Bitcoin Forum

Project Timeline:

August 4, 2013 - Low output capacitance reason for low overclockability, added 1000uF (from 2200uF) and achieved 350, added another 1000uF and can probably go to ~390 but software issues need to be corrected
August 2, 2013 - Trying to hunt down a bug that is causing the miner to stop working after some random time; As temp fix, wrote python script for RPi to reset miner via GPIO when not responding
July 31, 2013 - PCB fully populated with 10 chips; Hashing at 300 MHz (having issues above 300); Second PCB populated with remaining 4 chips; Board chaining appears to work
July 30, 2013 - PCB mistake found relating to SPI (corrected); Microcontroller correctly decoding nonces; Finally getting valid nonces!; Hashing with cgminer working (but with external FT232 chip)
July 29, 2013 - 10 chips received from zefir
July 28, 2013 - PCB populated (1 chip); PCB mistake found relating to microcontroller (corrected)
July 22, 2013 - PCBs received
July 12, 2013 - Simple prototype PCB etched and testing underway
July 11, 2013 - Heatsinks received; 4 sample chips received from zefir (Thanks!)
July 8, 2013 - Prototype PCBs shipped from manufacturer
July 3, 2013 - Components received
July 1, 2013 - Components ordered (minus heatsink)
June 30, 2013 - Board design sent out to be manufactured
June 29, 2013 - Requested 4 sample chips from zefir
June 28, 2013 - Double-sided board design mostly complete; Added microcontroller to board
June 19, 2013 - Microcontroller identified by cgminer using icarus protocol; Single layer board abandoned
June 9, 2013 - Project announced

Detailed Update Info:

July 30, 2013 -

Board fully populated...

http://s17.postimg.org/hnw0bima7/100_3034.jpg
http://s17.postimg.org/6pkqtbxov/100_3035.jpg

July 22, 2013 -

PCBs received...

http://s15.postimg.org/qh3ugkl57/100_3024.jpg
http://s15.postimg.org/8sc3oy9e3/100_3023.jpg

July 12, 2013 -

Temporary prototype testing board etched. Testing platform shown below...

http://s21.postimg.org/85ft36so7/100_3033.jpg

June 28, 2013 -

The board design is pretty much finished. As I mentioned in the thread, I decided to abandon the single layer homemade board and I focused my effort on the double layer board. I originally was going to modify the single layer board to double layer, but I started from scratch on the double layer board to improve the layout. It was a challenge to route the traces while trying not to excessively cut up the ground and power planes. I also decided to put a microcontroller on the board, the LPC1343. It's not the cheapest option but I'm most familiar with the LPC microcontrollers. I also added a thermistor. There is also some prototype circuitry that will probably not be needed in the final design (if I decide to even make a final design). The board design will be sent off for manufacture in the next few days. Here are the front and back layers...

http://s23.postimg.org/qxmzbmo9n/dantoplayer1_small.png
http://s23.postimg.org/w7ry2x8ij/danbottomlayer1_small.png

----------------------------------------

Original Post -

Hello everyone. Last month I announced my intention of building my own Avalon based mining board, but I didn't make an official project thread, until now. I wasn't going to start this thread until I had the board finished, but I've been working on it pretty slowly since I won't be getting a sample chip and the main batch of chips won't be arriving for quite some time.

Originally, I was going to design a single board with both the MCU and Avalon chips on it, but I decided to first make just a board with the Avalon chips and the user can use whatever MCU they want with it. I will probably make a separate controller board eventually.

I'm calling my project the "Davalon" miner for now (Dan and Avalon mixed together). This is a DIY project, I currently don't intend to sell anything. Design will be open source.

I am designing two versions of the board. One is for etching the board at home. The board is essentially single layer with a lot of jumper wires. This is the board shown below. It has a lot of pads for soldering fine 30 gauge wire to it. The other version is a double sided board for a PCB fab to manufacture. It will not have all the pads for soldering wires and it will have proper thermal vias.

http://s15.postimg.org/4k6r2hbrv/danavalonminerboard1.jpg

The area around the chips is pretty much finished, apart from the thermal vias. The right side of the board is still under construction. Mounting holes to be added.

Here are some board specifications:
- 10 Avalon chips
- 150mm x 50mm
- Double-sided board
- D12S05020 DC-DC Converter Module (20A)
- Stackable vertically and horizontally
- Boards can be chained together and controlled through one controller board
- 4-pin molex power or 3 wire terminal block

The board can be stacked vertically (one on top of the other) using board spacers. Now imagine you have several boards stacked on top of each other. Put this stack on its side and that's what I mean by "stackable horizontally".

The config and report lines broken out to a 2x4 pin header, apart from the main connector for the controller. This 2x4 pin header is used to chain the boards together while still using one controller. The top 4 pins are for the input and the bottom 4 are output. Using a 4-pin 0.1" female-to-female jumper cable, the output of one board is connected to the input of another to chain them. The config and report lines are also buffered similarly to the official Avalon miner.

Currently the design requires 12v and 5v, supplied by the 4-pin molex power connector. A 3-wire terminal block is added so that power can be distributed using wires in case there is not enough 4-pin molex power plugs on you power supply. The design may be changed to use only 12v.

