A text teaches you how to solve the problem of transmitting LOL (emission local oscillator leakage)

"Advantage" always stands with "challenge"

Even if it is called

"Black Magic in the Next Generation SDR Transceiver",

“Zero IF” is now facing a challenge that needs to be overcome –

Launch local oscillator leakage, referred to as “launch LOL”.

Uncorrected transmit LOLs can generate unwanted emissions within the required transmit range, posing a potential risk of violating system specifications. This article discusses the issue of launching LOL and introduces techniques that can be implemented in ADI's RadioVerseTM transceiver family to eliminate this problem. If you can reduce the emission LOL to a low enough level that it no longer causes system or performance problems, maybe people don't have to worry about LOL issues!

The RF mixer has two input ports and one output port, as shown in Figure 1. The ideal mixer will produce an output that is the product of the two inputs. In terms of frequency, the frequency of this output should be FIN + FLO and FIN – FLO, with no other items. If either input is not in the drive state, there will be no output.

Figure 1. Ideal mixer

In Figure 1, FIN is set to FBB with a baseband frequency of 1 MHz and FLO is set to a FLO with a local oscillator frequency of 500 MHz. If it is an ideal mixer, it will produce an output with two tones at 499 MHz and 501 MHz.

However, as shown in Figure 2, at FBB and FLO, the real mixer will also generate some energy. The energy produced at the FBB is negligible because it is far from the desired output and will be filtered out by the RF components after the mixer output. Regardless of the energy produced at the FBB, the energy produced by the FLO can be a problem. It is very close to or within the desired output signal and is difficult or impossible to remove by filtering because the filtering also filters out the desired signal.

Figure 2. Real mixer

This useless energy produced by the LO in the font of the small one or two is called LOL. The local oscillator (LO) that drives the mixer has leaked to the output port of the mixer. There are other ways for the LO to leak to the system output, such as through a power supply or across the silicon itself. Regardless of how the LO leaks, its leakage can be called LOL.

In a real-signal IF architecture that only transmits one sideband, the LOL problem can be solved by RF filtering. In contrast, in a zero-IF architecture that emits two sidebands, the LOL is in the middle of the desired output and creates a more difficult challenge (see Figure 3).

Figure 3. Unwanted energy produced in FLO (shown in red). This unwanted energy generated under FLO is called LOL.

Traditional filtering is no longer an option, as any filtering that removes LOL will also remove some of the desired transmitted signals. Therefore, other techniques must be used to eliminate LOL. Otherwise, it may eventually become unwanted emissions throughout the required range of emissions.

LOL cancellation can be achieved by generating signals of equal amplitude but opposite phase to LOL, thus canceling it out, as shown in Figure 4. Assuming we know the exact magnitude and phase of the LOL, a dc offset can be applied to the transmitter input to generate a cancellation signal.

Figure 4. LO leakage and cancellation signals

The complex mixer architecture is suitable for generating cancellation signals. Since there are quadrature signals of the LO frequency in the mixer (they are the key to how the complex mixer works), it is allowed to generate LO frequency signals of any phase and amplitude.

The quadrature signals used to drive the complex mixers can be described as Sin(LO) and Cos(LO)—these are orthogonal signals at the LO frequency that can drive two mixers. In order to generate a cancellation signal, these orthogonal signals are added with different weights. In mathematics, we can produce an output, I × Sin(LO) + Q × Cos(LO). Using different signed values ​​instead of I and Q, the resulting sum will be the LO frequency signal and can have any desired amplitude and phase. An example is shown in Figure 5.

Figure 5. Example of any phase and any amplitude cancellation signal generated

The desired transmit signal will need to be applied to the input of the transmitter. After applying a DC bias to the transmitted data, the output of the mixer will contain the desired transmit signal and the desired LOL cancellation signal. The specially generated cancellation signal will be cancelled in combination with the useless LOL, leaving only the desired transmitted signal.

As shown in Figure 6, the observation receiver is used to observe the emission LOL. In this example, the observation receiver uses the same LO as the transmitter, so any emission energy at the LO frequency will appear as DC at the output of the observation receiver.

Figure 6. Basic concepts of observing and correcting TxLO leaks

The method shown in Figure 6 has its inherent drawback: using the same LO for transmission and observation, the transmit LOL will appear as DC at the output of the observation receiver. Since the components in the circuit do not match, the observation receiver itself will have a certain amount of DC, so the total DC output of the observation receiver will be the emission LOL present in the transmit link and the native dc offset of the observed link. There are ways to overcome this problem, but a better approach is to use different LO frequencies for observations to separate the native DCs in the observation path from the emitted LOL observations. This situation is shown in Figure 7 below.

