PlutoSDR plugin and FIR decimation

[dm6tt]

Hi Edouard,
I'm playing around with the PlutoSDR plugin on 7.22.0 an have a question about the FIR decimation.
When enabling the FIR and setting the decimation to 4 I do not encounter a reduction of the ethernet bandwidth on that interface. It stays constant.

Referring to the RX Signal Path section in https://wiki.analog.com/resources/tools-software/linux-drivers/iio-transceiver/ad9361 the FIR decimation factor can be querried on cli with:
pluto:/sys/bus/iio/devices/iio:device0]# cat rx_path_rates

Example:
Wit disabled FIR:

Leads to:
[pluto:/sys/bus/iio/devices/iio:device0]# cat rx_path_rates
BBPLL:953601026 ADC:29800032 R2:9933344 R1:4966672 RF:2483336 RXSAMP:2483336

With enabled FIR:

Leads to:[pluto:/sys/bus/iio/devices/iio:device0]# cat in_voltage_sampling_frequency
2483336

[pluto:/sys/bus/iio/devices/iio:device0]# cat rx_path_rates
BBPLL:1271468041 ADC:79466752 R2:39733376 R1:19866688 RF:9933344 RXSAMP:2483336

In both cases the sampling frequency shows the same value:
[pluto:/sys/bus/iio/devices/iio:device0]# cat in_voltage_sampling_frequency
2483336

I would have expected that the Baseband sampling rate would be decreased with the FIR decimation.

Thanks
Marcus, DM6TT

[f4exb]

That's the AD chip business not SDRangel's. In fact the baseband rate stays constant but the actual sampling frequency of the ADC is higher when decimation is engaged. It shows in the top left corner number below the start/stop button. This displays the actual ADC rate. See: https://github.com/f4exb/sdrangel/tree/master/plugins/samplesource/plutosdrinput#14-adc-sample-rate

This is also shown in the ADC value returned by the cat command.

Note that the actual decimation factor from the ADC value is 2.67 not 4. And relative to the baseband rate this is 12 and 32 respectively. For details please refer to the AD9363/9361 documentation. I suppose there are some limitations on the ADC rate and hence the actual decimation that can be achieved.

[f4exb]

This is actually explained in the link to the AD documentation you gave. In this table:

The HW decimation control only controls the decimation of the FIR filter so RF/ RXSAMP and effectively they are the same with a factor of 1: RF:2483336 RXSAMP:2483336 and RF is 4 times RXSAMP when HW decimation is 4: RF:9933344 RXSAMP:2483336 The rest of the chain is not controlled by the user.

This is the calculation for the full decimation chain:

1: BBPLL:953601026  ADC:29800032 R2:9933344  R1:4966672  RF:2483336 RXSAMP:2483336
                                 /3          /2          /2         /1      => /12
4: BBPLL:1271468041 ADC:79466752 R2:39733376 R1:19866688 RF:9933344 RXSAMP:2483336
                                 /2          /2          /2         /4      => /32  

[f4exb]

Note that if the ethernet bandwidth is your concern you may use software decimation.

[dm6tt]

Hi Edouard,
thanks for your answers.
I tested the software decimation, it is no decreasing the bandwidth.
In my opinion the SW decimation is done in SDR Angel plugin and not in the pluto itself, is that right?

I understand out of https://wiki.analog.com/resources/tools-software/linux-drivers/iio-transceiver/ad9361#rx_signal_path

that the the FIR decimation is used for decreasing the bandwidth and the smallest one is 520.83 kSPS.
I will do some further tests and come back to you.

Marcus

[dm6tt]

I've done a test with SDR Console with two differnet bandwidth
The Erhernet bandwidth gets here significantly hiher when setting the higer RF Bandwidth:
576KHz:
pluto says:
[pluto:/sys/bus/iio/devices/iio:device0]# cat rx_path_rates
BBPLL:884736002 ADC:27648000 R2:9216000 R1:4608000 RF:2304000 RXSAMP:576000

2.688MHz
pluto says:
[pluto:/sys/bus/iio/devices/iio:device0]# cat rx_path_rates
BBPLL:1032191999 ADC:32255999 R2:10751999 R1:5375999 RF:2687999 RXSAMP:2687999

Since SDR-Conosle hides the internas to the user, I can't say what happens here in detail.
The curcial point in my eyes: The RXSAMP value reflects the choosen Bandwidth and the Ethernet payloads changes as well.