update on the blade, I can confirm that irfh7004 don't work on upper gate when you have dual 5300 on lower gate.
but CSD16321 works with dual 5300 on lower gate. I received a pod with this configuration, and I tried it on the blade.
It's happily hashing again.

next step will be external quad parallel fets on each gate to feed the up1509 before trying complete external powerboard thaqt could be piggybacked to a blade

we still need a good TO-220 candidate for lower gate

as a side note, if you do replace the power plug with screw terminal like me, put them on the back of the board, that will allow for more space to solder the upper gate mosfet when it will die.
I've melted half the screw terminal with my hot air station when desoldering and resoldering the fet.
It will also be easier to screw/unscrew.

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

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

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

Do you have CSD 17556 for the other ones too or still the old IRFH ones?

ferrite are still alive at 1063Mhz for weeks on my first blade, but I have forced cooling on the powerboard area.

Here is the newer version from Gridseed with ferrite replaced by 0 resistors and the vrm replaced by gridseed too

I can't see any reason for choosing 2 different ones, but one is for a channel, the second one is for a different channel.
Original design used 2 per channel and the same ones
UG and LG going to pin 2 and pin 19 of up1509

All fets were named 09N03 in the schematics.

here is the datasheet for it
http://www.infineon.com/dgdl/IPD09N03LA_Rev2.1_G.pdf?folderId=db3a304412b407950112b408e8c90004&fileId=db3a304412b407950112b429dca44214

On the 5 chips pod, the design shows only 1 for UG (pin2).

So we have
Upper gate, pin 19 -> 5302 or CSD16321
Lower gate, pin 2 -> dual 5300 or dual CSD 17556

We should be able to use the same fets everywhere.

I'd simply use the reset switch for the USB
It should be enough

A switch to disconnect your resistor from the board should also work.

Up to 50K, you should be fine if you use screw terminals (easier than direct wiring), I've already tested it.
47K with the standard connector, and the connector that plugs to the board have melted in a matter of minutes.
Up to 50k, the worst thing I've seen is the board stopping hashing after 1 or 2 minutes because the VRM went into heat protection and cut the line until they cool down.

IRFB7437 works, 1404 is ordered, but if you can have both at the same price, take the one that works for now.

I'll test the blade with IRFB7437 and 47K resistor in a day or 2

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

TO-220 replacement should be done safely with IRFB7437 in place of IRFH5300/5301/5302
I'll also try IRF1404 that I can get way cheaper and is easier to source.

If I understant you correctly, you will glue the heatsink on board "A" next to the USB chip, over the empty header holes, to screw on it the TO-220 connected to board "B"
Am I right?
It could work, but you will need precise work because the solder pad for the Gate of the TO-220 is small and the soldering can easily break if stressed.

He has blown ferrite beads, so he sent the pod to me to try to revive it.

one power mosfet was shorted, so I replaced it with IRFB7437.

You can see on the picture that when exploding, the ferrite beads have badly damaged the PCB near the solder points for the fan.

If you do TO-220 conversion, beware the Gate/Source/Drain could be misplaced.

So here comes Frankenpod







It works just fine, I disconnected it after a few seconds to avoid overheating the GC3355, but even the 7437 was hot. It will need active cooling with dissipator.

it is pin to pin compatible, higher amps, higher amps at high temperature, so we should get better overclocking with it.
But we'd better go for to-220 packages since we have much more room available on the blades.
I'm trying to revive CryptKeeper's pod with IRFB7437.
If it goes back to life, I'll try replacing all the mosfets with TO-220 on one PCB of my blades.

BTW, why did they decide to go for 2 different types?

IRF00/5301/5302 or CSD17556/16321 do have almost the same datacheets.
I'm not an expert, but they are all N-channel power mosfets with almost the same values.

Hi Wolfey,

I'm still waiting for my no clean SMD flux paste to arrive.
I received some heatsinks for the back of the PCB, I should receive my spacers for wide body at the beginning of next week.

I do expect that we are reaching the limits on the 2x IRF5300 and 1x IRFH5301 or 2x CSD17556 and 1x CSD16321 because they can stand a much lower constant load while they are hot than when they are cold.
So overcurrent built in protection closes the gate.

IRF7004 is a little better, but not that much.
Best bet would be something like this one as direct replacement
http://www.st.com/web/en/resource/technical/document/datasheet/CD00291104.pdf
The worst case will be that the up1509 reaches its limits, but as it's far from hot, it should still be ok for some extra amps.

Anyway, piggyback powerboard could be done, I'm still working on something.

Hi, I don't think it's woth the change. They still shut down when using 47K resistor.

I'll upgrade to IRF7004 when I'll get my new SMD no clean solder paste

Hi, I haven't changed IRF5300 and IRF5302 on my blade.
2x CSD17556 and 1x CSD16321 were on a new blade I just received and I'm voltmodding too.
They are pin to pin compatible.

I ordered IRF7004 (expensive) that I received today to replace the 5300-5302 ones on the "old" blade.

1100 MHz, 3.74Mh/s for the first board. I'll try higher resistor and try the second board one once I add the side fan to cool the VRM
1063MHz gives 3.62Mh/s

with 44k resistor, the board looks stable at 1100 MHz, 1113 gave a few HW errors.

Even my 0.75mm² wires are getting warm 
I'll check power draw tomorrow

but still hot on the powerboard, it will need side cooling even with the monster fan because the choke hides the VRM from the airflow. I'll try high heatsinks from accelero L2 kits on the VRM to bring some airflow on the heatsinks

44.2k resistors added, screw power terminal, plus all the extra heatsinks except the ones on the back of the PCB at the moment.
New fan 120x120mm ultra kaze by scythe 130CFM ready to be added in place of the stock one with wide body for the blade.

Pics and results soon.

It's the same layout, simply ferrite replaced with 0 resistors and VRM changed for maybe better ones pin to pin compatible.