Topic: Klondike 1 Avalon Asic chip Mini USB miner - page 14. (Read 70653 times)

Need to get a method of mass producing these, I'd take 5 at least!

interested

Very interested! Would take at least 10-20.

also watching said thread. I would be in for at least 10.

very interested in this. watching thread

I'm interested.

Whoops..... Wrong pins. I meant to refer to the report pins.

The CONFIG_PO / _NO are the daisy chain outputs. I believe they simply copy the _PI / _NI inputs to be passed onto the next chip, but I expect they must remove one nonce from the output, otherwise each chip doesn't know which data it's working on. So if 10 nonces come in 9 go out, and each chip passes out one less until the last chip. I'd expect it to be the same rate as data input.

I'll have to look again at reading input from REPORT_P /_N as it may indeed be tricky to detect data properly there if the rate is also 4 Mb/s. It's a bit of a weird method for sure.

Thanks for pointing out that the spec's are available.
That is awesome!!

I reviewed the clock divider portion and we should be able to achieve the desired hash rate with a 16MHz clock;
so, we can get rid of the clock using the chip you selected.

The way data is entered, I'm sure speed is not a concern. it can be transmitted slower so that is not a problem.

The only concern I have (I'll continue to dig through this spec sheet to resolve it if possible) is the CONFIG_PO and
CONFIG_NO pins. I cannot figure out what are the data rates on these pins and if it can be configured or how the clock is even determined.
I'm speculating that is a 250nS period like you pointed out. If that is the case, It's probably too fast to capture the data
with chosen PIC. I will continue to look into it.

I just basically chose the cheapest one that had USB with as much as possible internal, so that I didn't need extra crystals/oscillators, or other parts. I've worked with the 8 bit ones and not the 32 bit so it was more familiar for me. I don't think it needs more but if I was already comfortable with 32-bit PICs, and seeing the only small price difference, I'd probably have gone that way.

They have a PLL to multiply the clock internally. That's why they have a second 1.2V supply. It probably needs to be noise free and the main supply is bound to be quite noisy with so much activity in the hash engine. At least that's my guess.

The protocol docs are out now so you can read up on the config word for the clock multiply/divide section.

They did today. The data is shifted in encoded on the two data lines. I forget the name for this now - brain dead, but I do recall running into it before. Anyway, I believe the period for one bit was 250nS, which makes it 4 MBits/s. Since a UART or MSSP won't be able to do that (I think, I'd have to check), it will likely be bit twiddling, and tight timing wise.

Yes, I initially wanted to use the internal clock gen and output it as 32MHz but there appears to be no way to output the internal clock. The oscillator costs as much as the PIC roughly. But this PIC is also pretty short on memory. I think it'll be enough but with the USB stack it's already 30-50% gone (depending on optimization).

It sounds like a good idea. I don't really know much about this area. I've used a different buck reg and control the EN pin from the PIC. It has a soft-start control capacitor and internal current source to ramp up, and together with the PIC deferring start up of the ASIC supply I figured I'd get a workable solution (I don't really get how to calculate the soft-start current and timing from the info they provide, so if you see how then maybe PM me). The buck reg I use is the AP6502A from Diodes Inc. I'm curious if you think that will give enough control. There is a small section on programmable soft-start in the datasheet.

Thanks for checking it out.
What was your rationale behind the PIC selection?

At the time, there was no information on my end about how to interface with Avalon and the chip I selected had a lot of capabilities
when it comes to outputting a clock at different frequencies. Thought it might be necessary.

I was surprised when I saw on the schematic that they are feeding in a fixed clock. That means they must have the ability to adjust the
clock inside the AVALON ASIC for over/under clocking purposes.

Anyways, do you know what speed the data lines use? I was expecting a separate clock for the data, but there is none according to the
labels. This is my last concern before making a PCB. If data is expected to be around 32MHZ, we may have to consider a different micro
controller altogether. Hope Avalon releases more information about this soon.

I'm seriously considering the same PIC selection you made. It's too bad we can't double the internal frequency and output it on a pin
with that PIC; we wouldn't need an external oscillator at all. That is one of the first 8 bit PICs that has auto calibration for a tight clock
frequencies to meet USB spec's.

What do you think of the DCDC setup? There's a MOSFET transistor to isolate the USB port from too much capacitance and inrush current.
the transistor will need to be pulsed a few times initially to get the caps charged without drawing too much current. A large current spike
from a USB port can cause it to auto disable.

I also included an input inductor (after the MOSFET transistor) to lower the current ripple. It will make the miner go a lot easier
on the output circuity of the USB port. Given that this is going to run 24/7, it could put some serious abuse on a computer's USB port that is
very cheaply made. I know it adds a little bit more costs, but I think it will be worth it.

Thoughts?

Correct me if i've just read wrong (I'm in work atm so have only had time to skim through a few pages), but you're considering a USB stick with the asic for ~$35? Fully built? If that's right, I'm definitely interested. I'm in the UK though so I'm guessing delivery would be considerably more.


I may be interested in being a UK distributor (of fully made & working usbs) if you'd like to head down that kinda route.
If you accept paypal, then I'll probably order twice as much as my coins would get me, or more.

allten's design is very similar to mine. He's made some different part choices, but conceptually it's doing the same thing with a bit higher end PIC chip. I must say I'm impressed with how cheap the 32-bit PICs have become. I stuck with the 8-bit one to keep costs as low as possible but the difference is only about $0.50. Not being familiar with them I thought I'd end up with two oscillators if I went that way but I see it must have a PLL inside as he generates the ASIC clock from the PIC, which is good. The Erupter appears to have two oscillators, which is a bit costly.

I'm talking with someone in Bkk now about a local assembler. If things work out we should be able to get boards professionally assembled in a proper factory here. The kind that can place 500,000 smds per day and churn these out like confetti.

I'd buy more than a handful at <50$ each.

this

the point to buy all this things, is to make profit(any investment that doesn't make roi in 2 months, is not good for me)
asic blade and other usb product are garbage, they cost too much and it is just better to sell your btc instead  

at 35 i will surely buy 10, but unfortunately i think that it will too late by the time they will be ready

How many if at all chips have you gotten yet? I'm thinking I may just order a finished product directly from you, Also where are you located?

not too bad an idea. waiting for someone else to give feedback (no idea how the hardware works)

BTW, I propose we give this the name "Quarter Stick". It will run about a "quarter" of a giga hash and it will cost about a "quarter" of 100 dollars.

Hey,

https://www.dropbox.com/s/j0hkkl4vs9nfx39/DIY_AVALON.sch.0.2.pdf

Here's the updated schematic I have so far. I started in Eagle so it is not compatible with BKKcoins, but my next project i will make it happen in KICAD.

Please take a look. All feedback may help avoid a second revision which is always nice.

Update BOM coming in the next day or so.

BKKCoins,
     Hope you find something useful here. Let me know.

@ $35 and 280mh/s this device would return the investment in 26 days.

That's a pretty good deal. I'll buy a few the day you release them, as long as they take <1.5 months to return.

I couldn't get my head around those people buying the 10gh/s block eruptor boards for 50btc (roughly $5,500). Those would take over 119 days to pay for themselves, and the hash rate on the network has been increasing, devaluing the reward for mined work by 15%+ in the last month. That's a lot of incurred debt for a long time.

These USB sticks are just plain cool though.