Klondike - 16 chip ASIC Open Source Board - Preliminary

[samurai1200]

so yeah - saving a few cents here and there won't really matter.

Why not? I'm not saying the board design should prioritize price over quality, but there's also no point in spending more money than what is needed. 10 cents saved on a $150 purchase has the same value as 10 cents saved on a $1 purchase.

Your time would be much better served trying to buy BTC cheap then, since it's the majority of the cost.

what i mean is it's not worth trying to penny pinch, time is the #1 most important factor here, I would say design a PCB Assembly that you know will work (over engineer it where possible) and test it as throughally as possible prior to the actual chips arriving

This being said, I have seen QFN-48 clamshell chip carriers around with thick gold plated slugs for the center pad. Since Avalon sample chips are coming earlier than the full batches, would it be helpful to make a, say, 2- or 4-avalon board with all the necessary hardware (interface uC, smaller version of power supply) to test/debug software on? I'd imagine you could get at least 20-30 seconds of full speed hashing before thermal issues arose (whereas with an unmounted chip I wouldn't trust it hashing for more than 2-3 seconds). That way you could finalize your design with higher confidence, and possibly have near-complete boards ready and waiting when the full Avalon batches show up.

http://www.mouser.com/ProductDetail/3M-Electronic-Solutions-Division/288-4205-01/?qs=sGAEpiMZZMsF5fZXKZXt6HnXba9GH55j3eF08QDY59M%3d

It'd probably cost you about $1000 to develop this prototype/debug platform.

I'd be willing to lead the PCB design on this. My code writing is kinda weak, but how hard could it be to pass some data from usb/cgminer to I2C/SPI/UART? Anyway, I'm sure someone like BKKcoins would be willing to help =)

[ecliptic]

I believe BKKcoin has mentioned about wanting to make a test board either with the avalon sample chips or a "dummy" that emulates their functioning?

I certainly want to do this as well, the way i see it there are two major things to test

1. Board functionality - Ideally done with a sample chip, but you can test a lot of things (uC, interface, layout/assembly, power supply [dummy load the QFN pads with their DC current draw - although you can't test Vcc noise]) even without the chip.

2. QFN Heat dissipation - this is the tricky part.  I haven't been able to look but i was considering the possibility of getting another QFN chip that produces the same TDP and testing out the thermal resistance.  of course if we can get our hands on the sample chips, this makes it easier to test everything.

to be honest though i wouldn't expect it to cost in 1000$.  the most expensive part would be the 4 layer PCB and the annoyance of assembling the boards (or the price if you get someone to pick&place + reflow them)

[samurai1200]

I believe BKKcoin has mentioned about wanting to make a test board either with the avalon sample chips or a "dummy" that emulates their functioning?

I certainly want to do this as well, the way i see it there are two major things to test

1. Board functionality - Ideally done with a sample chip, but you can test a lot of things (uC, interface, layout/assembly, power supply [dummy load the QFN pads with their DC current draw - although you can't test Vcc noise]) even without the chip.

2. QFN Heat dissipation - this is the tricky part.  I haven't been able to look but i was considering the possibility of getting another QFN chip that produces the same TDP and testing out the thermal resistance.  of course if we can get our hands on the sample chips, this makes it easier to test everything.

to be honest though i wouldn't expect it to cost in 1000$.  the most expensive part would be the 4 layer PCB and the annoyance of assembling the boards (or the price if you get someone to pick&place + reflow them)

I just threw out a number, but I dont think its that far off. 4-layer PCB containing all those components + overengineered/backup components, probably $400-500 for 3-day turn for 2-4 boards. If you wanted to test work dividing, you'd need at least 2 avalons. So thats $200 in carriers. Add all the other components and shipping, and you start creeping up on $1000 very quickly.

Anyway I think i'm going to go this route regardless of community support. Time for me to brush up on my coding skillz.

