Solution
The Differential Error Vector Magnitude (DEVM) measurement in RFmx Bluetooth was restructured to align with the specification definitions, addressing several bugs in the Test Toolkit for Bluetooth code. RFmx uses an acquisition rate of 8 MS/s, whereas the toolkit uses 4 MS/s. Certain signal processing blocks were removed in RFmx to better adhere to the Bluetooth specification.
The Signal Conditioning Option (SCO) estimation and correction are more stable in RFmx Bluetooth compared to Test Toolkit for Bluetooth. RFmx Bluetooth also uses a fractional delay correction filter to correct sampling phase error and timing offset as defined in the standard. The RFmx Bluetooth TxPower code had minimal changes compared to the Test Toolkit for Bluetooth. The Test Toolkit for Bluetooth did not consider the SYNC and Guard portions of EDR packets, whereas RFmx Bluetooth does, as defined in the standard.
An issue with incorrect sampling duration in VBW filtering was fixed in RFmx Bluetooth. Averaging is handled correctly within packets first and then across packets in RFmx, unlike in the Test Toolkit for Bluetooth. In summary, the RFmx DEVM shows better numbers compared to the Test Toolkit for Bluetooth, while the difference in TxPower is minimal with the number of averages set to 1.