Block Erupter USB - Overclocking/ hacking ?

[btct22]

Thanks for your pictures     What is the current draw, it looks like you are supplying 5V separate from the USB port, was that the cause of the fluctuations / HW faults?

[Trillium]

Waterblock or heatpipe cooler, problem solved.

[amigaman]

I don't think an external supply of the core voltage will be possible, just because the longer lines will add inductivity and noise. Also, i don't think you'll have an external power supply that can be set up to .01V and be stable with max .01V fluctuation.

Secondly, the tupperware idea is nonsense. There will be no problem with the stick in the fridge, as long as you keep water from dropping on to the surface. It's the cold things water from warmer rooms condensates on, the stick will be the warmest thing around, so no condensation.
And the cheesebox is of thick plastic, so the air won't circulate, meaning you'll get it hot in the box, cold in the fridge, not the effect you wanted.
The membrane is for exchanging air pressure, not moisture, so the cheese won't become dry, but test yourself: blow against the membrane: no air going through. It's meant to exchange sloooow.

So: we'll need to improve cooling, and this mainly means "how to get that fricken thing screwed/glued/whatever to a sufficient heat sink, lowering the overall temp to <70°.
Clock & Power Supply seems not to be a problem. Although i'd like to know how much 5V power the 450mhash draw, even better: how much of it is the ASIC...

[Bluestreak66]

Thanks for your pictures     What is the current draw, it looks like you are supplying 5V separate from the USB port, was that the cause of the fluctuations / HW faults?

Yes main issue was the 5v supply its actually 5.05v now and needs to be separated from usb power, however the grounds need to be tied together otherwise ground loops are formed and noise is introduced resulting in errors. I haven't checked the current draw, I'm just going for long term stability atm. Rough measurements suggest it's drawing about 6.4 watts that may be on the high side I'm not setup to do a proper current measurement atm.

Waterblock or heatpipe cooler, problem solved.

Heat sink is cool but the asic is now as hot as ever maybe even a little warmer than the stock setup. The issue is with conducting the heat though the board and to the heatsink, I may try and find some better thermal material for between them. I bet if that's solved I can bolt it to a water block, up the voltage and almost triple the stock speed. When these were being designed I know that in the beginning simulations were done for these to run at 1Ghps I'm not sure how high hash rates were actually tested. Half way through the design the decision was made to lower the rate in the production due to the high power consumption and heat dissipation. It may not be feasible to run them that high but it will be interesting to see how high it will go.

[Trillium]

The issue is with conducting the heat though the board and to the heatsink, I may try and find some better thermal material for between them.

What are/were you using as the TIM?

[Bluestreak66]

The issue is with conducting the heat though the board and to the heatsink, I may try and find some better thermal material for between them.

What are/were you using as the TIM?

Currently just the stock silicone sheet. I have some Arctic Alumina adhesive here in front of me that look quite tempting but after the heat sink is on it makes it difficult to rework the board.

[btct22]

So the overclocking mods you've done at the moment are a 16mhz crystal with voltage regulator resistors R1 @ 1k2 and R2 @ 1k5  and the whole unit powered by 5.05V? 

I'm going to attach a RAM heatsink to a single BE100 chip to start with.  Then for higher speed overclocking was thinking about clamping a heat pipe + fan unit onto 4 block erupter units lying on a passive heatsink. Each unit's USB connector facing outward and the units offset so it should be possible to have a BE100 in each corner of the heat pipe collector heatsink.

It's good if the power doesn't come from the USB system, as then can wire in a non-powered hub for data and have a separate adjustable power circuit.

[Bluestreak66]

So the overclocking mods you've done at the moment are a 16mhz crystal with voltage regulator resistors R1 @ 1k2 and R2 @ 1k5  and the whole unit powered by 5.05V? 

I'm going to attach a RAM heatsink to a single BE100 chip to start with.  Then for higher speed overclocking was thinking about clamping a heat pipe + fan unit onto 4 block erupter units lying on a passive heatsink. Each unit's USB connector facing outward and the units offset so it should be possible to have a BE100 in each corner of the heat pipe collector heatsink.

It's good if the power doesn't come from the USB system, as then can wire in a non-powered hub for data and have a separate adjustable power circuit.

It doesn't matter what type cooler you put on it if you cant get the heat to the cooler. I have a huge cpu heat pipe cooler here with a peltier cooler on the bottom that i was playing with. It will ice the bottom of the board but the top is still hot. The whole contraption along with power supply was drawing ~70 watts, not really feasible if your trying to get bitcoin with these things. It did clearly show that the heat transfer to the cooler is the issue. Also be aware that the voltage regulator is at its rated limit running like it is. By my calculations is outputting 1.47v @ 3.05 amps, it does get hot. If anybody finds some cheap thermal material that could be used let me know. 

[systic]

What about submersing in oil like the oil cooled gaming rigs?

[rmyadsk]

On the hacking front - is there a way to get diagnostics or self-check data from the device itself - to see how it "thinks" it's mining? Some basic messaging through the USB to UART port?

[amigaman]

Too bad friedcat didn't think about the possibility for the unit to report it's temps back.
The CGMiner shows nothing where GPU show fan speed and temps, so i guess it is not possible in Hardware.
So i guess no, that is not possible.

[btct22]

What I was thinking was mounting the fanned active heatsink on the top of the BE100 chip and a big passive heatsink at the bottom of the PCB, so both sides of the chip are being cooled.

When the stuff I've ordered gets here I'll try it out and post some pictures.

