these open source projects.
A common host interface would save considerable amounts of development time.
I'm familiar with AVR, PIC32 and TI's Stellaris .

Regarding manufacturing: i can get larger batches made.

[loshia]

Im going to start work on a modular design.

Thinking of towers of cards with 10 or less chips per card.
One is the main controller that interfaces to the PC and power supply.
With a maximum of 5 to 8 cards per stack.
Cooling is my biggest concern at the moment.  

Any suggestions which micro we should use?
So we have a common base for these open source projects.
A common host interface would save considerable amounts of development time.
I'm familiar with AVR, PIC32 and TI's Stellaris .

Regarding manufacturing: i can get larger batches made.

+1 for modular design + external power

[fnordfnordfnord]

Anybody have a footprint for the Avalon ASIC? No sense waiting around.

[ProfMac]

Im going to start work on a modular design.

Thinking of towers of cards with 10 or less chips per card.
One is the main controller that interfaces to the PC and power supply.
With a maximum of 5 to 8 cards per stack.
Cooling is my biggest concern at the moment.  

Any suggestions which micro we should use?
So we have a common base for these open source projects.
A common host interface would save considerable amounts of development time.
I'm familiar with AVR, PIC32 and TI's Stellaris .

Regarding manufacturing: i can get larger batches made.

I have a strong and irrational fondness for the MIPS architecture.  This is the PIC32 CPU core, and the core in the TP-Link 703n.

[burnin]

Anybody have a footprint for the Avalon ASIC? No sense waiting around.

you can guess some of it by looking at the pictures from ngzhangs boards.
Thats what i'm doing.

Whats on Ngzhangs mining board:

• 1.2V switching regulator
• 3.3V Linear Regulator
• 25Mhz Quarz
• Clock buffer
• (buffer for the serial interface)

But there's a lot we can do even without the exact pinout.

1. Voltage regulators design
2. basic uC design
3. cooling design
4. basic board layout (what goes where)
5. Design or reimplement a communication protocol to host.


Edit:
MIPS it is then
Even though we dont know how much computational power we really need.
But a native USB interface is always good.
(and maybe CAN for inter-module communication)

[fnordfnordfnord]

Im going to start work on a modular design.

Thinking of towers of cards with 10 or less chips per card.
One is the main controller that interfaces to the PC and power supply.
With a maximum of 5 to 8 cards per stack.
Cooling is my biggest concern at the moment.  

Any suggestions which micro we should use?
So we have a common base for these open source projects.
A common host interface would save considerable amounts of development time.
I'm familiar with AVR, PIC32 and TI's Stellaris .

Regarding manufacturing: i can get larger batches made.

AVR (best open source toolchain availability IMO)