Pages:
Author

Topic: ANTMINER S5: 1155GH(+OverClock Potential), In Stock $0.319/GH & 0.51W/GH - page 19. (Read 450939 times)

hero member
Activity: 588
Merit: 500
"String" or "Module" refers to the way the Chips are powered. Might be better to think of it as Series or Parallel. In the S1 & S3 the chips are powered in Parallel. In practice it is slightly more complex with the 12V powering a number of Buck Converters that drop the 12V to the voltage required for the Core (0.7V -1.0V) Each Buck Converter then powers a number of chips, 8 in the S1 and 2 in the S3.

In the S5 & S7 the chips power is connected in Series (Just like old style Christmas Tree Lights) So in the case of the S5 there are 15 pairs of chips in the chain, 12V /15 = 0.8V / chip. All very neat and efficient as there are no Buck Converters needed, saves money and is very efficient, But....

If one of the chips in the chain fails then at best you loose all the chips after that one and often you loose the whole chain, dependant on the failure mode.  If a chip goes open circuit (Like a bulb blowing in the Christmas Lights) as the voltages to the chips are messed.

Also worth remembering that as well as the power all of the data & control signals in all of the Miners is chained through the chips. It seems that sometimes even if a chip is not hashing it continues to pass through these signals enabling the chips further down to keep operating , however with more major failures downstream chips will stop working.


Rich

It would be interesting to see a design based on rectified and smoothed 120V. No external PSU needed Smiley

Just as an aside on the Christmas lights, I have those ones where a dead bulb does not kill the lights. Problem is, it does that by shorting the connection through the bulb, leading to increased voltage to the other bulbs and more rapid failure. By the time you get to three dead bulbs in a row, it's game-over for that segment. Not sure this is relevant but I felt like sharing Smiley

If you rectified & smoothed 120V Mains would give 168V. You are going to need a very long string, like 210 Chips. Could be done but the strings work better with more than one chip at each stage. S5 has 2, S7 has 3. So that means 420 or 660 so not very practical, and without an isolating transformer, which will have losses, not very safe.  Smiley

Yes the bulb failure mode you describe can also happen in a Miner if a chip goes short circuit...  Sad

Rich
legendary
Activity: 2478
Merit: 2151
1RichyTrEwPYjZSeAYxeiFBNnKC9UjC5k
"String" or "Module" refers to the way the Chips are powered. Might be better to think of it as Series or Parallel. In the S1 & S3 the chips are powered in Parallel. In practice it is slightly more complex with the 12V powering a number of Buck Converters that drop the 12V to the voltage required for the Core (0.7V -1.0V) Each Buck Converter then powers a number of chips, 8 in the S1 and 2 in the S3.

In the S5 & S7 the chips power is connected in Series (Just like old style Christmas Tree Lights) So in the case of the S5 there are 15 pairs of chips in the chain, 12V /15 = 0.8V / chip. All very neat and efficient as there are no Buck Converters needed, saves money and is very efficient, But....

If one of the chips in the chain fails then at best you loose all the chips after that one and often you loose the whole chain, dependant on the failure mode.  If a chip goes open circuit (Like a bulb blowing in the Christmas Lights) as the voltages to the chips are messed.

Also worth remembering that as well as the power all of the data & control signals in all of the Miners is chained through the chips. It seems that sometimes even if a chip is not hashing it continues to pass through these signals enabling the chips further down to keep operating , however with more major failures downstream chips will stop working.


Rich

It would be interesting to see a design based on rectified and smoothed 120V. No external PSU needed Smiley

