Topic: Klondike - 16 chip ASIC Open Source Board - Preliminary - page 173. (Read 435369 times)

I was just pricing at 85C commercial parts. I really hope we don't run over about 60-65C. It's mainly the PIC if I recall because I think most of the others are rated higher, but I'd have to double check, and it's very little more for the industrial series. It's not right near an ASIC and generates little heat itself. And the ASICs, well, have no idea what they'll be but assuming they're rated higher since they sure burn the watts.

Right. If the temperature got above a certain limit, we could disable the output of the regulator supplying the ASICs and generate a fault.

As for parts, are we looking at just 85C parts, or are there some you want at 105C?

I've got a fan speed controller and temp sensor on board.

Naturally if the temp got too hot then it warrants cutting power to ASIC chip rather than trying to control fan (because fan may be dead or stuck).

Out of curiosity will there be any safety features to disable the chips if the temperature get too hot to protect the asics?

Kosta

I will be buying boards and parts to make 100 boards. I will be taking 50 for myself and sell the remaining 50.

I will most likely have a local company place the parts and reflow them. They will have all parts except for the ASICs. I will have a proper reflow oven and can place these parts on for anyone who wishes. I am located in Michigan.

Personally I am not going to try and built this at home.
16 ASIC = to much risk for me.
I don't think I would have a hard time finding out what is wrong if it didn't work.
That's what I do for a living.
But if the problem is say a short under one of the ASIC chips I don't know if I could fix it.

Keep in mind those are my costs from China and much less than any US/EU board house would charge. People should assume I won't turn around and sell at cost.

Generally there are PCB factories and board assemblers. One place could do both but not necessarily. You're not likely to get half assembled boards. You would provide ASICs to someone who assembles the whole thing at once.

Several people have expressed interest in that and I know a couple at least are pricing it out.

This is quite advanced as far as boards go. If you are the kind of person who is good at and likes DIY projects then now would be the time to read up on the net about home reflow process, stencilling etc. And visiting some of the open source board sites to get to work on a couple boards as practice before attempting this. There is still quite a while before you'd need to be up to speed.

There are two main challenges: the sheer parts count for manual pick n place on each board, and debugging a board when it doesn't work first time. This is by far the hardest part. So you make the board, plug it in and nothing happens - what next? Even worse you plug it in and it makes a cracking noise and smoke whiffs off the board. Oh, did you just put some component on wrong? You know like you didn't notice ASIC #3 was rotated and the 1.2V just shorted to GND...

ie. You need great attention to detail to pull this off.

While you could do a pick-and-place with tweezers, and do a reflow in a basic household oven, it's not for the faint of heart. Chance for failure is high for beginners, so it's not really suggested.

If you want something that "just works", buy a complete board with asics. Fwiw.

I am trying to psyche myself up about participating in an asic chip group buy, but I am still uncertain of what is needed to turn them into miners.

It sounds like the empty boards are pretty inexpensive from BKKCoins direct (30x$5).

Should any pcb board manufacturer which is capable of producing the PCB be able to mount and solder all the components if I supply the asics (and the opensource board design files)?

Is there a plan for someone to offer populated boards minus the asics?

Is adding the asics and "reflowing?" something of limited technical capabilities (i.e. doesn't own a soldering iron or multimeter - though knows what they are and in theory how to use them)?

What say the wisdom of the crowd?

Good luck!  I hope this project succeeds!

Yes, a melting temp of 183C compared to about 220C. We have shown a slight improvement of fatigue life using lead free, but with the higher reflow temp we have to worry about the higher temp on the all parts being used. I agree the SnPb is probably a better choice.

i think it's critical to use 63/37 as it will reflow at the lowest temp and has much better wetting

Am I assuming we won't be doing lead-free?

There is a lot of great information there as well. There are some characteristics we don't know, such as what mold compound is being used. If the CTE is low, we may have to increase the solder joint height to decrease joint failures. This is because we are laminating to aluminum (CTE237). The PWB is maybe 17, same as copper.

http://www.ti.com/lit/an/slua271a/slua271a.pdf  Some interesting reading here as well.

http://www.ti.com/lit/an/sloa122/sloa122.pdf

Yes, 0.3m via diameter.

The bottom of our boards do have a lot of signal paths and test points for testing, most of it is flooded with ground and is a major source of ground connections.

You do both, is the thing.

The internal ground and the heatsink on the bottom work in parallel to lower the temp better than either alone could

Frankly i'd be very suprised if Avalon's reference PCB doesn't use this technique.

ecliptic
That is good design information in your links.
I didn't see anything that says you have to have an internal ground plane I agree that an internal plane would spread the heat.
But in this design it seems the idea is to get the heat away from the board with an external heat sink.
I wonder if BkkCoins can specify the amount of copper plating in the vias under the ASICs.

All of the articles on thermal vias are focussed on dissipating the heat ONLY thru the board and maximizing this heat spreading. But in our case we know we will have a heat sink and the primary thing we want is moving the heat from the bottom pad to the heat sink. I'm not nearly as concerned with spreading the heat on the copper layers because with so many other chips on board this is going to be of little benefit.

After doing more reading I'm thinking of two changes: reducing via hole size to 0.3mm (the minimum I have available for std pcb boards), and changing the spacing to 1mm with a 5x5 grid. I don't think it's going to a big difference but at no extra cost it should improve the primary conduction to the heat sink, and is worth the small effort. (I believe we will have a 5mm width "flag" under the ASIC but we don't have specs on that yet. Maybe TSMC can provide standard reference for that packaging. I checked their web site but without customer status nothing is available)

Since the bottom layer is 1 oz. copper I'm still planning to keep it the main GND plane. But also the signal layers (0.5oz) will have copper floods that make them 99% copper anyway. The literature was helpful as it makes me conscious that near the ASIC pads I want to keep traces aligned outwards radially and not have them block heat flow near the pads. I believe these inner layers are really secondary to the purpose of getting heat to the heat sink. They are insulated by FR4 material, preventing heat loss. On a board with 1 chip they would help to spread and remove heat but with 18 chips pumping heat into the copper the best way is to ensure it gets out to the heat sink.