The ORS is a base station 4G/5G MIMO 2x2. The ORS is specified to emit 0.5W in each antenna.

In TDD (Time-division duplexing), the ORS emits and receives at the same frequency at different times. In order to reduce the price and the physical dimensions of the ORS, only two antennas are necessary, each of them is used alternately in transmission and reception. A switch connects the antenna either to the reception chain or to the transmission chain.

In FDD (Frequency-division duplexing), there is no need of such switch.

As we will understand in this article, the current switch used in the ORS (TDD) can generate distortion. The ORS switch will thus be changed in the near future to ensure the absence of distortion.

Meanwhile, this does not prevent using the ORS for TDD POC or test deployments. However, ORS should either be used at lower power than its maximum TX power in order to meet ACLR requirements or it should be placed in a larger bandwidth at higher TX power in order not to create interferences on adjacent channels.

This recommendation only applies for TDD versions of ORS.

Here is the schematic of the transmission part of the first ORS version.

At the reception of the first version of ORS, we have done power measurements. 

You can see 2 devices connected to the ORS :

1.  Signal Hound SA124B: a spectrum analyser up to 12.4GHz from -151 dBm to +10 dBm (7.943.10^-16mW to 10mW). So, it has to be connected at the end of the 50 ohms line with suitable attenuator if necessary.
   The power measurements made with the SA124B hound signal presented here are the result of the real power seen by the spectrum analyser to which the value of the attenuator has been added. In fact we use the spectrum analyser in order to measure the ACLR (Adjacent Channel Leakage Ratio)

2. the Anritsu MA24104A: a power sensor to do average power measurements from 600 MHz to 4 GHz for power levels as low as 2 mW to as high as 150 W. It’s a great device to measure power easily with very low insertion loss but it can not provide ACLR measurements.

The power meter measures an average power. In the 4G configuration used here the emission time is 3.8ms over the 5ms of the TDD period. The real power is in fact about 1.3 time the power read.

So we were really disappointed by the fact that instead of 500mW we had only around 150mW (or 300mW with Anritsu) for maximum emission. Also we can see in this table that starting from gain 83, the output power is not increasing as it should and the ACLR value should not be above -45dBc.

Then we investigated by measuring the power at different points in the chain emission.

Please note that this issue obviously does not exist in FDD models of ORS which do not need a switch, as we will understand later.

As you can see on the picture we have modified the hardware of the Tx2 path to be able to measure the LNA output (Tx2 path).

On the Table 2 we check that the signal at the output TQL9093 is as it was expected with all the ACLR values under -45dBc. The power measured here are too small to be measured with the anristsu power meter.

The next step is to look at the output of the two power amplifiers. This means a measure at the combiner output.

As you can see on the picture, we modified the hardware once again to measure the power at one of the combiner output. 

Whenever ORS gain is set up to 85 the ACLR values are always below -45 dBc. The power meter measures an average power at the output of the combiner of 591mW which gives an actual power during the transmit phase of the TDD period at 591 * 1.3 = 768 mW. Losses and distortion therefore come from the switch and the band pass filter.

This table recalls the power measurements made by the power meter at the output of the ORS (table 1 Tx1 path) and at the output of the combiner (table 3 Tx2 path).

It is assumed that the 2 transmission paths Tx1 or Tx2 are substantially the same. The subtraction of these 2 powers makes it possible to estimate the losses in the filter and the switch. This is shown in the fourth column.

The band pass filter is a linear device therefore it does not generate any distortion on the signal it's only impact is its insertion loss. The datasheet of USB039B band pass filter give an insertion of 1.4dB in the band. This means that the power loss is mainly dissipated in the Renesas F2912 switch

Since we have shown that the losses are essentially dissipated in the F2912 switch we are working on the redesign of this part of the ORS.

The F2912 is not suitable for switching powers of the order of one watt although this is not clearly indicated in the datasheet. However at low level (<0dBm) this switch has a very low insertion loss -0.5dB at 2GHz and -0.6dB at 3.8GHz . Also its got a very good isolation RF1 to RF2 up to -62dB at 2Ghz and up to -50 dB at 3.5GHz. We did not find a switch on the market that matched our needs low insertion loss at 30dBm and high isolation.

The solution that we have chosen and that we have started to implement uses 2 switches. The HMC784AMS8 where the receive and transmit RF power will pass through and the F2911 where only the receive RF signal will pass through.

The HMC784AMS8 has very low insertion loss 0.4 to 0.8dB from 2GHz to 3.8GHz with an RF input power of 40 dBm. He's got a good linearity:

• supply with 5V he's got an input power for 0.1dB compression minimum of 35dBm 38dBm typical;
• supply with 8V he's got an input power for 0.1dB compression minimum of 38dBm 40dBm typical.

But from DC to 4GHz its minimum isolation is only 24dB and 28 typical. This is the reason why we have chosen to use two switches

The sketch illustrates how the 2 switches are connected together. We will have some results from this new design in the next few weeks.

Here is an example of ACLR analysis:  the frequencies between the red zone (20MHz) and yellow zone (50MHz) are the leakage due to switch distorsion.

To mitigate the current situation, we suggest that: 

1. either ORS should be used at lower power than its maximum TX power in order to meet ACLR requirements (for instance, don't use a Tx gain more than 79dB according to Table 1);
2. or ORS should be placed in a larger bandwidth at higher TX power in order not to create interferences on adjacent channels (ex. reserve 100 MHz of N78 in Germany and use 40 MHz of bandwidth at the center of the allocated band).