Just as an aside on the Christmas lights, I have those ones where a dead bulb does not kill the lights. Problem is, it does that by shorting the connection through the bulb, leading to increased voltage to the other bulbs and more rapid failure. By the time you get to three dead bulbs in a row, it's game-over for that segment. Not sure this is relevant but I felt like sharing Smiley
sr. member
Activity: 287
Merit: 250
Global economic crisis? i hold my bitcoin..
help what happen with my S5
Code:
[23330.314174] Send new block cmd
[23330.693485] jiffies{11792321}dev->cgminer_start_time{9318878}time_elasp_ms =4830943ms dev->total_nonce_num{16040}
[23330.704256] bgNonce_average{125}
[23330.707641] chain0 Chain_nonce_nu[6811] Chain_nonce_nu_last[6811]
[23330.714027] UTC time :2015-10-13 14:36:36
[23330.730961] remap chain1(hardware chain0) power err
[23330.736093] bgNonce_average{125}chain_nonce_nu{6811}last{6811}
[23330.742200] Send new block cmd
[23330.745892] set_baud cmd_buf[0]{0x86}cmd_buf[1]{0x10}cmd_buf[2]{0x1a}cmd_buf[3]{0x17}
[23330.754097] send BC data:
[23330.754097] 0x0000: 0x03 0x00 0x00 0x05 0x86 0x10 0x1a 0x17
[23330.771084] send_to_fpga_work in queue
[23330.775029] send BC data:
[23330.775029] 0x0000: 0x03 0x01 0x00 0x05 0x86 0x10 0x1a 0x17
[23330.788667] send_to_fpga_work in queue
[23330.800381] send_to_fpga_work in queue
[23330.812098] send_to_fpga_work in queue
[23330.821860] send_to_fpga_work in queue
[23330.833596] send_to_fpga_work in queue
[23330.845310] send_to_fpga_work in queue
[23330.858979] send_to_fpga_work in queue
[23330.868736] send_to_fpga_work in queue
[23330.882409] send_to_fpga_work in queue
[23330.894144] send_to_fpga_work in queue
[23330.905850] send_to_fpga_work in queue
[23330.915607] send_to_fpga_work in queue
[23330.927333] send_to_fpga_work in queue
[23330.939054] send_to_fpga_work in queue
[23330.950771] send_to_fpga_work in queue
[23330.954728] Detect device for anyone power err
[23330.959377] send BC data:
[23330.959377] 0x0000: 0x03 0x02 0x00 0x1a 0x89 0x10 0x1f 0x00
[23330.974197] send_to_fpga_work in queue
[23330.978937] reset hash asic
[23330.981869] clear FPGA nonce buffer
sometimes, hashrate is very low (half of normal or just 0.00)
i change the PSU, cable, etc still same.

what about :
[23330.730961] remap chain1(hardware chain0) power err
and
[23330.954728] Detect device for anyone power err

how to fix it up?
thank
member
Activity: 68
Merit: 10
"String" or "Module" refers to the way the Chips are powered. Might be better to think of it as Series or Parallel.

Thanks for the explanation Rich.  Series & parallel quickly relate to my electronics experience.  "String design" sounds all mysterious and high-brow.  Wish people that like that phrasing would instead say "string power design", so some like me don't think it's referring to the logic of the processing that is the string design.
hero member
Activity: 588
Merit: 500
"String" or "Module" refers to the way the Chips are powered. Might be better to think of it as Series or Parallel. In the S1 & S3 the chips are powered in Parallel. In practice it is slightly more complex with the 12V powering a number of Buck Converters that drop the 12V to the voltage required for the Core (0.7V -1.0V) Each Buck Converter then powers a number of chips, 8 in the S1 and 2 in the S3.

In the S5 & S7 the chips power is connected in Series (Just like old style Christmas Tree Lights) So in the case of the S5 there are 15 pairs of chips in the chain, 12V /15 = 0.8V / chip. All very neat and efficient as there are no Buck Converters needed, saves money and is very efficient, But....

If one of the chips in the chain fails then at best you loose all the chips after that one and often you loose the whole chain, dependant on the failure mode.  If a chip goes open circuit (Like a bulb blowing in the Christmas Lights) as the voltages to the chips are messed.

Also worth remembering that as well as the power all of the data & control signals in all of the Miners is chained through the chips. It seems that sometimes even if a chip is not hashing it continues to pass through these signals enabling the chips further down to keep operating , however with more major failures downstream chips will stop working.


Rich
legendary
Activity: 1302
Merit: 1068
Really wish you could just bypass bad sections, ugh. But i guess they were designed to be cheap to make, not easy to maintain, repair or be resilient.
You often lose modularity in chain design as you drop entire chains rather than modules (or sub modules). The benefits are still worth it though.

I'm really curious about the differences and just googling string design / vs modual circuit design does not yield anything relevant. (Whoot string theory results?)
I'm suspecting thats not the proper terminology?

If i understood the differences then maybe i could bring up the bad instead of just bashing on the S5.
legendary
Activity: 1666
Merit: 1183
dogiecoin.com
Really wish you could just bypass bad sections, ugh. But i guess they were designed to be cheap to make, not easy to maintain, repair or be resilient.
You often lose modularity in chain design as you drop entire chains rather than modules (or sub modules). The benefits are still worth it though.
hero member
Activity: 798
Merit: 1000
I have 2 pcb defective S5 today I try to study the 18-pin connector, which seems stained, and short.
legendary
Activity: 1302
Merit: 1068
I am going to keep working on my boards and will probably get there in the end. One approach I will probably take if I cannot get the hang of replacing chips is what you are saying and that is to bypass the faulty stages.

