GTA04A4 starts shipping tomorrow

Dr. H. Nikolaus Schaller hns at goldelico.com
Mon Jan 30 21:13:04 CET 2012


Dear all,
you may have wondered a lot why it was so quiet about the GTA04A4
boards in the last weeks. And nobody did report about unboxing
experiences, because nobody did receive a new GTA04.

The reasons where manifold. It started with soldering issues where they
had to make a new stencil. This improved soldering quality a little but it
was still not perfect. So we got much less good boards out of the soldering
machine than planned and chips had to be replaced...

Then, we did test the boards completely. It turned out that they still had
some minor issues. The most maddening bug was that there was an error
in the GND flooding of the backup battery. Therefore it was shorted on the
PCB and could not work.

After fixing this and testing more, we found out that the new circuit that
is intended to allow the CPU to control the power of the UMTS modem
through a GPIO did not work as intended. Rather than shutting completely
down it did put the module into reset state. This would have been ok, if
reset would be the same as power down. But the module draws 16 mA
in this state. Even more than a not powered down modem in the GTA04A3
boards which draws 3-10 mA and can wake up the CPU...

So we finally found a way to solder a NPN transistor with built-in bias
resistor to the position of some test pads and other resistors to make
it work. This made us again bring the boards to the factory and
have them do some hand-rework. And wait some days...

Then, last week, we got the first boards that did pass all automatic
tests and could power down completely (to some 300 uA for RTC).

Happily we did install them into the first GTA02 cases to build the
"OpenPhoenux" (GTA04 Complete) units. They worked fine, until we
wanted to operate the GPS. It did not work well with internal antenna,
but really excellently with an external antenna (up to 11 of 12 satellites
used for navigation).

We found out that the U.FL antenna socket was mounted in the
wrong orientation (180 degrees rotated). This shorts the power
supply of the internal Antenna (which has an amplifier built in).

So, boards did go to rework again. Now, GPS with internal antenna
was a little better, but still very weak.

After 3 days of intensive research (incl. this weekend) we did find the root
cause: it is the same as it was for the GTA02A5 board [1]. The SD card
clock. It is 25 MHz and 25 * 63 = 1575 MHz, i.e. the GPS frequency.

This means that the GTA04 creates some noise which is neglectable
in more than 5cm distance with an external antenna, but not for the
active GPS antenna built into the case. And some boards and
antennas are a little better and others are a little worse so that it
results in a basically working, but very unreilable receiver.

The main problem was to find a measurement setup that gave
reproducible results so that we could not only compare devices, swap
antennas and try different shielding scenarios, but also check the
effectiveness of measures against the problem.

Like with the GTA02 there is a hardware and a software fix. The SW fix
was to re-enable a feature in the 2.6.32 kernel that we had disabled due
to problems with the SDIO driver. It appears that it is not at all necessary
to change anything in the 3.2 kernel. So please forget immediately that
there is a SW fix :) Because it fixes our unawareness and not a problem...

The real hardware fix is adding a small capacitor like for the GTA02A5.

You may wonder why we did not find it earlier. One aspect is testing time.
The more testing you do the more issues you find. But we simply did
not have enough GTA04A3 boards to do these tests...

This time it was better and we could set up, test and compare e.g.
5 GTA04A4 boards in parallel. And swap antennas between units etc.
This simply was not possible before.

So after having applied this last hand-rework, we have the first boards
and devices that are in our opinion good enough to really be shipped
and make you happy! More boards will become available next week.

And we have some boards to show during FOSDEM next weekend.
So please come and have a look!

Finally, it is still time to subscribe to the Group Tour (which will get
another board redesign where all these things are done right from
the beginning - but otherwise they are equivalent).

Nikolaus

[1]: http://wiki.openmoko.org/wiki/GPS_Problems




More information about the community mailing list