The question about rx noise floor of adrv9029

The noise floor at the output of the RX channel would be -174dBm+ Gain+ Noise figure of the device.

So, -174+18+12 will be ~-144dBm/Hz at the Rx channel output.

You can probably terminate the RF input port with 50Ohm and get the nooise floor. That would be the expected noise floor instead of measuring when feeding a signal.

www.analog.com/.../noise-figure-measurement-methods-and-formulas--maxim-integrated.html

I think the ADC conversion gain is 10dB, so the noise floor will be -174+10+12 =-152dBFs/Hz at the Rx channel ADC output.

This was the spectrum when terminated the RF input port with 50Ohm, rf frequecy was 2450MHz,  BW was 245.76MHz,  gain index was 255. The rx data rate was 245.76MHz, sample time was 20mS, so frequency resolution was 50Hz,

we can see the noise floor is -120dBFs/50Hz= -137dBFs/Hz,  the difference between theoretical noise floor(-152dBFs/Hz) is 15dB, it's abnormal.

If this calculation process is reasonable, looking forward to your reply, thank you!

The RX chain has a gain of 18dB at gain index of 255. In terms of dBm, the noise floor at the output of the ADC is -144dBm/Hz. The Rx full scale input power is -~12dBm which is -2dBFS. In terms of FS values, -144dBm/Hz is equivalent to -132dBFS/Hz.

Let me know if you need any further clarifications.

You mean that the analog chain has a gain of 18dB at gain index of 255, the ADC conversion has a gain of 12dB?

18dB is the total RX chain gain and 12dB is the noise figure of the chain.

So  the noise floor at the output of the ADC is -174+18+12 =-144dBFs/Hz， not dBm/Hz。

Test noise floor from IQ datas is -120dBFs/50Hz= -137dBFs/Hz， there was 7dB gap?

Ramarao said:

The RX chain has a gain of 18dB at gain index of 255. In terms of dBm, the noise floor at the output of the ADC is -144dBm/Hz. The Rx full scale input power is -~12dBm which is -2dBFS. In terms of FS values, -144dBm/Hz is equivalent to -132dBFS/Hz.

Let me know if you need any further clarifications.

Please see this once and get back. Also, the noise figure depends on the BW, it looks like you are taking -120dBFS/Hz as the average noise floor. But if you look close to the center, the noise floor is little better.

What is your application, are you facing any issue?

The device worked at use case 51.  the issue is that noise floor of RX chain is higher, I want to check the noise foor of adrv9025.

I think NF=12dB, so the noise floor at ADC input is -162dBm/Hz,  the RX chain has a gain of 18dB, so the noise floor at the output of the ADC is -144dBFs/Hz,  why is -144dBm/Hz? 

The device worked at use case 51.  the issue is that noise floor of RX chain is higher, I want to check the noise foor of adrv9025.

I think NF=12dB, so the noise floor at ADC input is -162dBm/Hz,  the RX chain has a gain of 18dB, so the noise floor at the output of the ADC is -144dBFs/Hz,  why is -144dBm/Hz? 

Please clarify which parts was covered by RX gain? 

I stand corrected.

At ADC output, 0dBFS is 7dBm. so, -144dBm/Hz is -137dBFS/Hz.

-174dBm/Hz is thermal noise floor. So, -174+18+12 is -144dBm/Hz. FS depends on the ADC input which is 7dBm.so, -137dBFS/Hz is the correct noise floor that you are measuring

Its the red path. In general, ADC will not have any gain.

Thanks very much!

If 0dBFS is 7dBm at ADC output, then -144dBm/Hz is -151dBFS/Hz ?

The gap is 14dB between standard noise floor and measrued?

Are there any factors that worsen noise floor, clock,power?

I will get back you ASAP, working on this.

Thanks for waiting on this.

In the calculations, we are taking -120dBFS/50Hz as an approximate value. However, this is just one capture and within this capture as well, the noise is varying down to -130dBFS/50Hz as well if we zoom in. So, basically, we need to take multiple captures and then take average bin noise.

In our characterization test, we do multiple captures, get a PSD from each capture, and average each individual bin across 20 captures of the same bin.

See below, i have captured the RX samples in EVB GUI and plotted in VisualAnalog tool and its giving the expected noise floor/ noise figure.