ADRV9025 ETM2 ISSUE

You can try DPD mode 2 , where M table is used for high power model and C table for low power. Details are available in the user guide. 

ez.analog.com/.../adrv9029-dpd-and-cfr-user-guide

DEAR. Vinod

Have you ever DPD tested with ETM2?
If you did the test, please tell us your case.

Thank you.

In Normal scenario's , since the Avg power of TM2 signals are much lower (I guess 22 db lower), Normally PA's are inherently linear and DPD is bypassed in most customer application.

You can set  minAvgSignalLevel in DpdTrackingConfig_t to deal with low power conditions in TM2.0 conditions.

For same reason we have not included TM2 testing in our normal DPD evaluation. 

Have you tried testing DPD with ETM3.1 with digital power scaled down by 20 db? The DPD performance for low power depends on the PA as well. 

If you need DPD to work at low power you should use DPD mode2 and C table as explained earlier.

DEAR. VINOD

Our system specifications must support both ETM2 and TM3.
By changing PeakSearchWindowSize in ETM2, stability Error disappears and DPD operates.
However, in TM3, the ACLR characteristics deteriorate compared to before the ACLR change.
ACLR is reduced by about -10dB.

The level difference between ETM2 and TM3 is 6dB.
ETM2 is 6dB lower.

Our setup is below.
How can I improve performance?

adrv9025_0 DPD Status:
ErrCode = 0x341d
PercentComplete = 0
IterCount = 383
UpdateCount = 0
SyncStatus = 4 (1=OK)
ModelTable = M (Table)
Statistics:
MeanTuPower = -30.57
PeakTuPower = -26.97
MeanTxPower = -30.57
PeakTxPower = -26.97
MeanOrxPower = -31.33
PeakOrxPower = -27.50
DirectEvm = 0.06
IndirectEvm = 0.06
SelectError = 0.04
IndirectError = 0.05
ErrorStatus0:
MetricsMask = 0x36a
ActionMask = 0x1
ErrorStatus1:
MetricsMask = 0
ActionMask = 0
PersistentErrorStatus0:
MetricsMask = 0
ActionMask = 0
PersistentErrorStatus1:
MetricsMask = 0
ActionMask = 0

# DPD configurations
dpd.cfg.dpdIndirectRegularizationValue = 25
dpd.cfg.dpdDirectRegularizationValue = 25
dpd.cfg.dpdSamples = 16384
dpd.cfg.dpdMThreshold = 13517608
dpd.cfg.dpdPeakSearchWindowSize = 1024
dpd.cfg.dpdUpdateMode = 2
dpd.cfg.minAvgSignalLevel = 519
dpd.cfg.dpdMu = 50
dpd.cfg.minAvgSignalLevelOrx = 519
dpd.cfg.dpdFilterSel = 0
dpd.cfg.enableDirectLearning = 0
dpd.cfg.dpdIndirectRegularizationLowPowerValue = 25
#dpd.cfg.dpdDecayP = 2

Thank you

Henry J said:

UpdateCount = 0

It looks like the DPD update never happened looking at the update count of 0.

Henry J said:

dpd.cfg.dpdMThreshold = 13517608

This sets the M-table threshold as -19dBFS(13517608) which is why the DPD update never happened as the input signal is below this threshold.

Can you set the M-table threshold to -40dBFS (107374) and check if it works?

DEAR. RAMARAO

In some cases, it operates normally with the above setting values. DPD unstable does not always occur.
In some cases, it operates normally with the above setting values. DPD unstable does not always occur.

If it is abnormal, I do not know why it is not updated in M-table mode.
Since MODE-2 is used, in the range of -19dbfs to -36dBFS, C-table is updated and DPD should operate in normal cases.

Another problem arises with ETM3.1 when setting the M-table threshold below -19dBFS in MODE-2. (ACLR degradation to a certain level)
I am using it with -19dBFS set.

As an additional experiment, after setting the M table threshold of -25dBFS and changing "dpd.cfg.dpdPeakSearchWindowSize" from 1024 -> 491520, DPD operates stably in ETM2/ETM3.1, but there is a problem that EVM characteristics deteriorate.
As an additional experiment, after setting the M table threshold of -25dBFS and changing "dpd.cfg.dpdPeakSearchWindowSize" from 1024 -> 491520, DPD operates stably in ETM2/ETM3.1, but there is a problem that EVM characteristics deteriorate.

