[Gta04-owner] GPS fix?

Dr. H. Nikolaus Schaller hns at goldelico.com
Tue Nov 22 09:07:16 CET 2011


Hi,

Am 21.11.2011 um 12:53 schrieb Boudewijn:

> On Monday 21 November 2011 11:31:11 Dr. H. Nikolaus Schaller wrote:
> > Hi Neil, Boudewijn, et al.,
> > 
> > over the weekend, we have done some more experiments
> > and now it is clear where the bug is sitting. It is indeed an IC
> > with 6 feet (like bugs) IC: the antenna switch (U1201).
> So, one lesson for GTA04: use IC's with at least 8 feet (like spiders) to catch more bugs. 10-feet (err.. ten-feeted) IC's might help it make it crawl back to land when dropped in shallow waters.
> 
> > Although we have copied the circuit 1:1 from the Freerunner
> > we did make one mistake. We have swapped the two antenna
> > inputs. But only for the antenna switch chip and not the control
> > signal which should have been inverted.
> How did you find this out? Is that by closely studying the board design, or by measurements on an actual board? I found U1202 on page 84 of the system manual (v3.1, 11 Oct)

By measurements. Measuring the voltages on all resistors and
components of the switch circuit. And soldering a control
wire to the CONT input of the antenna switch chip.

It turned out that senstitivity goes heavily up, if VCONT is grounded.
This is contrary to the logic of the circuit and GPIO144.

Then, a simple comparison of the schematics did reveal the issue.

The schematics of the GTA02 is here:

http://downloads.openmoko.org/developer/schematics/GTA02/Schematics_Freerunner-GTA02_A5-A7cumulative_public_RC0.pdf

and GPS is on sheet 10.
 
> 
> > So, if the current sensor U1202 detects an external antenna,
> > this tries to switch off the power for the internal antenna (T1201)
> > and moves the RF switch - to "internal".
> > 
> > And if there is no external antenna, it moves the switch to the
> > external input.
> The switch would be U1201, wouldn't it?
> 

Yes.

> > Additionaly, the Freerunner circuit design is quite critical since
> > it does operate the antenna switch in some linear mode.
> > We have never seen a clear control signal that is within
> > the specification - so I wonder why it ever worked in the
> > Freerunner...
> Haha, that startled my daughter when I laughed out loud ;-)
> 
> > We could also not see that it switches off the power of the
> > internal antenna completely. Therefore, the receiver always
> > sees a mix of internal and external antenna signals.
> > 
> > This was proven by temporarily connecting and removing
> > the internal antenna, while monitoring the satellites (GPGSV
> > record).
> > 
> > Adding the internal antenna reduces the SNR by 10dB.
> > Since the maximum SNR I have seen is 35dB these
> > 10dB reducting is enough to make all satellites invisible.
> > Even with external antenna.
> > 
> > Therefore, the only mode where you currently can get a
> > GPS fix with the GTA04A3 boards is after removing the
> > internal antenna and using an external antenna outdoors.
> > Indoors, signals are too weak.
> > 
> > Now the other experiment was to solder some wires to
> > manually control the RF antenna switch.
> > 
> > When forcing the RF-switch to "external", it is really like
> > removing "ear-protectors" and the SNR raises by 15 dB:
> My internals and externals are getting switched after reading the above... Correct me if I'm wrong: you forced the RF switch (U1201) to external, which actually is connected to the internal antenna, by making it belief there was an external antenna connected, so that the actual power was set on the internal antenna? 

Mostly. I forced the switch to Input "B" (Pin3) although the current detector U1202
sees the current drawn by the external antenna and tries to set the control
to a high volgate, which would mean "connect input A". So this was an override
of the switch control. And this improves the reception dramatically.

> 
> > red color: no signal
> > brown: weak signal
> > green: sufficient signal
> > light green: very strong
> > 
> > Now as we have identified the reason, what is a cure?
> > 
> > At the moment we do not know yet, but are looking into some
> > options. One is to really cross over the input signals by two tiny
> > wires. We have to discuss that with the SMD and RF specialists.
> Does it "only" involve soldering two tiny wires, or also disconnecting the original (incorrect) paths? About what size is the switch on the board?

A first estimate means that the traces must be cut. Or the chip must be
removed, two pins must be bent upwards, and the chip resoldered.

Then a 2mm and a 5mm long tiny wire must be soldered. IMHO quite
difficile if you don't have a stereo microscope.

> 
> > Maybe of interest for Open Hardware development lab stories:
> > 
> > Why didn't we notice earlier? The GTA04A2 board didn't
> > work well enough to test any GPS. And the first GTA04A3 boards
> > arrived in July. Back then, we did some tests with an external
> > antenna and got a location fix. So we thought it works and missed
> > the detailled analysis of the SNR values etc. And we had to solve
> > other even more important issues first (you may find a collection
> > in the list archives).
> > 
> > Conclusions:
> > 1. testing can only prove the presence of bugs, not their absence.
> > 2. there is always one more bug.
> 
> Thanks for the troubleshoot, testing and explanation!

That is what makes us different...

Nikolaus

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20111122/58a24e90/attachment-0001.html>


More information about the Gta04-owner mailing list