This is not quite as simple as it sounds as the chips and level shifters need to be removed then the signals connected directly to the next stage and the voltage reduced to allow for the shorter chain. It does however require the first stage to be operational as that has the oscillator.

If I can get to grips with repairing hash boards and also ideally making pre V1.91 boards undervolt I will post up all the information.

However If it fails, and will go with one board anyway, with Sidehacks latest recycling project which is a very attractive alternative.  Smiley


Rich

Then i wish you good luck.

I invite you to start a thread and act as a journal, so that people can keep track, learn and give feedback.

Personally it would be of interest to keep track of.

It's a bit bothersome for me since I have plenty of time to put time on repairing this board, but i'm afraid without some sort of map that show me what everything should be at, i can't figure out what the problem is. And once i figure it out, i don,t even have the pieces to repair the board.

In your case if you had several dead boards. You could more reliably just swap bad parts and probably be able to get a high % of dead board working by cannibalizing the rest of the boards.
hero member
Activity: 588
Merit: 500
I am going to keep working on my boards and will probably get there in the end. One approach I will probably take if I cannot get the hang of replacing chips is what you are saying and that is to bypass the faulty stages.

This is not quite as simple as it sounds as the chips and level shifters need to be removed then the signals connected directly to the next stage and the voltage reduced to allow for the shorter chain. It does however require the first stage to be operational as that has the oscillator.

If I can get to grips with repairing hash boards and also ideally making pre V1.91 boards undervolt I will post up all the information.

However If it fails, and will go with one board anyway, with Sidehacks latest recycling project which is a very attractive alternative.  Smiley


Rich
legendary
Activity: 1302
Merit: 1068
Yes that board looks ok, however dependant on how it failed you it may not have overheated. I bought an S5 from ebay that had burnt itself up on losing the internet connection and the fans had stopped. It had got sufficiently hot that the side plastic pieces were very distorted. Here is the picture.



You can see the darkening on the PCB around all the chips on the right, some of that has also been made slightly worse by me removing chips.

If anyone has any recommendations on what, reasonably priced, equipment is best for removing and replacing this style of chip that would be very helpful.  Smiley I have a hot air blower which has been fine for 64 pin QFP, but I am finding that these chips need to get very hot to remove & am concerned that I am damaging them in the process.


Rich

Really wish you could just bypass bad sections, ugh. But i guess they were designed to be cheap to make, not easy to maintain, repair or be resilient.

I guess Sidehack would be a good one to tell you this information. Making a thread just for this fiddling would be nice but i'm afraid there's no much more i can do with my knowlewdge.
legendary
Activity: 2478
Merit: 2151
1RichyTrEwPYjZSeAYxeiFBNnKC9UjC5k
I kind of feel like i don't want to spend money on getting more stuff than i already have on hand since i have no guarantee that this will allow the board to function. Hence, kind of wasted money.

Thermal paste is like $1 on ebay including shipping. $5 if you want to go for the fancy stuff.
hero member
Activity: 588
Merit: 500
Yes that board looks ok, however dependant on how it failed you it may not have overheated. I bought an S5 from ebay that had burnt itself up on losing the internet connection and the fans had stopped. It had got sufficiently hot that the side plastic pieces were very distorted. Here is the picture.



You can see the darkening on the PCB around all the chips on the right, some of that has also been made slightly worse by me removing chips.

If anyone has any recommendations on what, reasonably priced, equipment is best for removing and replacing this style of chip that would be very helpful.  Smiley I have a hot air blower which has been fine for 64 pin QFP, but I am finding that these chips need to get very hot to remove & am concerned that I am damaging them in the process.


Rich
legendary
Activity: 1302
Merit: 1068
OK I have reordered the measurements to represent how they are connected on the board. The chain bottom is connected to 0V is C17, they then go anticlock round the board ending up at C2 which is connected to +12V.

+12V

C2 5mV
C3 4.4mV
C5 0.885V
C6 1.08V

Right hand top corner of board

C7 0.95V
C8 0.995V
C10 0.91V
C11 0.97V
C12 0.9V
C13 0.87V
C14 0.951V

Left hand top corner of board.

C15 1.015V
C16 1.11V
C17 0.45V

0V


So the obviously odd voltages are those on the top 2 stages (C2 & C3) which are very low. This will probably mean that one or both chips at those stages are low resistance, probably because one or both of the Hash Chips on those stages are short circuit? Does the board look slightly blackened around those stages? C17 is also low, indicating perhaps that one of the first stage chips is also faulty? However this is just a guess....