[Bluestreak66]

On the hacking front - is there a way to get diagnostics or self-check data from the device itself - to see how it "thinks" it's mining? Some basic messaging through the USB to UART port?

The SPI port is accessible from the breakout pins so a lot could be done with that. I was going to try and pull the firmware off the chip if/when I get time.

Too bad friedcat didn't think about the possibility for the unit to report it's temps back.
The CGMiner shows nothing where GPU show fan speed and temps, so i guess it is not possible in Hardware.
So i guess no, that is not possible.

Theres no fan or thermistor on the board and I doubt theres one in the asic so there would be nothing to report back. A thermistor would have required at least another resistor and a pin on the micro so I can see when he didn't put one in there as theres not much room as it is. Plus I'm not sure I would want to see the temps sometimes they may get quite scarry sometimes! 

[btct22]

So the overclocking mods you've done at the moment are a 16mhz crystal with voltage regulator resistors R1 @ 1k2 and R2 @ 1k5  and the whole unit powered by 5.05V? 

I'm going to attach a RAM heatsink to a single BE100 chip to start with.  Then for higher speed overclocking was thinking about clamping a heat pipe + fan unit onto 4 block erupter units lying on a passive heatsink. Each unit's USB connector facing outward and the units offset so it should be possible to have a BE100 in each corner of the heat pipe collector heatsink.

It's good if the power doesn't come from the USB system, as then can wire in a non-powered hub for data and have a separate adjustable power circuit.

It doesn't matter what type cooler you put on it if you cant get the heat to the cooler. I have a huge cpu heat pipe cooler here with a peltier cooler on the bottom that i was playing with. It will ice the bottom of the board but the top is still hot. The whole contraption along with power supply was drawing ~70 watts, not really feasible if your trying to get bitcoin with these things. It did clearly show that the heat transfer to the cooler is the issue. Also be aware that the voltage regulator is at its rated limit running like it is. By my calculations is outputting 1.47v @ 3.05 amps, it does get hot. If anybody finds some cheap thermal material that could be used let me know. 

So it's drawing 3.05A with a 16mhz crystal?  That does seem like a lot of current.  Would the voltage regulator need heatsinking on top as well?

[Trillium]

There is no onboard temp monitoring because it keeps costs down. ASIC are internally simple (relatively) there are images of a the die on a site which I conveniently can't find where someone etched the epoxy away with hot nitric acid.

Adding a sensor would increase ASIC manufacturing cost and number of components in the circuit, and not really required for normal use.

[Bluestreak66]

So the overclocking mods you've done at the moment are a 16mhz crystal with voltage regulator resistors R1 @ 1k2 and R2 @ 1k5  and the whole unit powered by 5.05V? 

I'm going to attach a RAM heatsink to a single BE100 chip to start with.  Then for higher speed overclocking was thinking about clamping a heat pipe + fan unit onto 4 block erupter units lying on a passive heatsink. Each unit's USB connector facing outward and the units offset so it should be possible to have a BE100 in each corner of the heat pipe collector heatsink.

It's good if the power doesn't come from the USB system, as then can wire in a non-powered hub for data and have a separate adjustable power circuit.

It doesn't matter what type cooler you put on it if you cant get the heat to the cooler. I have a huge cpu heat pipe cooler here with a peltier cooler on the bottom that i was playing with. It will ice the bottom of the board but the top is still hot. The whole contraption along with power supply was drawing ~70 watts, not really feasible if your trying to get bitcoin with these things. It did clearly show that the heat transfer to the cooler is the issue. Also be aware that the voltage regulator is at its rated limit running like it is. By my calculations is outputting 1.47v @ 3.05 amps, it does get hot. If anybody finds some cheap thermal material that could be used let me know. 

So it's drawing 3.05A with a 16mhz crystal?  That does seem like a lot of current.  Would the voltage regulator need heatsinking on top as well?

I don't think it would hurt to have a heatsink on the regulator. Really it needs a bigger regulator, current will go up as frequency is raised. The voltage may be able to be lowered slightly to improve efficiency, I was limited on resistor values a high quality pot could be used to find the optimum voltage. It would be nice to find a pin compatible voltage regulator that would go to ~5 amps or so.

[amigaman]

And a "pin compatible" coil...

Regarding the NTC: It would not have increased the cost so much, but would have been a good indicator if the cooling needs to be revised.
But FC had the Problem of putting them chips on the market as fast as humanly possible to be able to sell them with a profit. So maybe they missed something not-so-vital.

[Bluestreak66]

And a "pin compatible" coil...

If your referring to the inductor, it's a standard footprint and depending on the chip used the stock one may able to be used especially if the switching frequency is increased, which would also increase efficiency.

[amigaman]

After seeing all those barely sufficient parts on there, and even the idea that a one thousandth cent worth of NTC isn't applied, you really believe the coil is sufficient for more than absolutely necessary?
I'd say it's at the end with supplying slightly above 2A, sizewise. More is converting more power to heat, but you don't notice that becaus the heat spreader on the back makes it all untouchable.

Another coil with more amperage on the same footprint would have far less inductivity, hence needing a massive increase in switching speed.
And you'll need a pin compatible control ic with this, good luck finding one, better luck getting your hands on some.
You'll better desolder all parts and make a new USB-SupErupter from it, with a massive 3+-layer Board (totally flat and pure copper with no holes on the underside besides some for heat transfer from the asic), and a power supply capable of 5+ Amps, also temps measurement.

[Lucko]

Just wondering how long chip would live like this... Isn't highest voltage on a blade 1,2 and about 400MH?