|
Title: Fixing Dragonmint T1 miners: Thoughts and observations Post by: lightfoot on February 14, 2020, 12:36:29 AM So I've had a few requests to fix broken Dragonmint miners. These are.... interesting units and the boards seem to fail a fair bit. As usual there is little to no documentation on the hardware itself, so in order to fix it you kind of have to reverse-engineer the thing. What fun....
First up is to take a look at non-working boards and see what they do. Since we can't talk to them on the spi bus we have to use other tools. In my case I like to start with a watts up power meter on the supply (to monitor how much the board is pulling and when) and an infrared camera to see exactly what is warming up/pulling heat/shorted on the board. To a fault, the first three bad boards exhibit the same symptoms: They start out pulling pretty much no current, then pull about 60-120 watts of current for a few seconds, then it drops back to zero. This correlates with the logs from the miner trying to initialize the bus and get it going. On a working board it will go to 120, then 250, then whatever the watts the board is tuned for (in the lowest power setting case around 355 watts). Second was to take IR pictures of the bad boards. And I spotted something pretty quickly: When the board started up one chip on the control circuit got very hot, then the heat went away once the power draw dropped to zero. Like clockwork. I focused in on the chip and found it's an 8 pin SOIC chip with a part number of KBA6F1806 on two of the boards and KBA6F1752 on the third. Pin 8 goes to ground, two pins go to an RC circuit, and one pin goes to the crystal oscillator. It doesn't seem to be a FET driver, I'm wondering if it is a scratchpad memory for the main processor circuit. However I haven't found anyhting on it, anyone know what it is? If it's overheating then we may have an answer to the problem: That chip goes bad and the board is dead. Currently I am cooling down a working board in the fridge to see if that chip warms up when the board is first communicated with. On a hashing board it does not show any sign of excessive heat but at that point the copper planes in the board are conducting heat pretty uniformly. Thoughts on what that chip might be? One way to test my theory is to remove the chip from the good board and swap one of the suspect chips in. Got to be something :-) another board Title: Re: Fixing Dragonmint T1 miners: Thoughts and observations Post by: lightfoot on February 14, 2020, 06:27:23 PM Meantime I was able to trace down the heat a bit more closely. Looks like the PIIC chip next to that oddball one. Which makes more sense from a heat point of view.
Working on the board that sort of works: Oddly enough it does seem to have a bit of a problem with one of the chips (61). Need to map out the board and figure out how Inno counted the chips, could be on either end of the board. Code: Feb 14 18:19:23 InnoMiner cgminer[1306]: hub_cmd_read_register spi crc error ! I'm also figuring out the bit about the power supplies: It's not that the bigger supplies sag or have ripple it's that the board seems to want to see voltage sag to 11v in order to power up. The corsair is dropping voltage under load which seems to be what gets the board going. Weird. Title: Re: Fixing Dragonmint T1 miners: Thoughts and observations Post by: NotFuzzyWarm on February 14, 2020, 06:45:59 PM Quote I'm also figuring out the bit about the power supplies: It's not that the bigger supplies sag or have ripple it's that the board seems to want to see voltage sag to 11v in order to power up. Methinks it's Inno's idea of a soft-start circuit? ::)Title: Re: Fixing Dragonmint T1 miners: Thoughts and observations Post by: lightfoot on March 20, 2020, 01:21:43 AM Maybe. In the meantime I picked up a working miner to figure out what makes them tick and how to find the bad chip. First step was to run it, and sure enough it hashes at around 4th per board. So I pulled the board, and soldered wires onto one of the Clk signals to see if I can see it from my scope.
Good news: Although the clock does not come up with board power up it does display the signal. Bad news: With the scope connected the board throws SPI errors and shuts down. Guess it really has very little power going from chip to chip. This could be a good thing though, it means I should be able to see the clock signal for about 10 seconds without the board trying to hash. Which means I can test it with the board out of the box, as it won't generate much heat if it can't get nonces to process.... Still I now have a way to start testing the chains. Next step is to try a dead board and see if the clock is on chip 63, then bisect the board to see where it drops. More in a bit. Title: Re: Fixing Dragonmint T1 miners: Thoughts and observations Post by: ali farmer on July 12, 2021, 10:36:29 PM Hello my friend
Were you able to find a way to find the broken Dragon 16th chip? Title: Re: Fixing Dragonmint T1 miners: Thoughts and observations Post by: BitMaxz on July 12, 2021, 11:57:50 PM Hello my friend Were you able to find a way to find the broken Dragon 16th chip? I think the model name of dragonmint t1 chip is "DM8575" try to look on the marketplace like Alibaba or Aliexpress with this ASIC chip name or try to Google it with that keyword. Here look I found one but they are out of stock Code: https://d-central.tech/product/dm8575-replacement-chip-for-dragonmint-t1/ |