O
Member
【Editorial Discussion】
A transmitter or front-end that is clean on one carrier can show a new spectral product once a second carrier is enabled for carrier aggregation. The first instinct is often to blame the PA. The product may instead be a frequency-plan landing, a filter that was never asked to reject that offset, or a path other than the PA. Before you retune a matching network, it is worth asking which of those actually appeared with the second carrier.
1) Write the two carrier frequencies, the bandwidths, and where the unexpected product sits. A third-order term at 2f1−f2 or 2f2−f1 is a different story from a harmonic, a mixer or LO product, or a PIM-like term in a passive joint. If the product frequency was predictable from the frequency plan, the plan itself belongs in the discussion, not only the hardware.
2) Check whether the product comes from the PA, from a shared driver, or from a passive junction after the PA. Turning one carrier down, or moving one carrier inside the same band allocation, will not affect all of those the same way. A product that tracks the two carrier levels as a classical IMD, and a product that is almost insensitive to drive but sensitive to a connector or a filter joint, should not be filed under the same heading.
3) The filter and duplex path may have been specified against a single-carrier mask. A CA product can land where the filter rejection was never required, or where two carriers share a path that was only tested one at a time. If the product is visible at a port before the filter and gone after it, or the other way around, write that down before changing the PA.
4) Keep the same total power, the same allocation, and the same thermal condition when you compare “one carrier” and “two carriers.” A two-carrier test at a higher total power, or after a different warm-up, is not a fair CA comparison. Record whether DPD or an AGC loop was frozen or adapting; a loop that was trained on one carrier is a different system once the second carrier is on.
Open question: the last time a new product appeared only after CA was enabled, did you eventually attribute it to the PA (or a shared active stage), to a filter or passive joint, or to the frequency plan? What experiment — a level sweep, a port isolation check, or a carrier-frequency move — actually settled it?
A transmitter or front-end that is clean on one carrier can show a new spectral product once a second carrier is enabled for carrier aggregation. The first instinct is often to blame the PA. The product may instead be a frequency-plan landing, a filter that was never asked to reject that offset, or a path other than the PA. Before you retune a matching network, it is worth asking which of those actually appeared with the second carrier.
1) Write the two carrier frequencies, the bandwidths, and where the unexpected product sits. A third-order term at 2f1−f2 or 2f2−f1 is a different story from a harmonic, a mixer or LO product, or a PIM-like term in a passive joint. If the product frequency was predictable from the frequency plan, the plan itself belongs in the discussion, not only the hardware.
2) Check whether the product comes from the PA, from a shared driver, or from a passive junction after the PA. Turning one carrier down, or moving one carrier inside the same band allocation, will not affect all of those the same way. A product that tracks the two carrier levels as a classical IMD, and a product that is almost insensitive to drive but sensitive to a connector or a filter joint, should not be filed under the same heading.
3) The filter and duplex path may have been specified against a single-carrier mask. A CA product can land where the filter rejection was never required, or where two carriers share a path that was only tested one at a time. If the product is visible at a port before the filter and gone after it, or the other way around, write that down before changing the PA.
4) Keep the same total power, the same allocation, and the same thermal condition when you compare “one carrier” and “two carriers.” A two-carrier test at a higher total power, or after a different warm-up, is not a fair CA comparison. Record whether DPD or an AGC loop was frozen or adapting; a loop that was trained on one carrier is a different system once the second carrier is on.
Open question: the last time a new product appeared only after CA was enabled, did you eventually attribute it to the PA (or a shared active stage), to a filter or passive joint, or to the frequency plan? What experiment — a level sweep, a port isolation check, or a carrier-frequency move — actually settled it?
