I don't know where to get them in small quantities. I get them as 5000 piece reels for 5 EUR from Farnell.

For a more durable way to overvoltage the boards you should replace resistor R01F. The nominal value is 1.0K, and to increase the voltage the resistor needs to be replaced with one of a higher value.

A value of 1.5K is generally still safe, and should increase voltage from 0.65 to 0.7. If you want to go higher, you can try 2.2K or maybe even 3.3K if you feel lucky.

Of course, as punin said, this will void your warranty, and may break your card. Keep a close eye on the regulator temperature. Also, the resistor is in a somewhat tricky position to desolder, and you risk getting tin on the pins of the regulator.

The resistor should be 0603 size to fit properly on the pad. Also, I recommend a 1% type with low temperature coefficient. Resistors with high positive temperature coefficient may cause thermal runaway: as the resistor heats up, the voltage increases, causing the resistor to heat up even more.



Note: if you accidentally mess up the capacitor next to the resistor (C10F), and you wish to replace it, use 1nF, 50V, X7R type. Similarly, resistor R02F is 10k, 1%.

These chips are designed to dissipate heat through the ground pad and to the bottom of the board. In practice, cooling the board works well enough.

What is "pencil mod" ?

Note there's no 30A fuse, and it doesn't blow. There's a resettable polyfuse in the 12A supply. If there's too much current, it will get hot, and as it gets hot the resistance increases, which will reduce current. It is resettable, which means that after removing power and letting it cool down, it will restore itself. There's no need to replace it.

Unfortunately, these polyfuses aren't particularly accurate or fast. Don't depend on them to save your board. Their purpose is to protect your power supply from blatant shorts on the H-CARD.

Maybe somebody replaced a resistor and increased the regulator voltage. I don't think there are regulators in the same shape that have higher current capability.

"Toch?" is Dutch for "Right?"

0603 size has a somewhat higher inductance, so it may not perform as well as 0402. Whether that's significant, I don't know.

Soldering 0402 has one advantage though: after you pick and place 10000 of them, suddenly 0603 seem as huge as bricks.

Right. You can't have too much overkill.

Maybe add a solder jumper from INCLK to GND.

As far as calculating bypass caps, to do a proper job you need to know the layout of the board and the exact specifications of the capacitor. As they say: "ain't nobody got time for that!"

More little caps in parallel have lower inductance than a few big ones. Maybe 15 caps is overkill, but it works, so why mess with it ?

I could try mounting fewer, but the problem is that if it doesn't work properly, I'd have to hand solder the rest, which is a terrible job.

I agree, the solder jumper will surely cause problems.

Yes, but only for small series, like one or two prototypes. This is a board for Foofighter to help fit a case for it.

I also made these boards (in Dave's first post) in a similar way.

Since the SPI signals form a long chain from chip to chip, there's probably a problem at chip 6 or 7. Check those extra carefully on the board. Note: the chips are numbered in white ink in the same order as in the software.

If the chips are mounted correctly, and one is broken, it could be that it prevents communication with next one in the chain. In that case, the chip could be removed, and the solder jumpers shorted. Beware: don't attempt to fix unless you've done stuff like this before.

That's flux residue from somebody fixing a short with soldering iron.

Here's a couple of photo's I took while assembling an S-HASH board.

No, 0.25A is not right.

I just measured it on the S-HASH board by putting a 1Ω resistor on the output of the 1.8V regulator. Voltage drop was 95mV, so current is 95mA for 16 chips, or about 6mA per chip, which sounds a lot better.

The PCB should be ready in a few days.

The internal clock has some jitter, which reduces the setup/hold windows inside the chip. Theoretically, with a stable external clock you should be able to go a little bit faster. But is it worth the effort ? The internal clock has an advantage that you can tune it per chip.

So the first 5 seconds you run at half speed. Not a big deal over the lifetime of the equipment. And with 500 BF chips, you should have a better solution than running getwork to a single host anyway.

If you notice that returned timestamps are increasing too fast, send work faster.