Klondike - 16 chip ASIC Open Source Board - Preliminary

[TomKeddie]

Also, whats the word on linking these together?  For example, buying them built as K16s and connecting them in groups of 4?

There was a post a week or so back from the guy doing the i2c code that he had the chips communicating on a breadboard and BKK was going to merge in his code.

Personally I think this is unnecessary, you can just run a USB per K16, USB hubs are cheap and you get better reliability.  If you have a lot hanging off the one K16 and something goes wrong you have a lot less redundancy.  I have 32 FPGAs each on their own serial port (4 serial ports per USB port and a 12 port USB hub), each one has its own pyminer instance etc.  I lose pieces from time to time but the rest keep on marching.

[Lollaskates]

Also, whats the word on linking these together?  For example, buying them built as K16s and connecting them in groups of 4?

There was a post a week or so back from the guy doing the i2c code that he had the chips communicating on a breadboard and BKK was going to merge in his code.

Personally I think this is unnecessary, you can just run a USB per K16, USB hubs are cheap and you get better reliability.  If you have a lot hanging off the one K16 and something goes wrong you have a lot less redundancy.  I have 32 FPGAs each on their own serial port (4 serial ports per USB port and a 12 port USB hub), each one has its own pyminer instance etc.  I lose pieces from time to time but the rest keep on marching.

yeah I was really hoping to avoid 16 usb connections

[TomKeddie]

Its not that the chips are available tomorrow. There is still enough time to revise everything. That doesnt need weeks.

We are getting close to the point where assemblers with multi week lead times need to place orders for components (especially for PCB manufacture at economic prices).  The chips might only be four weeks out (based first samples appearing a week or so ago and my recollection that their delivery was 4-6 weeks post samples).

*edit* wasn't directed at BKK, this is not news, was just replying to the original post.  I am confident BKK will get us there in good time and excited about the fact he's hashing.  I've brought up miners before, you get this weird feeling like you don't want to turn them off because they're earning money for you - you end up with wires everywhere and doing crazy things to avoid downtime (like moving miners across the room without any downtime).

[joeventura]

Its not that the chips are available tomorrow. There is still enough time to revise everything. That doesnt need weeks.

We are getting close to the point where assemblers with multi week lead times need to place orders for components (especially for PCB manufacture at economic prices).  The chips might only be four weeks out (based first samples appearing a week or so ago and my recollection that their delivery was 4-6 weeks post samples).

You are right. The fact is there is no working prototype yet.

So not much can be done until the nice folks here who are helping make the prototype say that "it's done"

[ecliptic]

Its not that the chips are available tomorrow. There is still enough time to revise everything. That doesnt need weeks.

We are getting close to the point where assemblers with multi week lead times need to place orders for components (especially for PCB manufacture at economic prices).

Apart from the PICs at 2 weeks, there are no components like this on the assembly (besides possibly your heat sink solution and the avalon chips themselves ofc)

you don't need 4 weeks to get an "economical" PCB.  You only need more than ~1.5-2 weeks if you're going for absolute lowest price possible for mass production

[fasmax]

Now getting correct Nonce result data.
Have receiver working and posting correct Nonce to host via USB.

Still a bit of funkiness sometimes so will need some work.
I'm not too happy with the capacitor delay for the result clocking. I played with it a bit and got it quite stable but I think I'm looking for something better. Looks like result data is ~400nS / bit so an ideal delay would be 100nS and I don't get that yet.

No time for more today as I'm off again for my Bkk run. Will be offline for ~36 hours.

So your clocking data into the shift resistor on the falling edge of the output of the NOR gate.
Do you know what the setup time is on the data input.
Also is the result data 400ns/bit time fixed or is it dependent on the clock frequency?
Great progress.
Thanks for the update.

[intron]

Now getting correct Nonce result data.
Have receiver working and posting correct Nonce to host via USB.

So your clocking data into the shift resistor on the falling edge of the output of the NOR gate.
Do you know what the setup time is on the data input.
Also is the result data 400ns/bit time fixed or is it dependent on the clock frequency?
Great progress.
Thanks for the update.

Yes. Would love to see a picture of a scope trace
capturing the real wave form.

intron

[Bicknellski]

Its not that the chips are available tomorrow. There is still enough time to revise everything. That doesnt need weeks.

We are getting close to the point where assemblers with multi week lead times need to place orders for components (especially for PCB manufacture at economic prices).

Apart from the PICs at 2 weeks, there are no components like this on the assembly (besides possibly your heat sink solution and the avalon chips themselves ofc)

you don't need 4 weeks to get an "economical" PCB.  You only need more than ~1.5-2 weeks if you're going for absolute lowest price possible for mass production

We don't? Funny... Not sure "EVERY" manufacturer is just sitting idle with no new orders coming in. The earlier you can get something to a fabricator in terms of orders etc the better off in the longer term. Cutting it tight with a 2 week or less window is just asking for delays should something not arrive. Give yourself as much LEAD time as possible. I do believe given what BKKCoins has said there is plenty of time before chips from Avalon arrive to run tests he has stated in GitHub for overclocking and any minimal redesign etc he is proposing can be done. If it can't obviously he will inform us of the choices.

