Topic: NanoFury Project - Open Source Design - page 5. (Read 75334 times)

Anyone got luck with cgminer 4.3.3 and more chips than 6? With 6 chips I got ~10.5 GH, with more chips it's dropping.

2 things:
1) make sure you've ran the nf1init program - bfgminer will NOT recognize the device unless the product string says NanoFury
2) if you already did 1) then make sure that the WinUSB driver has been uninstalled when testing with HID - the two are incompatible (and it's somewhat tricky to uninstall the zadig stuff). Run the nf1init program and if it detects the device then bfgminer should detect it as well. If it doesn't - then you're using most likely the winusb driver (and not the HID one).

Also - I'm presuming that you're trying with a fully populated board (not just barely the MCP2210 stuff).

Ok, so now I have fully populated the board. And it is now detecting the bitfury chip also, yet another victory haha.

with HID driver:


With winusb driver:

But the hashrate is still 0.
I will do some more testing, but I wanted to let you guys know I got a little further again.

Nice!  What's next, overclocking?
Have fun

Finally!

My theory was correct, there was a mistype in the BOM, it should be SOT23R. WITH the R!

Since I don't have those, I soldered the normal one on the board upside down and translated.
It doesn't look pretty, but it works(so far haha).

I did this with an empty board with only the components that  you guys recommended.
So I don't know if it is capable of hashing, will be posting updates with more testing soon.

Thanks for all the help guys.

On the BOM you can see U3 is a sot-23 package.
When you check this document, the manual:

http://www.diodes.com/datasheets/AP7313.pdf

You can see that pin 3 is GND, since 5V is applied on pin 3, shouldn't I need package sot-23R.
Package 23R has the input on pin 3 instead of pin 1.

that'll do it.

I did some testing:

-The usb input is 5V, correct
-The resistance of the bead is approximately 0 ohm
- the input of the regulator is 5V.
- But the output of the regulator is only 0.49V(+/- some uncertainty ofcourse)
- So the input of the mcp is only 0.49V, which is not sufficient.

Conclusion, I have the wrong regulator, or the regulator is broken, or it is not soldered correctly.

I checked the packaging, the regulator is 3.3V but maybe some guy put the wrong item in the bag.

I meant pins 4-13 where 1-12 are on the "left" side if you look at the chip as on this picture: (from the NanoFury WIKI page - https://github.com/nanofury/NanoFury/wiki/The-missing-bitfury-chip-documentation#pinout-another-better-source )


Just for reference - usually chips have a mark which is near "pin 1". Pins are counted counterclockwise from that corner.

This is correct. Pins 1/2/3 are connected to ground on the schematic (and hence on the PCB as well).

No, that's not the same. Inside the chip there are tiny wires connecting the pins to the crystal. Each of those wires can carry a certain maximum current. If you want more current you have to use more wires. Hence all power pins **must** be connected/soldered.

Due to the heavy switching currents involved
it is best to use as many pins as possible to
keep impedance low and thus ringing and
ground bounce. The GND connection has
"one big pin", that's the big pad on the bottom
of the chip. Power is applied using all unused
pins of the chip package. That has the same
effect as using "one big pin", but a normal chip
package can be used now. There are not many
standard packages that have this "just one
big pin" option.

And internally this "big pin" would be connected with
many gold wires to the actual die. So using normal
pins or one big one would matter that much.

Thanks, I will be on that.

Do you mean left with the usb port on the left like in the picture?



lol

It is sloppy but looks worse on the camera because of the flash, no pins are connecting if you look at them with a magnefier.
It is pointed out that some pins of the MCP are not soldered well, but they are touching, maybe not good enough.
I'll try do draw it since it is not visible on the picture.
http://imgur.com/xAB33RN


also if you look closely on the empty board you can see a short that is already present in the pcb:
http://i.imgur.com/J6xe5kZ.jpg
In this picture it is on the bottom right of the bitfury chip.

Other thoughts, if the chip is internally shorted, what is the point of having so many pins, just make one big pin?

+1 !!!

Yup - seems like some pins have no solder at all -



The one on the top is the 3.3V which appears to be in the air - without power you obviously won't get the MCP2210 to work
The one on the bottom is one of the crystal pins - same story.

As for the bitfury chip - that one seems to be fine. Only the pins on the left side matter - all others are internally shorted (so any shorts on the pins don't matter). The first 3 pins (near the mark) are all GND too, so that short on them is irrelevant as well.
Just double check that there are no other shorts to on any of the other pins on the left side - they are important.

Also, if we presume that there might be a similar story with the bitfury chip and bad solder points - it might be a good idea to run the soldering iron by the sides of that chip as well and just re-melt the solder between the PCB and the pins (put flux there, some extra solder on the iron and use solder wick to scoop out any leftovers - it will be messy but at least it will work)


So - do what intron said and just in case touch all pins and add a bit of solder to all of them - that should get you a bit further.

Maybe you should touch the pins of the chips
with a soldering iron and add a little bit of extra
solder here and there.

Bridges only matter on one edge of the chip

This soldering is terrible ! It's clearly visible that there's no solder paste on some pins of MCP and there's brigdes on bitfury chip. No wonder it don't want to work.

Pictures of the board:
http://imgur.com/a/zYquJ#2
http://imgur.com/FT7KJFB(solder 3 capacitors under microchip?)

I will check into those voltages tomorrow.
As power supply, a usb port supports up to 5V, that should be enough.
Unless you want the full 5V potential between ground and the '5V' connection.

I don't have an oscilloscope unfortunately to check the crystal.

No. But MCP2210 chip is

Is the crystal orientation sensitive?

By what you said so far it seems the MCP2210 is not talking to the PC. Can you take some close-up pictures from your boards?


Start by checking one thing at a time.

First - connect 5V from external source - just solder 2 wires onto the USB connector pins. (anything 3.8V-5.5V will work, but don't push it beyond 6V!)
- Check the input and output voltage at the 3.3V regulator. Between the USB port and regulator there is a ferrite bead (should be equivalent of zero ohms for this test).
If both 5V and 3.3V are good - move to the next step.
- Check for oscillations on the 2 pins where the crystal is. You will need an oscilloscope for that. (note - one of the pins is more sensitive than the other, and by touching with the probe you will probably kill the oscillator - that's fine, as soon as you lift the probe it should start oscillations again and they should be good 12MHz ones on the other pin)

For the MCP2210 chip to be recognized by the PC all you need is the 3.3V and that crystal. If you have another board you can assemble just that part of the board - the ferrite bead, USb connector, 3.3V regulator, at least one of the capacitors between 5V and GND and at least one of the ones between 3.3V and GND and the crystal along with the MCP2210.

Check those and see where that leads you.

https://github.com/nanofury/NanoFury_Init

This is a windows based c++ program that uses ms visual studio to compile. This should program the nf1 with the proper naming convention needed to get cgminer and bfgminer to recognize them.

Did the parts come in a kit or did you have to buy them separately through a distributor