Cheap and simple repair of S7 hash board

[RichBC]

This PIC and it's actions make you wonder if perhaps a different firmware in the PIC couldn't improve the S7 efficiency. Consider something like:

1) At power up for the PIC, send out the required value to safely and reliably start the chain (e.g. 10h)

2) Delay a minute (whatever it takes to get board warmed up and settled in)

3) Send out a different value that reduces the voltage some to up the efficiency.

It wouldn't really matter how long the board ran at the increased "get it started" voltage, since what really matters is the runtime voltage for the numerous hours after startup that matter.

just a crazy thought.

And with an additional connection to the PIC it's probable possible to monitor a voltage and do some dynamic adjustments.

Rich

[1715256661]

1715256661

[1715256661]

1715256661

[1715256661]

1715256661

[E]

I think Bitmain has 19" rackmount enclosures for S5+/S7 boards fitting 9 boards. We can see unpopulated power connectors using screws. We can also suspect other unpopulated connectors may be used for industry mining.

Still unclear, WHY it fails the same way and why Bitmain did not solved it yet. Just repaired another two boards with failed PIC.

I saw a lot of glitches on the pot's i2c interface -- haven't put a scope on but there were a bunch of random very short pulses showing up on the logic analyzer... If the pic isn't fused to prevent reprogramming it's possible random glitches could put it in a bad state. Just speculation at this point.

[RichBC]

I saw a lot of glitches on the pot's i2c interface -- haven't put a scope on but there were a bunch of random very short pulses showing up on the logic analyzer... If the pic isn't fused to prevent reprogramming it's possible random glitches could put it in a bad state. Just speculation at this point.

Could just be that it was a hasty addition with poor layout / grounding / decoupling / supply etc and with the high currents involved in the hash board it's prone to interference and crashes?

Rich

[E]

Looks like the digipot may be used for voltage binning of the the hashboards!

The two images below show chain voltage and i2c traffic from the PIC on two hashboards.

Both hashboards come from the same S7F1; both of them startup at 9.177V but about 8ms after the regulator starts to ramp the voltage the pic issues a write to the pot, and it's a different value and results in a different voltage for the two boards!



The third board in this S7 was also a '21' / 10.26V board.

So on these three boards we have:

Pot Value -> Vchain
16 -> 10.38V
21 -> 10.26V

I have 9 other regulated S7 hash boards on hand and I'll try to see what values are present on those.

[RichBC]

Looks like the digipot may be used for voltage binning of the the hashboards!

The two images below show chain voltage and i2c traffic from the PIC on two hashboards.

Both hashboards come from the same S7F1; both of them startup at 9.177V but about 8ms after the regulator starts to ramp the voltage the pic issues a write to the pot, and it's a different value and results in a different voltage for the two boards!


The third board in this S7 was also a '21' / 10.26V board.

So on these three boards we have:

Pot Value -> Vchain
16 -> 10.38V
21 -> 10.26V

I have 9 other regulated S7 hash boards on hand and I'll try to see what values are present on those.

Very Nice shots. This makes a lot of sense, however still does not explain why they have used a PIC as opposed to a non volatile digital pot and no PIC? Unless.... under different circumstances on a given board the PIC writes a different value to the Pot?

Rich

[E]

Very Nice shots. This makes a lot of sense, however still does not explain why they have used a PIC as opposed to a non volatile digital pot and no PIC? Unless.... under different circumstances on a given board the PIC writes a different value to the Pot?

You're right, and I should more properly have said, "Looks like the digipot may be used for voltage binning of the the hashboards, plus other unknown applications"

I've only looked at startup behavior so far -- no idea if there's any runtime behavior from the controller or the PIC (can the controller even talk to the PIC? I haven't investigated that at all). OP indicated that the PIC actively tries to adjust voltage on one of his modified boards:

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.

So it may not be coincidence that the new value is written right as the voltage ramp finishes.

Anyway, I'll keep poking at this in spare half-hours.

[RichBC]

Very Nice shots. This makes a lot of sense, however still does not explain why they have used a PIC as opposed to a non volatile digital pot and no PIC? Unless.... under different circumstances on a given board the PIC writes a different value to the Pot?

You're right, and I should more properly have said, "Looks like the digipot may be used for voltage binning of the the hashboards, plus other unknown applications"

I've only looked at startup behavior so far -- no idea if there's any runtime behavior from the controller or the PIC (can the controller even talk to the PIC? I haven't investigated that at all). OP indicated that the PIC actively tries to adjust voltage on one of his modified boards:

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.

So it may not be coincidence that the new value is written right as the voltage ramp finishes.

Anyway, I'll keep poking at this in spare half-hours.

Doing a bit of maths on the values measured equates to about 0.024V / step. The Pot has 127 positions this would give a full scale range of about 7.75V to 10.75V. On reset the Pot is set to mid position which equates to about 9.2V and this ties up with the measured start up voltage.

Would be interesting to trace the other connections to the PIC and see if it has the ability to communicate with the main controller or measure a voltage? The other strange thing is why this simple piece of circuitry seems prone to failure?

Still also think it's worth someone checking if it's not code protected and the PIC contents can be read, all the ICSP connections look to be on the board edge.

Rich

[RadekG]

Looks like the digipot may be used for voltage binning of the the hashboards!

The two images below show chain voltage and i2c traffic from the PIC on two hashboards.

Both hashboards come from the same S7F1; both of them startup at 9.177V but about 8ms after the regulator starts to ramp the voltage the pic issues a write to the pot, and it's a different value and results in a different voltage for the two boards!

[...]

The third board in this S7 was also a '21' / 10.26V board.

So on these three boards we have:

Pot Value -> Vchain
16 -> 10.38V
21 -> 10.26V

I have 9 other regulated S7 hash boards on hand and I'll try to see what values are present on those.

All my boards have the same voltage 10.59V when running.

Firstly I thoght the same like you and RichBC, I thought it can dynamically adjust voltage based on frequency and HW error rate. My opinion had good base, because CGminer starts firstly for "pre-heat" period of about 30s and then stop and restarts again. But on my boards, voltage didn't changed before or after cgminer starts or restarts. So question still remains. What is the purpouse of PIC and digi pot?

Possible answer: Public FW has removed voltage control feature. I even tried to pass "voltage" parameter with no luck.

[cryptichermit]

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)    

I concur with this assessment (for the most part).
I had a batch 16 board failure within 48 hours.  Measures 9.16V - doesn't work at all - reads 30 chips.
I have another board that was an RMA replacement from batch 9 that partially works (actually seems to work best when around 69-72c) reads 10.12 at start and 10.26 at load.
Normal-fully functional board measures 10.56 at start and up to 10.69 at load.

Ordered some trimmers today, looking forward to resurrecting boards - unless bitmain RMA gets back to me first.
So on the partial working unit will i need to remove the U2 chip?

[scyth33]

Does anyone know what IC is U101 and Q2? its located at the top left on the inductor side

[adaseb]

Found this for sale on eBay. Apparently its a dead S7 hash board. Any idea why they tried to mod it?

http://i.ebayimg.com/00/s/MTIwMFgxNjAw/z/Em4AAOSwQNRXLX-7/$_57.JPG

[RichBC]

Found this for sale on eBay. Apparently its a dead S7 hash board. Any idea why they tried to mod it?

I guess it is just a failed board as described in this thread, modded as described in this thread, does the text say anything?


Rich

[RadekG]

I guess it is just a failed board as described in this thread, modded as described in this thread, does the text say anything?

Second picture shows burnt VRM chip. It has no connection to this mod. The seller sells also another hash board with burnt VRM without mod.

[RadekG]

I have some burnt boards.Some coolers detached due to experiencing high temperature.Is there any way to fix?

remove heatsinks from unsoldered chips and resolder them back. Chips usually survives high temp, but double check partially desoldered chips which retain at its position, but short circuit some connections

[aladinsain]

Hi
i have done the repair on 2 hash boards  1 on a batch 10 which seems to be working ok and 1 on a batch  16 which is showing all a the ascics but will not get up to temp and hashing at about 100gh max  any help would be appreciated

[opmac]

Looks like I may need to do this with a board I have.

Could someone post a link as to where your purchase trimmers from. Thanks

[aladinsain]

Looks like I may need to do this with a board I have.

Could someone post a link as to where your purchase trimmers from. Thanks

Hi i got mine from maplins i'm in the UK

[pusttiu]

Hi
i have done the repair on one hash boards , the voltage form gnd and the coil it whas 8.4Vcc, i put the 50k and now i have 9,7Vcc .
the miner see only on  ( 1 )asic on board why???
pls help .
the board is hot, like is worching alll, no temp on all 3 board.

[sidehack]

Got a HEX dump off a working PIC. In free time I'm gonna play with either figuring out this firmware to see if it takes in voltage data, or maybe just write a new one that I know does. The 6-pin header on the board is indeed an ICSP header. The ICSP clock and data lines correspond to the PIC's clock and data lines for its half-duplex sync transceiver; the DPOT's SCL is tied into the same clock but SDA runs to a GPIO pin so the PIC must be bit-banging I2C for the pot. That might actually be good news; if the PIC is taking in volt-change commands on the header, that keeps the DPOT separate from getting garbage data and hosing up the setpoint. Tomorrow I need to get to manufacture but today I'm gonna play with the PIC, see if I can at least alter the default voltage by day's end. With the S9 dropping and the halving coming down, just having a lower-fixed-voltage option for the S7 that doesn't require obviously-warranty-voiding hardware changes would be nice. See about hitting 3TH at 700W stable.

Definitely will be pulling from some of the info in this thread.

EDIT - looks like RA5 on the PIC, which can be set up as a receive pin for the internal USART, is wired to the TX pin on the 18-pin controller jack. Doesn't look like anything is tied to RX, but the chip is definitely capable of listening to the controller. Makes me wonder if Bitmain has a secret cgminer that sends out volt-setting parameters.

I haven't looked at the control hardware enough to know what the TX and RX pins on the 18-pin are tied to. Is it a hardware UART that's easily library-accessible from the OS? Because that could make things interesting.

More EDIT - I decided to start a fresh thread, since though it's peripherally related to this "repair" the idea is altered and extended quite a bit.

[RadekG]

Hi
i have done the repair on one hash boards , the voltage form gnd and the coil it whas 8.4Vcc, i put the 50k and now i have 9,7Vcc .
the miner see only on  ( 1 )asic on board why???
pls help .
the board is hot, like is worching alll, no temp on all 3 board.

Please clarify your question. You added 50k pot and icreased voltage, but it is still not high enough. You need 10.5V, so turn your pot to set right voltage and check each board separately. Be sure all boards are fitted in S7 case, but only one is connected to PCIe and controller.