ASIC Repair - DIY Corner - Power Distribution
HTML-код
- Опубликовано: 8 сен 2024
- In this first video of the DIY Corner, we chat a bit about power distribution on a hashboard... because understanding makes us troubleshoot better and more efficiently!
Check out the wiki (work in progress): wiki.ASICRepair...
Reach me: On the Antminer Repair Discord server: / discord
or by email: info@ASICRepairCanada.com
Hi Spooky! I'm Enigma from the Discord server.
I just watched this video, understood a lot of things I didn't quite get before, and I wanted to say thank you for all this.
You explain things very well, it's easy to understand and it's a lot of valuable information.
What you are doing is fantastic for the community and for people like me who are just some months into ASIC repairing and need all information available, so videos like this are truly helpful.
Hope you continue with these, I'm willing to see all of them and learn.
Thanks again! And, you know, I'll probably be writing some things for your wiki, because that source of information for beginners will be also amazing.
Hey! Thanks for the kind words, it is appreciated. I like helping people so sharing what I know is the least I can do! I just wish I had more time to make videos and write the wiki!
God bless you for this information I am from Nigeria this is so helpful 👍
thank you sir very usefull content for miner repaire and maintenance thanks again
Thanks!!
Hi mate, great video. U got another subscriber from UK 🇬🇧.
Cheers! It is appreciated.
Salut!! Great video. I wish the boards I am working on were so clean. Mon Dieux. I am going to pass this video on to some users on the discord channel who have been asking me about where to start during their repairs. I always say, see if you have power from the mosfets, then start chip troubleshooting. I would add that more attention should be put on U7 which sends the power to the Mosfets Gate and is a major cause of power loss. Too many people start reprogramming the PIC and changing out Mosfets without looking at U7 the two diodes and the resistor leading to that gate.
Salut! Thanks for the comment. You are right that people tend to focus on the wrong things sometimes, probably because programming a PIC is easier than replacing a chip... There will be more videos coming, especially on the diagnostic side. I am just trying to lay the theory foundation first before going into it. Next video is probably going to be on signals...
Thank you sir!
Interesting video. Do you have a schematic of this board? That would be helpful in figuring out what you just said.
i learned a lot! thanks.
i have noticed S9 hashboards, some have boost circuit some dont, it baffles me a little, its my perception that maybe the boost is not really needed but very welcomed? shed any light on that?
I'm not sure that your description of how the 1.8V and 0.8V voltage domains is correct. According to schematics and a number of boards I've seen (taking the S17+ as an example), the 1.8V core domain voltage (and thus the 0.8V domain) is regulated down by the domain LDO straight from the 21V main supply voltage for all but the last three domains. The last three domains derive their 1.8V supply from the 24V boost regulator output. What you stated here is that the LDOs derive their input voltage from two domains prior to where they are, and I'm yet to see a board with that configuration - in fact, if you do have collateral showing this for models you've worked on, then It'd be really good to share that. As an aside, the domain voltages and power distribution of these boards are amongst the worst I've ever seen in high performance computing and FPGA designs in general, in over 25 years being a researcher and an engineer.
Thanks for the input. I'm not sure how your proposed layout would work.
1. On a S17+, you can easily see that the input pin of the 1.8v LDO of domain 1 is directly connected to the ground of domain 5 which is effectively the same as the positive side of the core voltage of domain 4 (3 domains higher, hence the need for a boost voltage for the last 3 domains). There is an internal trace, just check continuity and you will find it without a doubt.
2. We know that all domains use the same values for the components used to set the regulated voltage. If all LDOs were supplied by the board voltage, the output of each domain would be different relative to each domain ground since the ground of each domain varies relative to board ground.
3. The schematics call the LDO input VDD_IN where the board voltage is VDD_21V0.
@@asicrepaircanada159
Thanks for responding; I just love these discussions :)
Re: 1, without the benefit of X-Ray inspection of the internal layers, standard practice would dictate internal ground and power plains, as well as who-knows-what other layers inside; these will also double as thermal dissipation aids. I bet if we de-populated a board, we'd be able to identify both through and buried vias that achieve the routing. Not that I have any confidence in any schematics the Chinese produce, much less Bitmain, but I'm going by a) what the (silly excuse for) service manuals show and b) by measurements taken from repair jobs.
Re: 2, not sure I get what you mean here. All LDOs, save the last three, are fed by the 21V pre-boost. Using the same component values only makes sense, since you want the same voltage across the domains - 1.8 and 0.8. I've not done full testing on grounds, but suspect that grounds for the 1.8v and 0.8v domains are one and the same, hence you're able to derive the 0.8v straight out of it's 1.8v older brother.
Re: 3, you're neglecting the MP2019, and therein lies another ridiculous design quirk.
heating the chip from behind is a great idea. Do you have T17 boards to work on in a video?
Thank you so much for the valuable almost gave up fixing s17+ found out some a cap at chip side from domain to ground is shorted , also would like to ask if I don't only get 400mv in my domain how to I check ? I got 0.8 n 1.8 from ldo , chip problem ? Also some with low 0.9 but can find full asic but will get intermittent temp issue how do I bring up the domain voltage ? One of the chip resistance is low ?
That is a lot of question for a youtube comment my friend. Hop on the discord and you can get some help there. For domain voltages: depends on you are measuring it. Generally speaking, there are only chips and caps that can short a domain. If you have all chips I wouldn't start chasing that problem. Chips all have varying internal resistances....as long as it's not completely shorted, it's hard to tell if there's a real problem.
How much dc voltage is needed to test T17e hashboard ? Can we test the hashboard without its power supply ? Thanks for sharing.
Sorry just seeing your comment now... You need a power supply to test the board, either the one from the miner or a bench one. Somewhere between 16 and 21 volts will work. If you are using the same tester that I am using, you will need a bench power supply since it cannot control the miner PSU. Hope this helps!
So between the voltage domains we should have the same resistor measurement?
In a perfect world, yes. In the real worls, it should be close. The resistance of one domain is really the internal resistance of all the chips on that domain in parallel. You just want to make sure of of them isn't shorted. Also, a short on a domain could be one of the caps, typically on the back of the hashboard, although I have only seen that once.
It's like saying: should we have the exact same domain voltage for all domains. Ideally yes but close enough is good enough. Hope this helps.