And here is the firmware needed for the G-Blade

http://dl.free.fr/vLuvCPiD9

I'll make a quick detailed "how to" later with pictures.

Material needed:
-an USB to UART converter (CP2102 based is working for me)
http://www.aliexpress.com/item/Free-Shipping-1PCS-CP2102-Serial-Converter-USB-2-0-To-TTL-UART-6PIN-Module/623537804.html?s=p
-2mm spacing pin headers will make things easier
http://www.aliexpress.com/item/10pcs-2mm-40-Pin-Male-Single-Row-Pin-Header-Strip/1982700728.html

Software:
STM32 flash loader demonstrator
http://www.st.com/web/en/catalog/tools/PF257525#

Connecting the CP2102 to the Blade:

With the Blade unplugged from the USB

-Connect pin 8 (Ground) with pin 6 on the 8 pins header next to the STM32 chip.
Pin 8 is the closest to the JTAG 10 pins header
Pin 1 is the closest to the hashing chips.

-Connect pin 1 with pin 5

Now, your STM32 is configured to allow read and write of the flash.

Next, connect the pin 3 (RX) of the Blade header to TX on the CP2102 header
And connect pin 4 (TX) of the RX on the CP2102 header.

You can now plug the USB from the Blade to a computer
And the CP2102 to your computer.

The Blade will be found as an "unknown device".

The CP2102 will be assigned a port number that you will find in the hardware manager.
If it doesn't, you might need to install the CP2102 drivers from Silabs

http://www.silabs.com/products/mcu/pages/usbtouartbridgevcpdrivers.aspx

And uninstall anything you've done with Zadig earlier.

Start the flash loader demonstrator
and change the settings to match your com port
Baudrate: 115200
Parity: Even
Echo: disabled
Timeout: 10

Click "next"
You might be asked to unlock the device.
Do it, and don't bother with the note regarding erasing data since we will write a new flash anyway.

Your chip might be detected with 256k flash and say that "Target is readable"
All is good.

Select the target as 256k flash in the next window.
Click "next"

Choose the flash file to use, check the "verify" box, and proceed with the flash.

After about 1 minute, you should be good with your blade and happily hashing again.

You can send a few tips to my BTC or LTC address if this helped you.

It took me some time to do it, but here is the file someone would need to flash the Gridseed G-Blade if the STM32F103RCT6 failed or was erased for some reason.

http://dl.free.fr/vLuvCPiD9

I'll post a quick "how to" in my G-Blade 7MH overlocking thread

Quick update.
The stencil went quite good, the first board is soldered.

The bad news is that I managed to make 2 midstakes in the design, the USB hub isn't recognized in windows.

But the "useless" boards will allow me to solder the necessary components and to connect wires between boards to have a working ugly prototype and validate the new design.
I already updated the schematics and I'll try to have something working today or tomorrow.

and here is the stencil mask that I'll use tomorrow for toner transfert method.
It will be my first trial at homemade stencil.

Once finished, it will need a case too.

I'm looking forward to see it.
You already tested the design, so you won't discover if it works or not as I do.

Once tested and working, why not.
The idea was to offer assembled boards, but kits can be done for those who'd like to try assembling.
SMD soldering for some parts might not be easy if you aren't used to do it.
I'll probably get/make a stencil once the design is validated.

It's time for some soldering job.
 

No more stock on DPS 2000 and DPS 1200.
That was fast.

Next arrival of boards at the end of this week.

As usual, custom order for PSUs not listed are available with a minimum order quantity of 5 boards

If you need them for ASICs, I can offer Server PSU + breakout boards.
I still have a few DL580 G3 PSUs 1300w, quite loud, but 20€ each.
Small breakout board with 6x 6 pins PCI-E cables is 34€

Shipping within Europe will be around 15-20€

If you need ATX PSU, sorry, I can't help.

Yes, they are the good ones.
I will be short of cables for a week, and you might want to try and find the PSU locally.
My supplier wasn't reliable on them.
I've found another one, but each PSU would be 35$

A set with 1 PSU and 1 board would be 70$
For the cables, ask klondike-bar, he should have some stock.
And I remember you, I was going to throw you a message saying that the boards are in stock for when you'd need them.

If the customs could stop playing tennis with the package between Shenzhen/Hong-Kong, the PCB would arrive on Tuesday

I still have a few that are most likely reserved, next batch within 1 week.

And here is the DPS-1200 FB Common Slot board

I've got a custom order for the DPS-2000 BB, so added it to the available for order.
Still a few in stock.

here is the DPS 2000-BB with 6 fans headers to allow 3 fans to be installed on each side in push/pull setup.

That would mean that all the chips are performing as normal, but the results aren't read by cgminer, or something similar.
This could be helpful to fix the software.

I suspected it from the begining since all my chips were quite hot to touch as if they were all hashing full speed.

Looks like it's the same as I was using.
The one for the U2 on bitmaintech site.
3.8.5 from here
https://bitmaintech.com/support.htm?pid=007201404130633105772QFG6R2u0635

Hi, anybody who can help is welcome to.
Can you explain what is this mod supposed to do in the bitmain driver?
I assume it's for the original antminer control board.

I'm still looking for people able to "hack" the driver to allow the use of an open board using CP2102 (and CP2108) as the UART/USB interface. If you are interested in one of the prototypes to work on the code, let me know, I should have 1 or 2 spare.
First proto will be 2102, "big" boards might use the 2108.
The good thing about 2108 is that it seems no actual miner is using it. Plus it is a Quad UART>USB bridge, it saves space and components on board, but I need to be sure each board will be detected and work assigned correctly.

That's what needs to be sorted out software side.
Either 8 chips are working, or cgminer only receive/reads data from 8 chips.
Bridging 2 pins on each chip can be done and should result in single chain mode, but it's too hard to do for the Lambda user.
I've done half the chips on one board so far, and I might try to finish it on this weekend, just as a proof of (don't?) work.

Ok, that makes sense, you should be able to get 15-16 "real" GH/s when undervolted.
I was unable to go over 27 GH/s pool side with non modded board.