[someone42]

Generally start with one say, 0.1uF capacitance, and if it doesn't work, try changing the value lower or higher (depending on the frequency you need decoupling at) - once that is ideal, the best thing to do is add more of the exact same value capacitor in parallel.  this increases the capacitance but also DECREASES the paraistic inductance, so it gets BETTER and has more decoupling capacitance.

note : People used to add say, 0.1uF, 10nF, 100pF caps in parallel because in theory this would give a wide range of decoupling

the problem is it produces anti-resonance where the coupling gets worse.  it's a very, very tricky thing to try and fine tune, and should be avoided in 95% of cases.

my suggestion - put pads for lots of 0603 0.1uF capacitances, but only populate the reference PCB amount.  if you need more or have to tweak, the pads are right there for it.  It's standard practice to have pads for parts you don't actually populate going into production.

Thankyou for this. My experience has been exclusively with low-frequency stuff.

I think you could easily adjust the core voltage with some sort of programmable resistor on the buck reg. Not sure if such a thing is readily available but it should be. You could probably use a few FETs shorting out a binary series of resistor values to adjust the voltage divider.

eDiT: Oh geez, here you go...

http://ww1.microchip.com/downloads/en/DeviceDoc/22107a.pdf
(Now that I think about it you could probably use an analog output from the PIC as control voltage on the regulator but that would take some digging into to figure out)

According to the IR3895 datasheet, if Vref is grounded, then the output voltage can be adjusted by changing the voltage on the Vp pin. So you can easily add the capability to adjust the ASICs' core voltage.

[BkkCoins]

According to the IR3895 datasheet, if Vref is grounded, then the output voltage can be adjusted by changing the voltage on the Vp pin. So you can easily add the capability to adjust the ASICs' core voltage.

Thanks. I didn't catch that.
I may look into it as an option but with being able to fallback on std method in case it's not stable.

[yolo2222]

Interested.

Are you located in Bkk?

[greaterninja]

Looks impressive man.   Send me your PCB design, I can probably find chip to PCBA by the masses $100-$10,000 should be the cost.  Realistically I am thinking $5000 to $10000.  I am considering getting a group buy going for asic chips + pcb assembly.  The only other part will be to develop the software for the hardware.

-GreaterNInja

[fasmax]

BkkCoins
I want to thank you for being open with your design its great.
Couple of questions.
I looked at the data sheet for the PIC and it seems the PIC has only one USART port.
Do you think because the Avalon ASIC has 2 serial data lines you might need 2 USART ports?
Second question have you tested the 1.2 volt power supply circuit before ?
Thanks!

[BkkCoins]

BkkCoins
I want to thank you for being open with your design its great.
Couple of questions.
I looked at the data sheet for the PIC and it seems the PIC has only one USART port.
Do you think because the Avalon ASIC has 2 serial data lines you might need 2 USART ports?
Second question have you tested the 1.2 volt power supply circuit before ?
Thanks!

It's still unknown just what the data lines are but I considered it could be two separate hash engines. I don't think it is because of the way the cgminer driver prepares/formats the data it sends. But if I need two serial outputs then I have two choices: an external gate selector using another pin as address bit, or bit bang on a second port for one serial. Waiting on the docs. Maybe the 2 data in are synchronous serial, or one is already an address bit, or output enable.

This is my first time using the IR3895. I may build one on a proto-board before committing but I'm expecting it's good.

Interested.

Are you located in Bkk?

Not any more, but still in Thailand. I visit Bkk often.

[BkkCoins]

Actually, for $10 I can run a batch of small boards just to test the power supply. So I may do that as it will allow testing the specific layout of parts, which is important for a switching supply.

And having a bunch of 16A buck reg boards (settable voltage) on hand is useful anyway. This would also allow testing out the voltage tracking method with a control voltage too.

