On Nov 30, 2020, at 4:03 AM, Lukas Haase via USRP-users usrp-users@lists.ettus.com wrote:

Hello,

Does anyone know what the TX noise figure (output SNR) of the UBX160 is?
Roughly?
For simplicity, can neglect phase noise and assume white noise floor.

In words, if I transmit a full-scale sinusoidal signal, what will be the noise floor around that signal?

Of course, this will depend on TX Gain setting.

For example, a -3dBFS signal at TX gain 25 gives -3+25-5.56 = 16.44dBm. What will roughly be the noise power (and hence SNR)?

Thanks,
Lukas

PS: Usually we are interested in receiver (not transmitter) noise but the TX SNR can be relevant for self-interference.

USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Hi Marcus,
That makes sense, thanks.
Would you be willing to confirm if what I am doing here is correct?
To first order, the DAC has an SNR of 98dB (16 bit). Then I use Fries'
equation to get the NF of the following stages (for the filter and the
attenuator, the noise figure is equal to its attenuation). The NF is
dominated by the 2nd and third term.
Then I subtract the NF from the SNR which gives me an output SNR
somewhere between 92dB and 67dB. Does that sound right?

Hi Marcus,

Thanks again!

I did now the following experiment: I connected TX to RX back-to-back with
46.43dB attenuation in between. I set TX gain and RX gain to 20dB and
transmit a single CW at -3dBFS.
This means my output power is Pout=11.44dBm (cross checked with
spectrum analyzer) and on RX I sould have Pin=-34.99dBm. Indeed,
calculating the RMS of the received signal and converting to dBm, I get
Pin=-35.0224dBm. Spot on!

The red line is what I receive on the PSD (blue is the TX that I send):

As you can see from the annotation, the measured "SNR" of the received
signal is only 38.7dB. I think this is mainly caused by the phase noise
skirt (and potentially the I/Q image).
In order to keep only consider thermal noise, I add random noise to the
original CW (using randn(...)+1i*randn(...) in MATLAB) until it matches
roughly the white noise floor of the received signal. It's
SNRoutput=50dB (yellow line).

Now, according to our discussion below, at Gtx=20, we should have
SNRoutput=72dB (assuming thermal noise only).

Where could the 22dB difference in SNR come from?

Thanks!
Lukas

PS: I am aware of phase noise, DC offsets, I/Q imbalance etc. But as you
can see from my plot, I am *only *considerung thermal noise. The thermal
noise of the receiver should be orders of magnitude lower (at least
-102dBm) so the receiver noise should not limit the results either.

Gesendet: Montag, 30. November 2020 um 17:08 Uhr
Von: "Marcus D. Leech" patchvonbraun@gmail.com

An: "Lukas Haase" lukashaase@gmx.at
Cc: USRP-users@lists.ettus.com
Betreff: Re: [USRP-users] UBX160 TX "noise figure"?
On 11/30/2020 01:54 PM, Lukas Haase wrote:

Hi Marcus,

That makes sense, thanks.

Would you be willing to confirm if what I am doing here is correct?

To first order, the DAC has an SNR of 98dB (16 bit). Then I use Fries'
equation to get the NF of the following stages (for the filter and the
attenuator, the noise figure is equal to its attenuation). The NF is
dominated by the 2nd and third term.
Then I subtract the NF from the SNR which gives me an output SNR somewhere
between 92dB and 67dB. Does that sound right?

For the attenuator term, just assign it a NF (in dB) of (31.5 - TXGAIN).

The noise figure of an attenuator is just the attenuation value--similarly
for the filter.  Just pretend it's a fixed attenuator with 0 gain.

So the 'noise figure' after the DAC is just  2+(31.5 - TXGAIN) then factor
in the gains and noise figures of the amplifiers.

USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Hi Marino,
Which digital attenuator do you mean?

• The 46.43dB attenuation that I use in the TX-RX loopback is just
  composed of static, passive attenuators
• The receive gain is set to 20dB. Hence I assume I do not make use
  of the 6dB "digital gain" in the ADC and from the remaining 31.6dB
  gain I use 20. Hence I the "digital attenuator" should be 11.5dB
  if that answers your question?
• At a receive gain of 20dB, the noise figure is ~5dB, according to
  specs and measurements. -174+5+dB10(5e6)=-102dBm which is exactly
  what I measure without a signal applied.

