Question: In the attached plot (labeled Figure 10) from the ADL5375 IQ Modulator Datasheet, LO Leakage degrades significantly vs. temperature at higher frequencies after compensation at 25 degC. What is going on here?
Answer: It is important to understand that this plot shows the change in LO Leakage over temperature after compensating the LO Leakage at 25 degC. So as you can see from the plot, at all frequencies, the LO Leakage at 25 degC looks good. This is simply because the LO Leakage is being compensated at each of those frequencies (i.e. each frequency point has different calibration coefficients). What happens next is that we change the temperature while holding the 25 degC calibration coefficients constant (the calibration “coefficients” are simply offset levels on the I and Q inputs).
So why does the post-calibration LO Leakage degrade over temperature so much above 2.5 GHz?
To understand this we need to understand the LO Leakage mechanism a bit better.
LO Leakage is generally assumed to be caused by internal offset voltages on the I and Q inputs. This is why we apply compensating offset voltages to these inputs to reduce the LO Leakage. However, there is also an LO Leakage path between the LO input port to the RF Output (see attached image). So the net LO Leakage at the output is the vector sum of the signals leaking through these multiple paths. The LO Leakage path from the LO Input to the RF Output has some important characteristics.
1. As the frequency changes, the amount of leakage through this path will change with the leakage level tending to increase with increasing frequency.
2. As the temperature changes, the amount of leakage through this path will change.
The attached plot (labeled Figure 9), shows how uncompensated LO Leakage changes with frequency and temperature. The increase in LO Leakage with frequency is consistent with the above explanation.
If we compensate LO Leakage at, say, 1 GHz, we are compensating for LO Leakage that is mostly caused by I and Q offsets with LO-to-RF Leakage playing a less significant role. When we cycle over temperature, the 25 degC compensation tends to hold up well and the LO Leakage at +85 degC and -40 degC does not change by much (see plot labelled Figure 10).
At 4 GHz, things look different (referring again to the plot labelled Figure 10) . At this frequency, the initial (un-compensated) LO Leakage is dominated by LO-to-RF leakage and I/Q offset based LO Leakage is less significant. Even though most of the LO Leakage is not coming from the I and Q input offsets, we can still use these inputs to compensate the overall LO Leakage. So regardless of where the original LO Leakage comes from, we can always compensate it by applying compensating offset voltages to the I and Q inputs.
However, when we change temperature (or frequency) without re-compensating, the leakage of the LO-to-RF path changes and this causes the overall LO Leakage to change (usually it degrades). So the benefit of compensating at 25 degC quickly disappears, particularly at the highest frequencies where LO-to-RF leakage dominates.
You can read more about the topic of IQ Modulator Error Correction in AN-1039, Correcting Imperfections in IQ Modulators to Improve RF Signal Fidelity