[2CC][20M+20M] CFR correction pulse to reduce EVM/ACLR

It looks like incorrect IQ rate is being used at the input of the CFR block. For eg, for UC50 NLS, the IQ rate to be used is 122.88MSPS and for UC51 NLS, the IQ rate should be 245.76MSPS.

Also, you need to give exact offset for the carrier from the LO frequency. For the second carrier, you need to give 10.1MHz as the offset.

For IQ rate issue, our chip vendor use 51C_NLS but with sampling rate=122.88Msps and lane rate of 8.11Gbps.

For 1CC case, our DUT use JESD8Gx2 and can get good EVM/ACLR as Case C) below.

Since chip vendor claim that 2CC case need to use JESD8Gx4 as the previous Case A) and Case B), I also try JESD8Gx4 with 1CC case as Case D), it seems that

with JESD8Gx4 speed, CFR BW effect has been doubled. That's why I'm asking if the CFR effects about BW is related to JESD speed.

Back to CFR generation question, should I still keep fsHz = 122.88e6 to generate pulse?
What's the correct way to match JESD8Gx4 with sampling rate=122.88Msps waveform for 2CC case?


Case C)
JESD8Gx2, 1CC case, use CFR NR40MHZ30KHZ

EVM: 1.53%, ACLR: < -49% 

Case D)
JESD8Gx4, 1CC case, use CFR NR40MHZ30KHZ

EVM: 1.36%, ACLR: -41%

CFR Input rate is not dependent on no of lanes or JESD rate.

For a give profile, look at the Tx settings in the GUI and provide that as the CFR input rate. CFR block is before DPD block.

I use sampling rate=245.76Msps and exact offset for the carrier from the LO frequency to generate CFR pulse.
The CFR effect on bandwidth is correct, however, EVM is not reduced.

CFR:
bpid = {'NR20MHZ30KHZ','NR20MHZ30KHZ'};
fsHz = 245.76e6; % CFR rate in DFE
osr = 4; % Oversampling rate
freq = [-10.02e6, 9.99e6]; % Frequency offset for CCs

Waveform:
CC1(3615.00 MHz, CBW: 20MHz) + CC2(3635.01 MHz, CBW: 20 BW),
One ADI LO (3625.02 MHz)

However, EVM is still 4.95%, 4.94%, here is the updated test results:
 

What parameters could I check further?
Is osr = 4 OK for 2CC case?
(DUT use osr=4 for 1CC and performance is good)

Here is another 2CC (20M+20M) test, basically,

2CC could get EVM 1.73%/1.72%, ACLR L1:-47.99dB, U1:-48.20 from SA as below screenshot



By using adi_adrv9025_CfrStatusGet, almost peak waveforms are detected in the 1st CFR engine, no waveform is hardclipped, as below


However,
the dpdDirectEvm value, got by using API adi_adrv9025_DpdStatusGet to get dpdStatistics->dpdDirectEvm value,
fluctuate between 1.76 and 3.95.

Is this fluctuation dpdDirectEvm expected for 2CC waveform case? What can I check this issue further?

Here is my 2CC (20M+20M) test background:

# CFR file:
bpid = {'NR20MHZ30KHZ','NR20MHZ30KHZ'};
fsHz = 245.76e6; % CFR rate in DFE
osr = 2; % Oversampling rate
freq = [-9.99e6, 9.99e6]; % Frequency offset for CCs

# CFR config of DUT:
"cfrPeakThreshold": 0.69 ,
"cfrInterpolationFactor": 1 (ADI_ADRV9025_CFR_INTERPOLATION_2X),
"cfrHardClipperThreshold": 0.71

# Waveform:
CC1(3615.00 MHz, CBW: 20MHz) + CC2(3634.98 MHz, CBW: 20 BW),
One ADI LO (3624.99 MHz)

Here is another 2CC (20M+20M) test, basically,

2CC could get EVM 1.73%/1.72%, ACLR L1:-47.99dB, U1:-48.20 from SA as below screenshot



By using adi_adrv9025_CfrStatusGet, almost peak waveforms are detected in the 1st CFR engine, no waveform is hardclipped, as below


However,
the dpdDirectEvm value, got by using API adi_adrv9025_DpdStatusGet to get dpdStatistics->dpdDirectEvm value,
fluctuate between 1.76 and 3.95.

Is this fluctuation dpdDirectEvm expected for 2CC waveform case? What can I check this issue further?

Here is my 2CC (20M+20M) test background:

# CFR file:
bpid = {'NR20MHZ30KHZ','NR20MHZ30KHZ'};
fsHz = 245.76e6; % CFR rate in DFE
osr = 2; % Oversampling rate
freq = [-9.99e6, 9.99e6]; % Frequency offset for CCs

# CFR config of DUT:
"cfrPeakThreshold": 0.69 ,
"cfrInterpolationFactor": 1 (ADI_ADRV9025_CFR_INTERPOLATION_2X),
"cfrHardClipperThreshold": 0.71

# Waveform:
CC1(3615.00 MHz, CBW: 20MHz) + CC2(3634.98 MHz, CBW: 20 BW),
One ADI LO (3624.99 MHz)