How can I improve performance?
How can I satisfy the characteristics of TM2 and TM3.1?

DPD NG : 0x341d DPD stability error.
ACLR NG: ACLR degradation in a specific level section.
EVM NG: EVM characteristic degradation.

Thank you.

With M Threshold of -19dBFS, and Peak search window Size of 491520, how much is the EVM going bad from a good case? Are you using CFR?

With M threshold of -25dBFS, Peak Search Window size of 491520, how much ACLR is going bad? Can you increase the Window size to 4*491520 and play with Regularisation value ? Hope you are using indirect learning.

DEAR.RAMARAO

I will reply as below.
Please let us know how we can improve it.

With M Threshold of -19dBFS, and Peak search window Size of 491520, how much is the EVM going bad from a good case?
A) ETM2 is measured around 5%.
Deterioration occurs about 3-4%.

Are you using CFR?
A) CFR is bypassing.

With M threshold of -25dBFS, Peak Search Window size of 491520, how much ACLR is going bad?
A)The problem with ACLR is that when M-threshold is set to -25dBFs, DPD does not work in the range of -15dBFS to -25dBFS.
At -25dBFS, both ETM2 and TM3 do not work.
And M-TABLE is not updated.

Can you increase the Window size to 4*491520 and play with Regularisation value ?
A) dpd.cfg.dpdIndirectRegularizationValue = 25
dpd.cfg.dpdDirectRegularizationValue = 25
dpd.cfg.dpdIndirectRegularizationLowPowerValue = 25

Hope you are using indirect learning.
A)I was still testing with indirect .

Thank you

Henry J said:

A) CFR is bypassing.

Are you using pre-clipped waveform? If not, can you test with a pre-clipped waveform. Also, from the log that you shared, the PAR is around 3.5dB. Which type of signal are you using?

Whats the digital input signal level in both ETM2 and ETM3 case? 

ETM3 with -13dBFS back off signal level is what generally used. From the log, that you shared the signal level looks to be of -30dBFS level which is lower, Is it for ETM2 ?

I would suggest to set the M-table threshold to -40dBFS in both ETM cases and the input signal level of -13dBFS.

Henry J said:

Can you increase the Window size to 4*491520 and play with Regularisation value ?

Please test with this case as well.

DEAR. RAMARAO

Our system specification should support both TM2 and TM3.
With one API setting, both TM2 and TM3 should work fine.

Can't support both TM2 and TM3 with one API setting of ADRV9025?
Have you ever tested TM2 on ADRV9025EVB?

Mass production requires solving this problem.
We don't have time.

We answer your questions as follows.

Q.Are you using pre-clipped waveform? If not, can you test with a pre-clipped waveform. Also, from the log that you shared, the PAR is around 3.5dB. Which type of signal are you using?

A)I don't know what the "pre-clipeed waveform" means.
what does it mean?

A)If it means CFR, the function is used on the MU side, and the adrv9025 built-in CFR function is not used. (The structure is simply shown below.)
MU(RF) -> MU(ADC(AFE7920)) -> MU(DSP,CFR) -> CPRI(OPTIC) -> RU(DSP) -> RU(DAC,ADRV9025,DPD) -> RU(RF,LTE850M)

A) ETM2 log.

Q.Whats the digital input signal level in both ETM2 and ETM3 case?
A)ETM2 : -25dBFS (Measured by RU DSP.)
ETM3 : -15dBFS (Measured by RU DSP.)

Q.ETM3 with -13dBFS back off signal level is what generally used. From the log, that you shared the signal level looks to be of -30dBFS level which is lower, Is it for ETM2 ?
A) ETM2 log.

Q.I would suggest to set the M-table threshold to -40dBFS in both ETM cases and the input signal level of -13dBFS.

A)If the M-table threshold value is less than -19dBFs, DPD does not work or the side effect of slowing the reaction speed is severe, so it cannot be applied.
For reference, the provided reference value was -15dBFS, which was set to a value that satisfies both the DPD response speed and ACLR in EMT3.

Q.Can you increase the Window size to 4*491520 and play with Regularisation value ?
A)If the window size is increased, DPD stability increases, but ACLR characteristics deteriorate.

Thank you