# AD9364 baseband digital filter design

We are using AD9364 in one of the project where we need to achieve around 58 - 60dB ACPR (Adjacent Channel Power Ratio) for 6MHz OFDM waveform with PAPR(Peak to Average Power Ratio) of 12dB. We are passing the input baseband samples to a mask filter whose output has 65dB ACPR in baseband. This filtered output is given to AD9364 input.

We used ADI's MATLAB filter wizard to generate the FIR coefficients with the sampling rate of 6.912MSPS at the AD and base-band interface. At AD9364 Tx output, we are getting ACPR of 50dB only, irrespective of the Tx attenuation of 5 and 10dB. Now, we tried to increase the rejection in AD9364 filter by using the wizard. In wizard, it shows the rejection of 70dB at 3.1MHz away from the center frequency. But when we loaded those coefficients in the Evaluation board, we are getting a rejection of just 20dB at 3.1MHz away from center frequency.

Why are we seeing this mismatch between the filter response plot by MATLAB wizard and eval board response when loaded with the same coefficient?

Is there anyway to get 58dB ACPR in AD9364?

We ensured that the input baseband signal has the ACPR of 65dB.

Parents
• Can you provide the generated filter and a screenshot of TX and RX configurations of the Wizard.

-Travis

• I've attached the filter wizard screenshot for the coefficients we are using,

The screenshot of Wizard window,

The screenshot of the filter response,

The filter coefficients are attached below,

```# Generated with AD9361 Filter Design Wizard 16.1.3
# MATLAB 9.2.0.556344 (R2017a), 23-Jul-2019 18:55:25
# Inputs:
# Data Sample Frequency = 6912000 Hz
TX 3 GAIN 0 INT 2
RX 3 GAIN -6 DEC 2
RTX 884736000 110592000 55296000 27648000 13824000 6912000
RRX 884736000 110592000 55296000 27648000 13824000 6912000
BWTX 6421171
BWRX 6337804
-82,101
-183,96
-248,16
-155,-38
24,-70
130,65
53,48
-89,12
-99,-112
39,-15
126,72
23,100
-125,-73
-95,-119
91,-8
153,161
-18,69
-186,-130
-77,-164
166,80
179,215
-94,29
-250,-241
-31,-147
268,192
175,271
-207,-87
-310,-346
71,-81
384,356
126,267
-367,-264
-334,-437
235,81
501,532
-2,178
-564,-514
-296,-457
482,351
585,695
-235,-50
-788,-813
-146,-353
818,750
599,780
-621,-464
-1015,-1134
177,-38
1272,1296
470,706
-1242,-1164
-1224,-1436
827,650
1932,2082
42,295
-2397,-2451
-1398,-1701
2365,2302
3302,3632
-1434,-1210
-6091,-6429
-1615,-2157
12717,12783
25904,26752
25904,26752
12717,12783
-1615,-2157
-6091,-6429
-1434,-1210
3302,3632
2365,2302
-1398,-1701
-2397,-2451
42,295
1932,2082
827,650
-1224,-1436
-1242,-1164
470,706
1272,1296
177,-38
-1015,-1134
-621,-464
599,780
818,750
-146,-353
-788,-813
-235,-50
585,695
482,351
-296,-457
-564,-514
-2,178
501,532
235,81
-334,-437
-367,-264
126,267
384,356
71,-81
-310,-346
-207,-87
175,271
268,192
-31,-147
-250,-241
-94,29
179,215
166,80
-77,-164
-186,-130
-18,69
153,161
91,-8
-95,-119
-125,-73
23,100
126,72
39,-15
-99,-112
-89,12
53,48
130,65
24,-70
-155,-38
-248,16
-183,96
-82,101
```

• What is your LO frequency? You may get better ACPR at lower LO frequency.

As per datasheet, the typical ACPR value is around 50 dBc.

• My LO frequency is 725MHz.

• Is it possible to test your waveform using our EVB?.

• ofdm_eval_transpose.zip

I've attached the .mat data along with the filter coefficeints which you can load in the EVB.

Children
• We are checking on this. Meanwhile, I am attaching here the LTE standard 5MHz waveform and .ftr file.

Can you please try measuring ACLR with these files and check whether you are seeing any improvement.

Waveform file

lte_5mhz_7p68datarate_64QAM.txt

Filter File,

LTE5_MHz.zip

• Hi,

In the filter file that you shared above, what is the sampling rate of the TX and Rx chain?

I added sampling rates of Tx/Rx chain as per our requirement. For this waveform and settings, we are only getting 54dB ACPR at 2.5MHz from the LO frequency.

• In the filter file that you shared above, what is the sampling rate of the TX and Rx chain?

The sampling rate is 7.68MSPS.

Attached the file which supports IIO Oscilloscope.

8737.LTE5_MHz.zip

Can you test with our waveform file and the attached profile?

• We have checked the LTE waveform that you provided.

Followings are my concerns,

1) We did the digital loopback(digital Tx -> digital Rx) and capture the response in IIO Oscilloscope GUI and the spectrum is attached below,

Here it seems that the data itself is passed through some mask filter to give a sharp cut-off at 2.5MHz from center frequency.

2) But the output spectrum looks like below,

Here, at 2.5MHz away we are getting only 47dB ACLR. There seems a hump in the spectrum and we want to know why this hump is getting created. Without it, we are getting 53dB ACLR.

Why are we getting this hump near the edge of the spectrum?

Is it because of some noises getting introduced by AD9364?

All the test are done in Eval board. In our custom board also we are facing the same issue.

• Please find below results from our testing with LTE signal and filter file shared earlier.

As per LTE standard, some roll-off (Guard Band) is allowed.

For ACLR measurement you have to take the channel power measurement.

Can you please try to measure ACLR a similar way for your custom board, filter, and waveform file?.