[fasmax]

I am concerned that 2 asynchronous USART ports may be needed.
And we still don't know what the bypass data is could be asynchronous also?
Because the Avalon internal clock is from an internal PLL I don't think that any of the I/O would be synchronous data (could be wrong though ).
I wish some one with a Avalon could look at the 2 input data lines with an oscilloscope and see if the data appears to be asynchronous (try 115200 baud).

One other concern each ASIC needs about 1.6 amps so for a 16 chip board the total 1.2v supply current needs be 25 amps or more.
Is the power supply capable of that?

Sorry I keep bugging you.

[FinShaggy]

How much longer do you think it will be before this is for sale/compete?

[sensei]

BkkCoins
I want to thank you for being open with your design its great.
Couple of questions.
I looked at the data sheet for the PIC and it seems the PIC has only one USART port.
Do you think because the Avalon ASIC has 2 serial data lines you might need 2 USART ports?
Second question have you tested the 1.2 volt power supply circuit before ?
Thanks!

2 data lines = 1 serial in and 1 serial out?

There should be a clock input for the data.

[sensei]

How much longer do you think it will be before this is for sale/compete?

I would think that within several days of Avalon releasing the specs we should have a board layout ready to test. This is my hope anyway.

[BkkCoins]

I am concerned that 2 asynchronous USART ports may be needed.
And we still don't know what the bypass data is could be asynchronous also?
Because the Avalon internal clock is from an internal PLL I don't think that any of the I/O would be synchronous data (could be wrong though ).
I wish some one with a Avalon could look at the 2 input data lines with an oscilloscope and see if the data appears to be asynchronous (try 115200 baud).

One other concern each ASIC needs about 1.6 amps so for a 16 chip board the total 1.2v supply current needs be 25 amps or more.
Is the power supply capable of that?

It would be nice to know more about the serial data sooner. I'm running on instinct

There are two IR3895 buck regulators rated at 16A each, with inductors rated for continuous 18A.
I'm expecting 12A, 1.2V on each, 24A total, or 2.3A, 12V at 87% efficiency. I didn't choose the 12A regulators because I wanted some room for over-clocking which could run 14A each.

See my github repo for all data sheets and full parts list.

How much longer do you think it will be before this is for sale/compete?

As a finished product, not any time soon. I'll be offering kits and can work with others who want to build boards. No chips specs, docs, no chips and time for testing/debugging limits how fast this can be brought to market. But I do hope to be among the first as long as sample chips turn up.

[FinShaggy]

How much longer do you think it will be before this is for sale/compete?

As a finished product, not any time soon. I'll be offering kits and can work with others who want to build boards. No chips specs, docs, no chips and time for testing/debugging limits how fast this can be brought to market. But I do hope to be among the first as long as sample chips turn up.

Well... We have a chip source, and a board source. We just need someone to do the delicate wiring, but we have a guy that can make a blueprint of it for us. I'm part of a larger group working on this, and we would definitely be interested in working on this project.

[sensei]

I will have a reflow oven, but wonder at the part count whether it would be better to a local firm do the part placement. I will have to get estimates.

I have 45 of these boards to build just for myself, but am willing to also supply assembled boards/systems in the USA/Americas.

[FinShaggy]

Has anyone taken apart an ASIC they bought? To like, reverse engineer?

What is keeping this project from completing?

[BkkCoins]

I will have a reflow oven, but wonder at the part count whether it would be better to a local firm do the part placement. I will have to get estimates.

I have 45 of these boards to build just for myself, but am willing to also supply assembled boards/systems in the USA/Americas.

This is slow but interesting as a demo of an improvised P&P machine using a CNC cutter. I bet a conversion could likewise be done on an old Roland plotter. But how much time to figure out...

http://www.youtube.com/watch?v=-CqpbsTfVtM

There are a few home made P&P machines demo'd on youtube.

[Bicknellski]

Used P&P and SMT stations? Why reinvent the wheel should be plenty of older machines out there for the right price. But really I think if you got more than a few chips and a few boards you have to take that to someone with the equipment and the skills sets. Most major cities will have these fab houses. Start looking.