Rich


http://imgur.com/a/q0xgX

I would say no burn marks, judge by yourself. The black smudges are thin layer of cooked thermal paste i believe. They were not there when i cleaned it. The board did end up at 205C~ when i tried to oven cook it.

My camera derp out and lose focus on the chip top side, which is weird, but anyways, even though that side is blurry, its still decent enough to see i guess.
hero member
Activity: 588
Merit: 500
OK I have reordered the measurements to represent how they are connected on the board. The chain bottom is connected to 0V is C17, they then go anticlock round the board ending up at C2 which is connected to +12V.

+12V

C2 5mV
C3 4.4mV
C5 0.885V
C6 1.08V

Right hand top corner of board

C7 0.95V
C8 0.995V
C10 0.91V
C11 0.97V
C12 0.9V
C13 0.87V
C14 0.951V

Left hand top corner of board.

C15 1.015V
C16 1.11V
C17 0.45V

0V


So the obviously odd voltages are those on the top 2 stages (C2 & C3) which are very low. This will probably mean that one or both chips at those stages are low resistance, probably because one or both of the Hash Chips on those stages are short circuit? Does the board look slightly blackened around those stages? C17 is also low, indicating perhaps that one of the first stage chips is also faulty? However this is just a guess....

Rich
legendary
Activity: 1302
Merit: 1068
OK so the indutor has shifted but must still be soldered to the pads. Also worth checking the Core Voltage on each of the stages. Easiest way to do this is at each stage to measure across the Black Capacitor (C17 on the first stage) Should be about 0.8V on each stage.

However I have been trying to repair a board without success... So although checking those voltages is instructive I do not have the experience yet to help you repair the board. Balance of probability is that one of the Hash Chips is faulty.

Rich


Okay i have checked every black rectangle between the chips. Lets get educational then.
I have labeled them according to what seem to be their name from top left, normal reading order;

C14 0.951V
C15 1.015V
C16 1.11V
C17 0.45V
C2 5mV
C3 4.4mV
C13 0.87V
C12 0.9V
C11 0.97V
C10 0.91V
C8 0.995V
C7 0.95V
C6 1.08V
C5 0.885V

I suppose this is easy to compare to a healthy S5 board.
hero member
Activity: 588
Merit: 500
OK so the indutor has shifted but must still be soldered to the pads. Also worth checking the Core Voltage on each of the stages. Easiest way to do this is at each stage to measure across the Black Capacitor (C17 on the first stage) Should be about 0.8V on each stage.

However I have been trying to repair a board without success... So although checking those voltages is instructive I do not have the experience yet to help you repair the board. Balance of probability is that one of the Hash Chips is faulty.

Rich
legendary
Activity: 1302
Merit: 1068
The reason you cannot see L1 is because the way the inductor has moved is to cover the silk screen.  Smiley Best thing to measure is from ground (The 3 easy to get at pins on the Molex connector). And the bottom of R213 which is where you have put your red cross. Should be 14V and is the supply to the LDO regulators for the last 3 chips in the chain.

Rich

Ah gotcha. It reads 14.11V. I've checked a few whatever they are between the chips. Some of them are at 0.4V some are near 0 like 45 mV some are 0.9-1V.

I'm not sure you can figure out what the problem is and help me fix it but if it happens, i'd gladly scrounge up a small bounty that i can for it.
hero member
Activity: 588
Merit: 500
The reason you cannot see L1 is because the way the inductor has moved is to cover the silk screen.  Smiley Best thing to measure is from ground (The 3 easy to get at pins on the Molex connector). And the bottom of R213 which is where you have put your red cross. Should be 14V and is the supply to the LDO regulators for the last 3 chips in the chain.

Rich
legendary
Activity: 1302
Merit: 1068

Thanks for posting that I did not know of that revision.  I was thinking something different.

The odd lock of many screws being different and look made me thing it was possibly a mod - https://i.imgur.com/v2poeaL.jpg?1 .  But appears just a mess of different screws no fun story I guess then.

Looking at the picture L1 on the 14 Step up Converter (Top Right Corner) looks very displaced from it's pads. Would be worth checking that there is 14V coming from it? You can measure between Ground and the bottom end of R213.

Rich

I don't see a L1 written so i'm maybe blind or i have no idea what L1 is. Anyways i guess its the obvious thing i first noticed that is so obvious i was sure it couldnt be the problem;

https://i.imgur.com/1DJmNNU.png

By putting ground on either black X and + on red X i get 2.2v.

Also at big the black rectangle next to it, from both its side it look like 12v go through and i mesured voltage accross the black rectangle all over the board. I can swear the first time i checked there was no voltage at all accross the board. Now its 0.91-1 volt.
Pages:
Jump to: