XIEGU G90 - ATU problems

Some considerations about Xiegu G90 internal ATU tuning problems.

1. The ATU is a "relay" based.

2. The ATU is driven by the Main Unit MCU STM32F429.

3. The mismatch (SWR ratio) is measured with a Bruener phase compensated bridge, like the one in this picture (too lazy to draw it myself): 

Here is a picture of the SWR bridge from the Xiegu G90:

The outputs of the bridge are fed into a dual OpAmp then sent to the MCU ADC where the values are used to calculate the SWR that is used to find the match solution and it is also sent via UART to the Front Panel MCU to be displayed on the LCD.

4. In some FW versions the ATU seems to do a better job in matching various non-resonant antennas. FW revisions that use high RF power gave more accurate matching solutions than those that uses low RF power.

5. How the tuning sequence is done (in a simplified and presumable description): 
    Frequency is stored in EEPROM;
    -LC relay configuration is stored in EEPROM;
    -SAMPLE(Direct and Reflected are sampled);
    -oldSWR is calculated; oldSWR is stored in EEPROM;
    -change is applied to LC relay configuration;
    -SAMPLE;
    -newSWR is calculated;newSWR is stored;
    -newSWR is compared with oldSWR; decision is made upon this (continue or stop finding a match solution).

In high quality "tuners", the RF is applied only during the SAMPLE and SAMPLE is made when the relays are settled, in a stable state.

In the G90 the RF is present always during the TUNE process.

Depending on the voltages applied function of time, the relays can be in 3 states: ON. OFF and UNSETTLE, the last being an uncertain position between the first two ones and this parameter is defined as “operating time” which, depending on manufacturer can or cannot include the bouncing time.The bouncing time is a time interval in which the contacts are closing and opening erratically for mechanical reasons.

If the SAMPLE is done during the bouncing or inside the operating time interval, the result can be massive altered. (bouncing during the RF presence can make the LC’s to “ring” on various harmonic frequencies also).

This is particulary valid on those relay ATU’s that are working very fast and the ATU in the Xiegu G90 IS A VERY FAST one!

Because all of this, a wrong LC configuration can be selected for that particular frequency thus allowing the radio to transmitt full power on a mismatched antenna system.

This can be a source for errors in ATU sequence and those can be identified very quick by activating the tune sequence several times on the same frequency and the same antenna system. 

Beware! The short press will recall a matching solution already stored in memory for that frequency therefore a long press is desirable, in order to erase that memory and start the matching process from the scrap. (Thanks Dave).

If the results are different, most probable is due to settle time. This time interval is in the 5-10msec range.

The problem can be mitigated by various means, all in the FW code; the most efficient one is to switch on RF after a 10msec timeguard from the LC solution is applied to the relays.

6. Errors in SWR value can came from the intrinsec construction of the SWR bridge. The major source of error are the diodes in the bridge. Injecting low RF (around 1W) into a Bruener bridge based on Si diodes (even Schottky) can be into the non-linear range of the transfer function which also is altered by variation in the ambient temperature. (https://www.analog.com/en/technical-articles/integrated-diode-based-rf-detectors.html#)

Another factor that alter the SWR ratio could be the loses on the Reflected path. On loosy antenna systems, the Reflected power is attenuated by some factors like the lenght in the coaxial cable. This leads to a lower voltage in the non-linear transfer function of the diode therefore reducing more and more the reflected detected voltage and finally affecting the SWR ratio to a point that the MCU "feel" a proper match is achieved even if a non-match solution is set in the LC relays.

Therefore, the SWR calculated value could reflect or not the true SWR. The result is a wrong tuning solution.

This problem can be mitigated by applying one of these solutions:

a.     Replacing the Direct and Reflected detection diodes with lower threshold ones. The actual diodes used in the Xiegu G90 has to be measured to find out if they are normal Si or Schottky.

b.     Biasing the actual diodes to the linear transfer range.

c.     Replacing the diodes with logarithmic amplifier like AD8307

Actually, the last method can be the most successfull one it is also the most complicated and require major modifications that might be not reproductible by most hams so the first and the second one could be further explored.