GRIDSEED G-BLADE Overclocking 7Mh/s, improvements and repair

[Blisk]

wasn't there on old baldes just 2 ferrites?
I see here now 6 resistros 0 ohm I suppose?
They are current asorbers because of high current here.

[wolfey2014]

this  IRFB7437 can replace all 3 chips?  IRFH5300/5301/5302

Why than they use two  IRF5300
and one IRFH5301?

what is difference with IRF1404 and IRFB7437? less power?

I found IRFB7437 for 1,17€ pcs

here is pic what I think will do with to-220 mosfet
this is look from a side

The orientation of the pqfn isn't the same on the 3 ones, so you won't be able to do it like this.
plus the legs of the TO-220 will be too short to bend them back to the other side

Since there will be 3 TO-220 next to each other, I'll probably go for a heatsink for the 3 of them and I'll let it stay on the PCB without soldering it, maybe glued with non conductive glue.
Look inside of a computer PSU to have an idea of what I mean, except in the PSU, the heatsink is soldered to the ground of PCB.

I have a few dead old PSU that I can take the heatsinks off for this project

Yo J4! How ya doin?
I see you've been busy in your mad scientist lab hah hah!  
Just my 2 centerinos: I don't recommend adding FETs or messing with the overall design.  Doing so will add leakage inductance to the switching circuit and could cause failure.  Installing bigger MOSFETs might (/might/) be an option, but if the gate charge (basically gate capacitance) is higher than those installed it may go !bang! as a protest. Yes! More MAGIC BLACK SMOKE! WOO HOO!  
WoOf

Edit: Another thought.....You might be able to use a 'separate' gate driver to switch the added FETs'. They MUST absolutely be 100% in sync with the main FET's gate driver which is where you'll tap the drivers gate signal from.
Maxim makes some great gate drivers if you're interested in trying this approach. Simply Google maxim gate driver...Mouser carries them.
Again, this is just an off the cuff idea without putting much thought into it but it's been my experience that the best way to drive a gate is with a gate driver, not direct output of the actual driver so you achieve nice clean square waves where you need them the most.
I presume you have the requisite test equipment including a good high speed analog or digital CRO. Gewed luck, doctor!

[Blisk]

ferrites are for leaking inductance.

[GenTarkin]

I put a small heatsink on the backside of the board, right behind the DC powerjack .. over the area where the hottest ferrite beads are. They seemed to have cool down significantly, so has the general area of the board where the small heatsink is at.

Also, Im waiting for someone to try the vdroop mod =P ... not knowing much about PCB's, I dont know if it would have adverse affects or kill something =/
If I knew, Id be the first to try it =P

Small update on my bridge mod for 2x gblades ... 913mhz @ .35% HW overnight , avg hasrhate 12.36mh

I know that if we could address the vdroop problem, if all chips saw 1.3v ... hashrate would be decently higher, w/o much additional load on the VRM's ...

[mangodream]

Hi there ! i think there's a problem on my 4 x G-Blades, dunno if it's the diff @512, or cgminer itself (i wanna change to cpuminer 1.0b, but it's not really compatible with Betarigs, since cpuminer and cgminer don't shows dead/alive pools the same way), or anything else  ...
I'm using a 80+ silver PSU, FSP brand, Win 7 x64.


(click here if image isn't showing)

[GenTarkin]

Hi there ! i think there's a problem on my 4 x G-Blades, dunno if it's the diff @512, or cgminer itself (i wanna change to cpuminer 1.0b, but it's not really compatible with Betarigs, since cpuminer and cgminer don't shows dead/alive pools the same way), or anything else  ...
I'm using a 80+ silver PSU, FSP brand, Win 7 x64.


(click here if image isn't showing)

Um.... what do you think the problem is? All looks normal to me...

[mangodream]

Different pools only reports ~ 13-15 Mh/s, i have also 40 Mh/s all in GPUs and the hashrate reported by the pools is pretty accurate with them...

EDIT : What about the HW errors, what's the "normal" rate ?

[J4bberwock]

Different pools only reports ~ 13-15 Mh/s, i have also 40 Mh/s all in GPUs and the hashrate reported by the pools is pretty accurate with them...

EDIT : What about the HW errors, what's the "normal" rate ?

Hi,

you are around 3.2% HW errors.
186*512/(2978896+16096)

I like to stay under 0.5%

High difficulty can give you some strange hashrate pool side.
If you can, check the 24h average.

For 7.2Mh/s, betarigs always show between 6.9 and 7.4Mh/s 2 hours average

[GenTarkin]

Different pools only reports ~ 13-15 Mh/s, i have also 40 Mh/s all in GPUs and the hashrate reported by the pools is pretty accurate with them...

EDIT : What about the HW errors, what's the "normal" rate ?