So far the only solution I have for the heatsink is a 100mmx50mm aluminum heatsink secured to the board using adhesive thermal pads.

I was going to add variable voltage control using a programmable resistor, but there were some design issues that I didn't want to deal with. The main issues were getting the voltage back to the controller from each board to be measured and chaining I2C between all the boards to communicate with the programmable resistor.  If I had a MCU on each Avalon chip board, then these issues would not be a problem.

That's all for now. Thanks for your interest.

interesting project..  how long need to wait for you for board?  ;)

Are you gonna sell some? I would love to buy one of you for a reasonable price.

I wouldn't advise anyone attempting to reflow solder a board of that size at home, but at least they are not BGA chips (eg BFL) which are virtually impossible to DIY with any reliability.

PS, just saw "single layer with jumper wires". NO, just NO! You won't get proper power distribution and decoupling, so even if it works at all its going to be unreliable as f*uk. Stick with the experts folks, don't DIY!

I suppose I could sell a small number of boards, but the problem is that I can't test the board until I receive chips. So, the design won't be ready until then.

Ok, will keep checking more info on development.. Let us know how it goes.   :-*

Does it work with Avalon as additional module?
I want to add an extra module for my 3-module Avalon(after it delivered to me).

+1

This board will NEVER work.

I think it it would work, but I'm probably going to abandon the board with jumper wires because I don't really feel like dealing with the homemade thermal vias I had planned and the board would be pretty messy with all those wires.

will this be compatible with official avalon miners?

I don't think so. The interface can be changed so it's compatible but I did not take the board dimensions of the official miner into account and I believe the board on the official miner is one large PCB.

First of all, I will admit that I'm not an expert. Of course, a homemade single layer board is not going to be ideal. But, I'm going to have to ignore your suggestion of "don't DIY". Electronics is my hobby and I live for DIY. I bought the chips fully expecting that I will make my own board for it. I am selling nothing to the community and it's your own choice to use my designs or not. In fact, it will go out and say that you probably should not use my designs. I will use this thread to show the progress of my board whether it be a failure or not.

Sure, sure. Don't get me wrong, I dabble myself (but not going as far as full size SMT boards). It was more of a warning to other forum users who may get the idea that they can just plonk their chips down onto a DIY board and get excellent results. Not only is there the risk of burning their precious chips, but power supply is critical to getting these high performacne chips working at full speed. Take a look at BitFury's Alpha Testing thread https://bitcointalk.org/index.php?topic=228677.0 ... lots of very relevant info there.

Good luck with your efforts, at least you're having fun  ;)

*stalking*

I think you can have lot of fun doing it like pros

Klondyke https://bitcointalk.org/index.php?topic=227186.0

Bitfury https://bitcointalk.org/index.php?topic=228677.0

I don't think I'm any less capable of designing a functioning PCB for the Avalon chips. The problem is I decided to take an unconventional approach and announce that I was making a homemade single layer (mostly) board and I received some negativity from the community and it made me look like I was completely incompetent and that I didn't know what I was doing. Ok, I get it. I'm not making a homemade board anymore due to the various challenges pertaining to thermal vias and all the jumper wires I would have to solder. I already follow the Klondike and Bitfury projects closely.

Anyway, if anyone out there wants a real update on this project, I pretty much know how I'm going to get my board communicating with cgminer now using the Icarus protocol. It's nice and simple and I was able to get cgminer into detecting a microcontroller as a miner. And, I'm still working on the new board design. I'm hoping to have it ready for manufacture by the end of this month.

Original innovation is been encouraged.

I'm interested in this. Right now there are only two real options when it comes to Avalon PCB designs. If we can get some of the demand for miners spread across more designs, we will all probably have fewer problems with deployment when the chips finally ship en masse.

I love novelty accounts

Thank you for your interest. The project is still moving along and I hope to have the new board done by the end of this month. I've just been messing around with my BFL miner for the last few days.

3 actually not including this one :p

Board design is pretty much finished. Check OP for details.

Very nice! Can your board handle overclocking?

Overclocking is done by the firmware. If you mean to ask if my board can handle additional power load and heat from overclocking, that is yet to be seen.

Do you know how much overhead your components have...?

The regulator module I'm using is 20A. Assuming that each chip uses 2 watts, that is ~16.7A under normal operation. So, I have ~3.3 more amps to work with.

Is there a reason you're using a 20A regulator vs something bigger?

How does heat get from bottom of the ASIC chip to the heatsink?

Good stuff man. 

Do any of the online PCB places sell short runs of custom boards that others have designed?  I've seen sites like 4pcb.com and pad2pad.com, but they seem to only make boards if you design it in their software.

I'm just using the cheapest 20A regulator module on DigiKey.


There is a 5x5 array of thermal vias of 0.3 mm hole diameter on each chip. You can't see them in the image because the image is taken directly from the gerber files which don't show the holes.


I know of iteadstudio and seeedstudio.

Olimex will, if the original designer gives them a permission letter.
I have not checked lately what the limits of their trace size are.  A few years ago, I think 8 mil was their limit.