Linked here is a video that will show you the steps required to replace any component of you Avalon 841. In reality their designs are so similar, that it is good for all models starting from the 721 series. Feel free to ignore the guy in the video calling it an 821
. The outer shell has an 821 label on it, and I didn't notice the difference until my final repair.
Video Thanks to Steven Mosher with Canaan for giving me the opportunity. It was simple I was sent an assembled unit that was non-functioning; and was tasked with identifying the issue and performing repairs. I was provided with an After Sales Support Engineer contact from Canaan, Yuxing Wei. I was able to update Yuxing on what I thought was the issue and how I was going to proceed; he also confirmed with me when I had completed the repairs.
Shipment Contents*Will update post with screen shots *First test - Getting a baseline for your issues
I was able to log into the controller, and the unit was visible. Digging a little deeper into the API log it was apparent that none of the 4 hashboards were doing any work. There are a few different issues this could have been.
- 4 faulty hashboards
- A combination of faulty hashboards and faulty PMU's
- 2 faulty PMU's - Despite them showing proper numbers for voltages, there can be an issue with the actual connections to the boards.
So I began as deep in as possible, and replaced all 4 hashboards and reassembled the Miner. I do have footage of this, but felt the published video was of better quality while still showing everything necessary.
Second test - How to identify if the problem is a Hashboard or PMU
This test revealed forward progress. I now had 3 hashboards that were operational, with MW3 still appearing dead. I was left with 2 possibilities:
- The replacement hashboard was a dud
- There was an issue with the PMU2 - Attached to MW2 and MW3
To identify the problem board at this point, I swapped Hashboard positions. I removed Hashboard MW3, and MW0 and swapped them. These boards both sit at the top position of the miner but are each powered by separate PMU's. So I am taking a known functional board and switching it's position with the problem board.
Doing this allows me to see if the board is the problem or the board position. After the switch if the problem is the Hashboard, the API log would show MW0 as dead, and MW3 hashing. If the problem is PMU related you will still see MW3 showing dead, despite there being a different functional Hashboard in that position.
In my case the issue was an issue with PMU2. This was confirmed by MW3 still showing dead.
Third testArmed with the knowledge that the Hashboards were all functional, I replaced PMU2. This was a fairly straightforward replacement, but it did surprise me. Using the API log I didn't notice anything that tiped me off to a faulty PMU as all my "VI" and "VO" number looked good. It was only through the previous Hashboard swap that I was able to confirm this issue. So it goes to show that you should not jusmp to conclusions and take your time in identifying the issue.
After swapping out the PMU, everything is functioning as expected.
Identifying PartsWhen changing out the PMU units I noticed that one side is marked with an "L" and the other is marked with an "R"
