Cheap and simple repair of S7 hash board

[RadekG]

My S7 and also my friend's S7 stopped hashing many times, even when boards arrived from RMA, so I decided to stop expensive sending of boards to Bitmain and wait so long without mining. This mod is only suitable for 135 chips version.

After repair I am able to adjust S7 very efficient, my B8 runs at 600MHz (4.05TH/s) with 0.22W/GHs DC, so it is around 0.235W/GHs at the wall.

Repair is very simple, every board I'v seen had malfunctioning or not working PIC microcontroller adjusting voltage for chips, so I decide to override this by 50k potentiometer. Now I can adjust voltage for each board manually.

This is original board without fix. You need to connect potentiometer between U2--R17 connection and GND which can be found on C76.

http://pantin.cz/20160209_155344.jpg

Firstly, use silicone or any other suitable glue to glue potentiometer to the board. After it dries out, you can solder its pins to R17 and C76.

http://www.pantin.cz/20160401_095409.jpg

Once you are done, you can use small screwdriver and turning clockwise potentiometer will adjust lowest voltage, about 9.3V which should be enough to start miner at 500MHz.
You can adjust voltage for each board even when miner is running and check instantly number of HW errors during operation.

I hope it will help you. My opinion is that PIC malfunction is intentional from Bitmain to lower diff after RMA period.

[adaseb]

You sure this will work? I've heard that the chips are involved in a chain series and not in parallel like the S3, so if you lower the voltage there, the chips near the end of the chain won't have enough voltage.

[RadekG]

You sure this will work? I've heard that the chips are involved in a chain series and not in parallel like the S3, so if you lower the voltage there, the chips near the end of the chain won't have enough voltage.

Yes, I am sure iit works. I ressurected 4 boards. Chips are in chain and B8+ has only 15 chips in chain and DC/DC convertor which lowers and stabilizes voltage to 10.5V, so it is 0.7V per chip. You can adjust voltage now from 9.3V to 11.5V, so it is 0.62 to 0.77V per chip.

If you want to mod working board, you need to remove U2 chip!

[sidehack]

I've been wondering since they released the 135-chip version if that was possible, but haven't had any to play with. Nice work. Fixed-voltage bucked string is actually worse than regular string, but adjustable bucked string is pretty much the best topology.

[HerbPean]

This is awesome !

thanks

[VirosaGITS]

This would be great even on working hashboard. Volt control is pretty great, it would allow for overvolt too if needed i guess?

[flikflak]

Can you tell us what kind of trimmer you have used?

I've found a quite simmilar one (Bourns Inc. 50k 1W): 3059P-1-503LF

Thanks a lot again.

[RadekG]

This would be great even on working hashboard. Volt control is pretty great, it would allow for overvolt too if needed i guess?

Yes, you can also overvolt. Maximum overvolt can be 1.013V per chip when using 16V input voltage (controller can accept up to 20V, but there are 16V capacitors), but with 12.6V input voltage it should be able to overvolt to 0.8V per chip which is maximum recommended by Bitmain.

[adaseb]

What was the lowest efficiency you achieved with the lowest voltage?

I take it 0.6V is the absolute lowest and can't go any lower than that?

[RadekG]

Can you tell us what kind of trimmer you have used?

I've found a quite simmilar one (Bourns Inc. 50k 1W): 3059P-1-503LF

[...]

You can use any 50k multiturn (cermet) trimmer. This one is also suitable. You can use cheapest one. I bought mine from sale at 10c each.

[RadekG]

Thank you. I will give it a try and let you know (and post pics).

How do you know if the PIC is working?. What symptoms did your dead hashing board experience?.

Symptoms are very easy to check. Measure voltage between inductor and GND. If the voltage is lower than 9V, you can be sure PIC is not working. If the voltage is higher and less than 10.5V, PIC is malfunctioning, partially working and you need to remove it or remove U2.

Correct voltage of string: 10.5V
Partially working voltage: 9-10.5V
PIC not working voltage : less than 9V (usually 8.7V)     

[RadekG]

What was the lowest efficiency you achieved with the lowest voltage?

I take it 0.6V is the absolute lowest and can't go any lower than that?

Most of my boards need about 9.5V to start, so it is 0.633V per chip.

[adaseb]

Most of my boards need about 9.5V to start, so it is 0.633V per chip.

So from the wall that should be about 0.20W/ghs ?

[sidehack]

I've noticed with BM1384 stuff that it likes a higher-than-chart voltage to start, probably in part because of comm latency - the first chips start hashing a bit before the last ones do, so they're relying on node-level capacitance to provide current in the instant before chips are conducting from upstream. A bit more node voltage gives a bit more stored energy to work with before the node undervolts and drops out (and takes the string down with it). There's also going to be a voltage imbalance once chips fire up, since they're not all exactly matched and some nodes will drop a slightly higher voltage than others. Taking the whole string to minimum voltage will likely undervolt some nodes while others are perfectly comfortable.

I'm assuming you've tried a minimum starting voltage and then lowering the total voltage from there?

[RadekG]

So from the wall that should be about 0.20W/ghs ?

No, I reached about 0.22 at the wall. Still much better than original 0.3+

[RadekG]

I've noticed with BM1384 stuff that it likes a higher-than-chart voltage to start, probably in part because of comm latency - the first chips start hashing a bit before the last ones do, so they're relying on node-level capacitance to provide current in the instant before chips are conducting from upstream. A bit more node voltage gives a bit more stored energy to work with before the node undervolts and drops out (and takes the string down with it). There's also going to be a voltage imbalance once chips fire up, since they're not all exactly matched and some nodes will drop a slightly higher voltage than others. Taking the whole string to minimum voltage will likely undervolt some nodes while others are perfectly comfortable.

I'm assuming you've tried a minimum starting voltage and then lowering the total voltage from there?

I completely agree with your perfect explanation. I am not native english, so I appreciate it.

I do not want to play everytime chips start and adjust voltage, so I left it at start-up voltage. I belive that PIC is able to decrease voltage right after start-up of chips, one of my board is only 9.5V which PIC always adjust (tries to override my pot settings) after chips start. Simple mod can't easily check when chips starts.

Maybe new PIC firmware can set voltage from WWW and auto-adjust start-up voltage, but I am not programmer. By the way, there must be a serious reason why all PICs are failing so frequently.

[yslyung]

thx for the find. more pics, info or updates will be great to the community.

thumbs up for a fix. any idea for the older versions with 54 chips per board ?

[RadekG]

thx for the find. more pics, info or updates will be great to the community.

thumbs up for a fix. any idea for the older versions with 54 chips per board ?

Older versions do not fail so often, they don't have PIC or voltage controller, so they are safe against this bug. You can lower input voltage directly to acheive better efficiency. No simple solution yet for repairs from me.

[SLEI]

How much resistance is used for the adjustment range, I have 10k multiturn trimpots from S1 adjustment?

[RichBC]

Very nice simple modification. The 135 chip Hash board has just been waiting for something like this. 

Strange that the PIC should fail, did you spend any time looking into what has failed in the original circuit?

I assume the 50K pot just replaces the original MCP4017 Digital Pot? Did you remove the connection to the Digital Pot?


Rich