Nominal HW rate is 2% and below. Anything above is a negative return.

[mangodream]

Thanks for the advices guys, resolved the issue.

So, i downloaded Hashra image for Blades, put it on my Rasp' and all is working like a charm with BFGminer 

[schnebi]

Hi there ! i think there's a problem on my 4 x G-Blades, dunno if it's the diff @512, or cgminer itself (i wanna change to cpuminer 1.0b, but it's not really compatible with Betarigs, since cpuminer and cgminer don't shows dead/alive pools the same way), or anything else  ...
I'm using a 80+ silver PSU, FSP brand, Win 7 x64.


(click here if image isn't showing)

Hey, i tried to run multiple G-Blades on Win 7 as well. But it only works with one Blade at once. As soon as i connect the second Blade cgminer crashes at atartup, or isn't finding any of the Blades. I am plugging both G-Blades directly into the Computer USB Ports (so 4 Ports). Have you encountered similar Problems? Or maybe do you think it will work if i use a powered USB-Hub?

[Blisk]

first connect blades and than run cgminer.
I use cpuminer it works much better

[schnebi]

first connect blades and than run cgminer.
I use cpuminer it works much better

Of course i connect the blades first and then run cgminer. But thanks for the hint with cpuminer. I haven't tried tis yet.

[schnebi]

first connect blades and than run cgminer.
I use cpuminer it works much better

Of course i connect the blades first and then run cgminer. But thanks for the hint with cpuminer. I haven't tried tis yet.

Thank you. I tried using cpuminer and a powered Hub and its working 

[mangodream]

Yeah you MUST use a powered HUB, and cgminer is a shit on windows for blades... Now running 5 blades on Hashra img, individual clock tuning, smooth interface, nice as fuck !

[Blisk]

Yeah you MUST use a powered HUB, and cgminer is a shit on windows for blades... Now running 5 blades on Hashra img, individual clock tuning, smooth interface, nice as fuck !

I don't use raspberry pi.

[J4bberwock]

Quick update after reading hundreds of pages on buck converters, it looks like upper and lower gate mosfets can't be tne same ones.
There must be difference in a few parameters to allow the gate to work.
I'll give more details once I understand it correctly.

In the mean time, basic powerboard redesign has started.
I'll try to use mostly through hole components and when it isn't possible, pqfp of similar components with legs to allow easier rework/replacement.

buck converter will be tps40140, well documented and that will allow any voltage from 0.7 to 2.2v easily.
That will allow the powerboard to be re-used for newer asics later.
That mean independant powerboard that will connect to the chips board and probably independant communication board for usb.


I've also blown a mosfet on one of my blades, as usual, it's the upper gate one that is alone. I suspect that shrinking the design from 2 upper + 2 lower to 1 upper + 2 lower fet puts stress on the single when overclocked. I'll solder a replacement one tomorrow. IRFH7004 doesn't seem to work in this place or something else is shorted. I'll try irfb7437 that is proved working on a pod or I'll replace with the newer components used by gridseed as I have some.
experiments so far:

Upper gate, pin 19 -> 5302 or CSD16321 or irfb7437 (works on pod) irfh7004 doesn't work on blade?
Lower gate, pin 2 -> dual 5300 or dual CSD 17556

I've found alternate soldering points on the blade for the gate of lower fets that will make experiments easier with fets soldered outside of the blade

[Blisk]

So you suggest to use a  CSD 17556 with IRFB7437?
is there any to220 instead of CSD 17556

I still didn't overclock mine because I have strange things with my not overclocked blades.
Both stops to work after fev hours, cgminer works but hashrate drops to under 1 Khash.
Had someone the same thing? What can be wrong?
I already changed PSU

[J4bberwock]

So you suggest to use a  CSD 17556 with IRFB7437?
is there any to220 instead of CSD 17556

I still didn't overclock mine because I have strange things with my not overclocked blades.
Both stops to work after fev hours, cgminer works but hashrate drops to under 1 Khash.
Had someone the same thing? What can be wrong?
I already changed PSU

Could be overheating on the power mosfets.
when too hot, there is continuous connection between source and gate, same as when they are dead.
even when they are cooled down, the blade needs restarting.
try heatsinks on the fets and/or a secondary fan blowing by the side of the blade on the powerboard, just to check.
Try cpuminer.
cpuminer solved every trouble I had with cgminer

[Blisk]

yes I tried cpu miner from link you post It makes a little lower hashrate but works great. But not for blades because works fev hours and than stops to mine. my blades are in room at 20 oC so overheating can't be reason and it must not happend on not overclocked ones.
I will try with some extra fans blowing from a side but I think it won't solve a problem, why is this happening now after more than a month working ok? And I didn't do any changes.