Revisions are part and parcel of the process I guess but again how many revision will BKKCoins be willing to put in? I would hope that he gets some time to enjoy retirement and get things in his life back to normal if at all possible. Someone else may have to take the reins on future development given the time and effort already sunk into this BKK we can't simply expect that to continue. Given what information is available the longer term solution and the short term pain is to make these boards overclock now if at all possible if that can be tested and mods made give his willingness and time constraints. Doing more now can afford him more down time later.

I suspect knowing his drive he will want to push the design pretty hard to see the potential. He is obviously methodical in his approach so I have no doubt we will have nothing short of a great working board in a timely manner regardless if it can overclock or not I am keen on supporting this board and BKKCoins with burn in time.

[BkkCoins]

So your clocking data into the shift resistor on the falling edge of the output of the NOR gate.
Do you know what the setup time is on the data input.
Also is the result data 400ns/bit time fixed or is it dependent on the clock frequency?
Great progress.
Thanks for the update.

Setup time (or what they call Hold before CLK in Sync. USART Receive Mode) is 10nS.
I don't know if the data rate changes with clock yet but I suspect I'll find out once I get a heat sink mounted and push up the speed. I haven't mounted one yet because I'll just have to take it off to solder on another chip or two.

Yes. Would love to see a picture of a scope trace
capturing the real wave form.

intron

I tried to capture the wave to put up here but for some reason it didn't save on the USB flash stick in the scope. I got a zero-byte file instead and didn't feel like messing around figuring out why before leaving. I'll be trying again, probably more than a few times.

Now that I have nonces coming back I want to do a bit of work on the driver and see if I can get it going full circle. That would get all the basics in place and then the work would be refining them to achieve full speed, full board results.

As usual I got lots of nifty new stuff in Bkk. Vacuum pump PnP hand tool. Heat sinks and thermal compound samples, Kaplon heat tape (for when adding more ASICs), more ASIC samples, TP-WR703n, smd pin headers, linear bearings etc. But mostly right now I just need to lay down some new code and solder more chips on.

I'm getting so many requests for support now that I'm sorry if I haven't responded yet - just too much; people even asking for jobs (as if I can even pay myself a salary) - will respond when I have time unless I just don't have an answer. I'd ask people not to send items to me without clearing it first with me as my mail receiver has complained and I don't want them to cut me off. I may have to set up another mailbox and that won't be right away.

[eros]

regarding KICAD - I had better experience with the version compiled from sources - this worked on Fedora 18:

# https://code.launchpad.net/~kicad-stable-committers/kicad/stable

cmake -DCMAKE_BUILD_TYPE=Debug -DKICAD_STABLE_VERSION=20130528 . # wxWidgets and boost needed

One month newer (0.201306251957):
https://launchpad.net/~adamwolf/+archive/kicad-testing-daily

Maybe alien can fit it fedora?
Ubuntu (etc&):
deb http://ppa.launchpad.net/adamwolf/kicad-testing-daily/ubuntu lucid main

[intron]

Yes. Would love to see a picture of a scope trace
capturing the real wave form.

intron

I tried to capture the wave to put up here but for some reason it didn't save on the USB flash stick in the scope. I got a zero-byte file instead and didn't feel like messing around figuring out why before leaving. I'll be trying again, probably more than a few times.

Ok, thanks anyway. Maybe more luck next time...

intron

[Jumpsu]

PIC16LF1459-I/SS, is now avaible on elfa ,15 left.

https://www.elfaelektroniikka.fi/elfa3~fi_fi/elfa/init.do?item=10-350-25&toc=0&q=PIC16LF1459-I%2FSS

[sensei]

I really hope the next project is on Eagle. I really don't like Kicad much.

As an assembler/fabricator, I will be ordering almost 200k resistors and capacitors within the next. week. I have spent many hours going through the BOM looking for the best value from countless suppliers and looking at alternate parts as some parts have 13 week lead times. There is one part that for a run of 500 K16's, I need almost 100,000 of them.

My son and I will be building 50 of the K16's for ourselves and I will run most of them separately on USB hubs. A few we will klego together just to prove out the capability.

[joeventura]

I really hope the next project is on Eagle. I really don't like Kicad much.

As an assembler/fabricator, I will be ordering almost 200k resistors and capacitors within the next. week. I have spent many hours going through the BOM looking for the best value from countless suppliers and looking at alternate parts as some parts have 13 week lead times. There is one part that for a run of 500 K16's, I need almost 100,000 of them.

My son and I will be building 50 of the K16's for ourselves and I will run most of them separately on USB hubs. A few we will klego together just to prove out the capability.

Sent you PM

[BkkCoins]

I really hope the next project is on Eagle. I really don't like Kicad much.

As an assembler/fabricator, I will be ordering almost 200k resistors and capacitors within the next. week. I have spent many hours going through the BOM looking for the best value from countless suppliers and looking at alternate parts as some parts have 13 week lead times. There is one part that for a run of 500 K16's, I need almost 100,000 of them.

My son and I will be building 50 of the K16's for ourselves and I will run most of them separately on USB hubs. A few we will klego together just to prove out the capability.

