Unwanted Oscilation
Problem:
When powering-up v0.9.2 there is an oscillation present at 968MHz, It can be picked up at RF testing point (TP6) before the RF switch as well as at the antenna connector (look at notes of tests).
Conditions
The board is connected to a PSU 3.3V drawing ~0.290mA, PSU has current limit at 3A. The board is placed on PQ9ish dev board.
MCU has no firmware and AX5043 is unconfigured. The default state of power amplifier in ON, it turns off once AX is initialized by MCU
Effects of Problem
- The MCU STM32L476RGTx either does not start or delays to lock. Debugging shows that it hangs during internal clock configuration. Oscillation is present in either situation.
Tests
Notes:
- The board has 3 RF test probes, (disconnect the RF line)
- TP5, connected at AX5043 output
- TP3, connected at AX5043 input
- TP6, connected at PA output before the RF switch
- A general block diagram for TX line:
- AX5043-->Matching network-->TP5-->Matching network-->Power Amplifier (PA)-->Matching network-->TP6-->RF Switch-->Antenna
- PA power (On/Off) is controlled by power switch which is controlled by AX5043 (PWRAMP pin). In our tests the power switch is bypassed in order to remove a factor of oscillation in power line.
- When a probe is inserted at TP5 the oscillation dissappears and reappears once removed
- If the board is powered up having a probe at TP5, MCU boots succesfully and has control of RF and PA.
- With MCU operating, when PA is turned on either on its own or during a transmission, oscilation is still present
Test-1
TP5 was rotated in order to feed the power amplifier (PA) with signal from a signal generator. The idle current of board with the RF probe in TP5 is 3.296V @ 0.041A and without oscillation (signal generator is connected but isn't feeding any signal to the PA).
- RF-in (AX5043) is F=435.0MHz, Pin=2dBm, FM, sine, Rate: 9.6kHz, Deviation: 9.6kHz.
- RF-out in antenna is a signal, F=435MHz, channel power: 27.81dBm (integration BW: 50kHz (big...)
- RF-out 1st harmonic at F1=870MHz, channel power: 4.09dBm. DC power consumption is 3.297V @ 0.552A.
Test-2.1
TP5 was returned to its initial orientation in order to measure AX5043 output. Idle current with RF probe in TP5 is 3.298V @ 0.06A and without oscillation (the RF probe is connected to the spectrum analyzer). RF output power of AX5043 is 3dBm (programed 0x2FF), GFSK 9600, 432.5MHz, 1Hz frequency. Measured power at TP5 is ~3.8dBm (integration BW: 10kHz) and F=432.5MHz. DC power is the same with idle, PA ON (idle state) and input of PA is floating. Maybe in this measurement the AX5043 isn't programed with 0x2FF, 3dBm, because measured power is larger than AX5043 programmed output. For that reason measurements of AX5043 RF output where taken as seen in table Test-2.2.
Test-2.2
Register/Power(dBm) | Channel power (dBm) | Integration BW (kHz) | F0(MHz) |
---|---|---|---|
0x7FF/8 | 9.5 | 10 | 432.5 |
0xFFF/16 | 12 | 10 | 432.5 |
0xEFF/15 | 12 | 10 | 432.5 |
0xAFF/11 | 11 | 10 | 432.5 |
Maybe integration BW is too large for GFSK 9600. Any thoughts?
Test-3
Given the tests above, it seems that the two blocks, AX5043 and PA, work properly on their own. A thought to be wrong calculation of PA input impedance drove us to change the value of L1, the matching component at the input of PA. Change was from 3.9nH to 22nH. As a result, oscillation dissapeared and MCU booted properly.
RF-in source is AX5043, 0x2FF(programed), GFSK 9600, F=432.5MHz, 1Hz frequency.
RF-out is from antenna, channel power=27.8 - 28dBm, Integration BW=11kHz (span=150kHz), F=432.500750MHz.
PA power detection pin
And ripple of the same signal
Note: Once oscillation disappeared, we addd again the power switch, as noted in test-notes.
Test-4
Now, using the same configuration of Test-3 (look also in the note-Test-3) but with a +5V PA power supply: RF-out:
- F0=432.500500MHz, channel power=31.7-32dBm, integration BW=11kHz
- F1=865.000500MHz, channel power=6.36dBm, integration BW=20.9kHz
Test-5.1
3.3V board but with L1 = 15nH.
Test-5.2
Look at issue #26 (closed)
Test-5.3
3.3V board. L1=3.6nH. Power (dBm) isn't correct because the measurement is at the antenna and RF switch does not pass the RF signal (isolate). Same oscillation with L1=3n9 but in different frequency.
Test-5.4
3.3V board. L1=1.2nH. Power output is almost the same as with with 5.6nH, (test-5.2).
Spurious signals appear after some (n=10) loops of the program when the transmitter signal is GFSK 9600, with 1Hz frequency or after some time of CW (continuous) transmitting signal. Also in this test and test-5.2 the clocks of MCU delay to lock (when the MCU is in debugging mode).
@drid ,@acinonyx , @papamat , @pierros , @surligas take a look.