Figure 7. Using different LO emissions and observations

Since the frequency is different from that of the transmitting LO, the energy of the transmitting LO frequency does not appear in direct current in the observation receiver. Instead, it will appear as a baseband tone with a frequency equal to the difference between the transmitted LO and the observed LO. The native DC in the observation path will still appear as DC, so the observed DC is completely separated from the transmitted LOL measurement.

For simplicity, Figure 8 illustrates this concept using a single mixer architecture. In this example, the transmitter's input is zero, so its only output is the transmit LOL. The frequency shift is done after observing the receiver, moving the energy observed by the transmitting LOL to DC.

Figure 8. Separating the receiver DC from the Tx LOL

The output of the observation receiver is divided by the transfer function from the transmit input to the output of the observed receiver, and the resulting result is compared to the expected emission to find the desired correction value. The transfer function involved is shown in Figure 9.

Figure 9. Transfer function from emitter input to observing receiver output

The transfer function from the transmitter baseband input to the observation receiver baseband output consists of amplitude scaling and phase rotation. This is explained in more detail below.

Figure 10 shows that if there is gain or attenuation in the loopback path from the transmit output to the observing receiver input, or if the gain of the transmitter circuit is different from the gain of the observed receiver circuit, the amplitude of the transmitted signal reported by the observing receiver may not represent The actual amplitude of the transmitted signal.

Figure 10. Amplitude scaling due to loopback attenuation

Now look at the phase rotation. It is important to realize that the signal will not be transmitted instantaneously from point A to point B. For example, the signal passes through copper at about half the speed of light, which means that the 3 GHz signal transmitted along the copper strip has a wavelength of about 5 cm. This means that if multiple oscilloscope probes spaced a few centimeters are used to detect the copper strip, the oscilloscope will display multiple signals that are out of phase with each other. This principle is illustrated in Figure 11, which shows three oscilloscope probes separated by copper strips. The frequency seen at each point is 3 GHz, but there is a phase difference between the three signals.

Figure 11. Distance versus phase, 5 cm trace, 3 GHz signal, and probe points at 0 cm, 2 cm, and 4 cm

It is important to note that moving a single oscilloscope probe along a copper strip will not show this effect because the oscilloscope will always trigger at 0° phase. The relationship between distance and phase can only be observed using multiple probes.

Just as there is a phase change along the copper strip, a phase change occurs from the transmitter input to the observation receiver output, as shown in Figure 12. The LOL correction algorithm must know how many phase rotations have occurred in order to calculate the correct correction value.

Figure 12. Phase rotation caused by physical distance in the loopback path

Determining the output from the transmit input to the observation receiver

Transfer Function

The transfer function shown in Figure 13 is obtained by applying a transmitter input signal and comparing it to the output of the observation receiver. But some points need to be kept in mind. If a static (dc) signal is applied to the transmitter input, it will produce an output that emits the LO frequency and the transmit LOL will be combined with it. This will prevent the transfer function from getting correct. It should also be noted that the transmit output can be connected to the antenna, so intentionally applying a signal to the transmitter input may not be allowed.

Figure 13. Determining the transfer function from the emitter input to the observation receiver output

To address these challenges, the ADI transceiver uses an algorithm that applies a low-level dc offset to the transmitted signal. The offset level is adjusted periodically and the output of the observation receiver shows these disturbances. Then, the algorithm analyzes the difference between the observed value increment and the input value, as shown in Table 1. In this example, no user signal is transmitted, but the method still applies when the user signal is present.

Table 1. Comparison of observation increments and input value increments

The transfer function is obtained by performing the subtraction of the two cases and eliminating the constant emission LOL from the equation. It can be expanded to more than two situations, and many independent results can be averaged to improve accuracy.

The LOL correction algorithm will be able to learn the transfer function from the transmit input to the observed receiver output. The output of the observation receiver is then divided by the transfer function to derive the input to the transmitter. The transmit LOL is determined by comparing the expected DC level of the emission with the observed DC level of the emission. Finally, the algorithm will calculate the correction values ​​necessary to eliminate the transmitted LOL and apply it as a DC offset to the desired transmit data.