Hold off on R47, C274 as they are changing. R47 may be dropped or moved.
Make sure you check current parts list as R45, R46 just changed.

The 0.47 uF cap is a nuisance as it's high qty tends to push you out of the stock-on-hand category, but there are quite a few alternates that could be used and some have typically had quite high stock levels.

I think that's all I have for the moment.

[sensei]

I really hope the next project is on Eagle. I really don't like Kicad much.

As an assembler/fabricator, I will be ordering almost 200k resistors and capacitors within the next. week. I have spent many hours going through the BOM looking for the best value from countless suppliers and looking at alternate parts as some parts have 13 week lead times. There is one part that for a run of 500 K16's, I need almost 100,000 of them.

My son and I will be building 50 of the K16's for ourselves and I will run most of them separately on USB hubs. A few we will klego together just to prove out the capability.

Hold off on R47, C274 as they are changing. R47 may be dropped or moved.
Make sure you check current parts list as R45, R46 just changed.

The 0.47 uF cap is a nuisance as it's high qty tends to push you out of the stock-on-hand category, but there are quite a few alternates that could be used and some have typically had quite high stock levels.

I think that's all I have for the moment.

I pulled the BOM a couple of days ago (the latest still) and I see that the quantities of the 100k ohm have dropped and shifted to the 1k ohm. I'm guessing these are the pullups for the data lines. The movement of one or so from one value to another isn't so much an issue if the part is already on the list as I'm buying these in 5k or 10k reels, but when a new value is added or one dropped, this is why I've been holding off on ordering.

I spent a lot of time on those 0.47 uF caps. I did find a replacement for about 0.5 cents each.

[cp1]

I really hope the next project is on Eagle. I really don't like Kicad much.

Maybe we could take up a collection to buy an Eagle license for Bkkcoins

[USMCGrunt0307]

I really hope the next project is on Eagle. I really don't like Kicad much.

Maybe we could take up a collection to buy an Eagle license for Bkkcoins

A little more then $1,600 bucks for the full featured license...im sure the community can scrape together 16BTC.

[BkkCoins]

I really hope the next project is on Eagle. I really don't like Kicad much.

Maybe we could take up a collection to buy an Eagle license for Bkkcoins

A little more then $1,600 bucks for the full featured license...im sure the community can scrape together 16BTC.

I'm not a big fan of Kicad but the main reason I used it was so that anyone could edit and work with the files without having to buy a license of any sort.  So I'm not really looking for a commercial version of Eagle. You have to remember that where I live I could have easily picked up Altium or Eagle for only a few $. Using Kicad was more a philosophical issue. Despite the troubles I had on the way I'm actually kind of impressed that it actually does work. I would change a bunch of things if I had the time and energy to contribute or fork the project - but I don't for now.

*****

Tonight I've been working on the cgminer driver. It was a bit tough at first seeing how it worked but I've finally figured it out I think and made some headway. I'm hoping that tomorrow I can get an initial version running for testing, and then I'll improve it as I can.

One thing I'm still not sure about is whether it should (or has to) show every slave as an individual device line (like a GPU) on the screen or whether it can consolidate them into one total per USB port. Personally I'd prefer it as an option but I'm not even sure that is possible.

And I still have to work out where it gets the cfg for each device. Presumably that is stored in the conf file like with GPU devices but I haven't quite sorted that out yet. Each device has a default config built in but I'd like user settings to be restored from the conf file upon start up, and I didn't see that in the driver code for bflsc.

******
My current opinion / standing on the over-clocking power issues is that I'll get the current version working fully and then do a version 2 board that uses the IR3847 buck reg chip instead. The circuit is almost identical and just has some position/orientation changes. The sourcing may be more difficult but it's still the better choice for boosting the on-board power without too much redesign. Most of the parts are value changes to similar parts. I think the main issue will be the Cyntec 0.215 uH inductor (PCDC1008-R215EMO) and for those looking at sourcing and supply I'd start digging into how to get test samples and place larger orders of that part. It doesn't seem to be commonly available at Mouser/Digikey et al.

Just some forward looking comments at this point.

On the K1 it's possible to replace the 6502A with the 6503A and use a powered hub to get up to 3A of juice for the ASIC over-clocking. It's almost the same price anyway but would require a small bump in inductor choice.

[TomKeddie]

My current opinion / standing on the over-clocking power issues is that I'll get the current version working fully and then do a version 2 board that uses the IR3847 buck reg chip instead. The circuit is almost identical and just has some position/orientation changes. The sourcing may be more difficult but it's still the better choice for boosting the on-board power without too much redesign. Most of the parts are value changes to similar parts. I think the main issue will be the Cyntec 0.215 uH inductor (PCDC1008-R215EMO) and for those looking at sourcing and supply I'd start digging into how to get test samples and place larger orders of that part. It doesn't seem to be commonly available at Mouser/Digikey et al.

Thanks, that seems sensible.  Perhaps it is possible on the rev 1 to use some 0R resistors to allow people to split the asics into two quads and power the second quad from a daughter board (with thick wires)?

This might calm some people.