Thanks,
Lukas
Gesendet: Montag, 07. Dezember 2020 um 03:23 Uhr
Von: "cyberphox" cyberphox@gmail.com
An: "Lukas Haase" lukashaase@gmx.at
Cc: "Marcus D. Leech" patchvonbraun@gmail.com,
USRP-users@lists.ettus.com
Betreff: Re: [USRP-users] UBX160 TX "noise figure"?
Hi Lukas,
What setting do you have the digital attenuator set to?
Kind regards
Marino
On Mon, 7 Dec 2020 at 02:05, Lukas Haase via USRP-users
<usrp-users@lists.ettus.com mailto:usrp-users@lists.ettus.com> wrote:

 Hi Marcus,
 Thanks again!
 I did now the following experiment: I connected TX to RX
 back-to-back with 46.43dB attenuation in between. I set TX gain
 and RX gain to 20dB and transmit a single CW at -3dBFS.
 This means my output power is *Pout=11.44dBm* (cross checked with
 spectrum analyzer) and on RX I sould have Pin=-34.99dBm. Indeed,
 calculating the RMS of the received signal and converting to dBm,
 I get *Pin=-35.0224dBm*. Spot on!
 The red line is what I receive on the PSD (blue is the TX that I
 send):
 As you can see from the annotation, the measured "SNR" of the
 received signal is only 38.7dB. I think this is mainly caused by
 the phase noise skirt (and potentially the I/Q image).
 In order to keep only consider thermal noise, I add random noise
 to the original CW (using randn(...)+1i*randn(...) in MATLAB)
 until it matches roughly the white noise floor of the received
 signal. It's *SNRoutput=50dB* (yellow line).
 Now, according to our discussion below, at Gtx=20, we should have
 *SNRoutput=72dB* (assuming thermal noise only).
 Where could the *22dB difference* in SNR come from?
 Thanks!
 Lukas
 PS: I am aware of phase noise, DC offsets, I/Q imbalance etc. But
 as you can see from my plot, I am /only /considerung thermal
 noise. The thermal noise of the receiver should be orders of
 magnitude lower (at least -102dBm) so the receiver noise should
 not limit the results either.
 *Gesendet:* Montag, 30. November 2020 um 17:08 Uhr
 *Von:* "Marcus D. Leech" <patchvonbraun@gmail.com
 <mailto:patchvonbraun@gmail.com>>

 *An:* "Lukas Haase" <lukashaase@gmx.at <mailto:lukashaase@gmx.at>>
 *Cc:* USRP-users@lists.ettus.com <mailto:USRP-users@lists.ettus.com>
 *Betreff:* Re: [USRP-users] UBX160 TX "noise figure"?
 On 11/30/2020 01:54 PM, Lukas Haase wrote:

     Hi Marcus,
     That makes sense, thanks.
     Would you be willing to confirm if what I am doing here is
     correct?
     To first order, the DAC has an SNR of 98dB (16 bit). Then I
     use Fries' equation to get the NF of the following stages (for
     the filter and the attenuator, the noise figure is equal to
     its attenuation). The NF is dominated by the 2nd and third term.
     Then I subtract the NF from the SNR which gives me an output
     SNR somewhere between 92dB and 67dB. Does that sound right?

 For the attenuator term, just assign it a NF (in dB) of (31.5 -
 TXGAIN).

 The noise figure of an attenuator is just the attenuation
 value--similarly for the filter. Just pretend it's a fixed
 attenuator with 0 gain.

 So the 'noise figure' after the DAC is just  2+(31.5 - TXGAIN)
 then factor in the gains and noise figures of the amplifiers.


 _______________________________________________
 USRP-users mailing list
 USRP-users@lists.ettus.com <mailto:USRP-users@lists.ettus.com>
 http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Hi Marcus,
Thanks again!
I did now the following experiment: I connected TX to RX back-to-back
with 46.43dB attenuation in between. I set TX gain and RX gain to 20dB
and transmit a single CW at -3dBFS.
This means my output power is Pout=11.44dBm (cross checked with
spectrum analyzer) and on RX I sould have Pin=-34.99dBm. Indeed,
calculating the RMS of the received signal and converting to dBm, I
get Pin=-35.0224dBm. Spot on!
The red line is what I receive on the PSD (blue is the TX that I send):
As you can see from the annotation, the measured "SNR" of the received
signal is only 38.7dB. I think this is mainly caused by the phase
noise skirt (and potentially the I/Q image).
In order to keep only consider thermal noise, I add random noise to
the original CW (using randn(...)+1i*randn(...) in MATLAB) until it
matches roughly the white noise floor of the received signal. It's
SNRoutput=50dB (yellow line).
Now, according to our discussion below, at Gtx=20, we should have
SNRoutput=72dB (assuming thermal noise only).
Where could the 22dB difference in SNR come from?
Thanks!
Lukas
PS: I am aware of phase noise, DC offsets, I/Q imbalance etc. But as
you can see from my plot, I am /only /considerung thermal noise. The
thermal noise of the receiver should be orders of magnitude lower (at
least -102dBm) so the receiver noise should not limit the results either.