## Tuesday, 9 May 2017

### The TT7F v1.5 Revision

My plan is to launch several floaters over the course of a few weeks so I don't have to repeatedly rent a new H cylinder and instead do it with one. For that I needed to put together a few more trackers. Since ATSAM3S4 was pin compatible with ATSAM3S8 I wanted to take advantage of the lower prices on Aliexpress and ordered four chips from a Chinese seller. However, only one worked while the other three exhibited peculiar behaviour and were never detected by the USB host. In short, during investigation of the problem I order a set of new TT7F boards from OSHpark. Since I had to adjust the design to satisfy OSHpark's stricter design rules, I used that opportunity to fix a couple of mistakes and add a few practical bits to the design.
The new schematic and board files can be found following these links or on github.

Changes:
• Narrower MT9D111 footprint.
• Re-wired traces inside MT9D111 connector.
• Covered small vias.
• Matching re-done for Si4463.
• Larger pads for SOLAR+, BATTERY+ and GND.
• Separated LTC3105 output and TPS63031 input - either use a jumper or bridge the pads.
• Added three large vias at the top for easier suspension.
• Removed VBUS powered info line (PA1) from the USB interface.
• Added TIMEPULSE connection from Ublox to ATSAM3S.
• TPS63031 equipped with a 3.3uH inductor (originally 2.2uH).
I ordered three boards with the 2oz 0.8mm thickness service for $10.60 and received them in 19 days. Above are the assembled trackers. I added the SMA connector only temporarily for easier testing. The layout also allows connecting/disconnecting power to the tracker with a jumper. The backside provides larger pads for a battery/supercap and solar panels. I placed a ground pad in the upper left corner for a dipole's ground element as well. Forgotten in the original design was something to tie a string to to hang the tracker easily. That is taken care of with three 1mm vias at the top now. Since the Ublox GPS module is one of the priciest parts on the tracker I looked for an alternative. These three Biwin GM10 modules from Aliexpress cost$20.96, almost half the price of a genuine Ublox MAX-M8Q. They should be built on the UBX-G7020-KT chipset and are pin compatible to the MAX-7/8 modules. So far I equipped only one of the trackers with this one to verify its functionality. Based purely on to which backup command the chipset responded I'd say it ran an older firmware, MAX-7 perhaps. It's consumption seemed higher than in genuine Ublox. Acquiring fix took longer with less satellites as well, but eventually it began outputting accurate positional data even from inside a wooden house.
I wanted to learn something more about the possible matching setups for Si4060/Si4463 transmitters so I decided to utilize the AD8318 and measure the output power of the individual trackers. I did two kinds of measurement. In the first case I had the tracker increasing in steps the value of the PA_PWR_LVL register at 144MHz and than at 434MHz transmitting carrier wave. In the second case I had the tracker transmitting CW at a fixed power setting (0x23) and stepping through its whole frequency range (2MHz steps). The AD8318 was provided with a stable 5.05V supply and sampled with an Arduino NANO. A 30dB labelled attenuator was placed between the tracker and the power meter. Comparison of the data with and without the attenuator suggested rather 33dB attenuation.
Later I accompanied the power data with current consumption measurements. I used Arduino DUE's 12-bit ADC to sample the voltage drop across a 1Ω shunt resistor. The tracker ran the same PA_PWR_LVL increasing code at 144MHz and 434MHz with the GPS module in backup mode and ATSAM3S running of the 12/16MHz XTAL as before. The consumption data in the following charts is the measured value minus the tracker's steady state consumption with the transmitter turned off so it should correspond to Si4060/Si4463's (plus the TCXO) consumption only.
The transmitter's output was terminated with a 50ohm termination. The LiPo battery providing power to the tracker was at 4.05V.
The first matching/filtering network I tested was based on (with minor changes) the 10dBm CLE layout for Si4060 from application note AN627. Implemented still on the old TT7F board with Si4060.
33dB were added to all of the power data points to compensate for the attenuator except for the 900MHz band which was measured directly. Both measurements suggest potentially higher output power at 144MHz.than at 434MHz but at twice the current. The 434MHz consumption seems to correspond to the datasheet stated 18mA at 10dBm.
The second network derived from the 20dBm CLE match for 434MHz from application note AN648 (modified to components I had at hand) was used on three different trackers all equipped with Si4463, one on the old TT7F board and two on the new v1.5 boards.
First the old board implementation again suggests much higher output at 144MHz, but this time at comparable current consumption. The 434MHz output doesn't even reach 10dBm.
For comparison these charts show the same matching network with the new layout on the v1.5 board. The 434MHz output seemed to be boosted as well as the consumption. The datasheet states 20dBm output at ~80mA.
This is data for the second v1.5 board. It seems to be more or less consistent with the data from the other board. The similarity in the measured output in the last two cases (new boards) in contrast to the first case (old board) suggests the different physical layout of the matching network plays significant role in the resulting output power and consumption. In all three boards it seems the filter cuts the output too early for the 434MHz output. None of them came close to 20dBm.
The last network I tested was 20dBm SQW match for 169MHz as in application note AN648 (with minor changes). The datasheet states 20dBm output at 70mA.
At maximum the 144MHz output reaches 16-17dBm with current consumption up to 95mA. The 434MHz output is then attenuated to -21dBm with the current consumption skyrocketing.

To conclude, in the original board I tried to follow the recommended layout as shown in application note AN629. On the other hand in the v1.5 board I focused more on fitting it all inside limited space. Looking at the data above I assume that proper layout plays a crucial role in achieving datasheet stated performance. Some of the deviations from the ideal could also be attributed to not following precisely the recommended component values.
Final look at the five trackers tested and currently being prepared for, hopefully, long-term operation. Four of the trackers are equipped with ATSAM3S8 and one with ATSAM3S4. One is testing the Biwin GM10 GPS module. Four utilize Si4463 and one Si4060. Four could potentially transmit SSDV. More on further preparations to come.