Hi USRP users,

I have a (rather strong ?) 1/f shaped noise floor with the SBX daughterboards (rev 4 I think, acquired end of 2012).
Running the uhd-calibration utilities does not give any visible difference, see links to pics below.
As a comparison, the noise floor on the RFX2400 db is rather flat. Maybe generally a little higher but without the strong 1/f peak.

Q1: Is this a normal and expected (1/f colored) noise floor with the SBX db?
If I pump up the uhd gain (>0 dB) the 1/f shaped noise floor remains and gets amplified correspondingly.
I know 1/f noise is a common notorious feature in electronics so perhaps it is included with the SBX too.

Q2: Should I see (or not) a visible improvement with the uhd-calibration enabled (i.e., with the calibration values in .uhd/cal/) ?
With the calibration enabled the gr-application tells me that the appropriate cal values are "loaded", which sounds reassuring…  but maybe I do something wrong.

Link to picture of PSD without calibration:
https://www.dropbox.com/s/vl7x8c8i7qn3ueh/UHD_uncalibrated_SBX.png

Link to picture of PSD with calibration (its identical to my eyes):
https://www.dropbox.com/s/ngc98l71ugifqs4/UHD_calibrated_SBX.png

Thanks for any comments,
Rickard

The SBX is a direct-conversion receiver, so you'll see some 1/f noise. You can use offset tuning to move DC out of your passband, and as a happy consequence, move the 1/f noise out of your passband as well.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

Hi USRP users,

I have a (rather strong ?) 1/f shaped noise floor with the SBX daughterboards (rev 4 I think, acquired end of 2012).
Running the uhd-calibration utilities does not give any visible difference, see links to pics below.
As a comparison, the noise floor on the RFX2400 db is rather flat. Maybe generally a little higher but without the strong 1/f peak.

Q1: Is this a normal and expected (1/f colored) noise floor with the SBX db?
If I pump up the uhd gain (>0 dB) the 1/f shaped noise floor remains and gets amplified correspondingly.
I know 1/f noise is a common notorious feature in electronics so perhaps it is included with the SBX too.

Q2: Should I see (or not) a visible improvement with the uhd-calibration enabled (i.e., with the calibration values in .uhd/cal/) ?
With the calibration enabled the gr-application tells me that the appropriate cal values are "loaded", which sounds reassuring… but maybe I do something wrong.

Link to picture of PSD without calibration:
https://www.dropbox.com/s/vl7x8c8i7qn3ueh/UHD_uncalibrated_SBX.png

Link to picture of PSD with calibration (its identical to my eyes):
https://www.dropbox.com/s/ngc98l71ugifqs4/UHD_calibrated_SBX.png

Thanks for any comments,
Rickard
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

How much bandwidth are you using?

With offset tuning, your desired center frequency is still slap-dab in the middle, but UHD has arranged to tune the analog hardware differently, and then use the DDC to move DC to somewhere else, and still have your desired Fc exactly where you want it.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

OK, thanks! However, I want to utilize all the bandwidth I can get so the passband need to stay in the centre. As a consequence then also the 1/f noise peak has a relatively small total power, except when the received signal is very weak which leads me to another related question I have...
On May 29, 2013, at 6:26 PM, Marcus Leech <mleech@ripnet.com> wrote:

The SBX is a direct-conversion receiver, so you'll see some 1/f noise. You can use offset tuning to move DC out of your passband, and as a happy consequence, move the 1/f noise out of your passband as well.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

Hi USRP users,

I have a (rather strong ?) 1/f shaped noise floor with the SBX daughterboards (rev 4 I think, acquired end of 2012).
Running the uhd-calibration utilities does not give any visible difference, see links to pics below.
As a comparison, the noise floor on the RFX2400 db is rather flat. Maybe generally a little higher but without the strong 1/f peak.

Q1: Is this a normal and expected (1/f colored) noise floor with the SBX db?
If I pump up the uhd gain (>0 dB) the 1/f shaped noise floor remains and gets amplified correspondingly.
I know 1/f noise is a common notorious feature in electronics so perhaps it is included with the SBX too.

Q2: Should I see (or not) a visible improvement with the uhd-calibration enabled (i.e., with the calibration values in .uhd/cal/) ?
With the calibration enabled the gr-application tells me that the appropriate cal values are "loaded", which sounds reassuring… but maybe I do something wrong.

Link to picture of PSD without calibration:
https://www.dropbox.com/s/vl7x8c8i7qn3ueh/UHD_uncalibrated_SBX.png

Link to picture of PSD with calibration (its identical to my eyes):
https://www.dropbox.com/s/ngc98l71ugifqs4/UHD_calibrated_SBX.png

Thanks for any comments,
Rickard
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Yes, the lo_off parameter basically tells UHD to program the analog hardware for an offset (by lo_off) tuned frequency, and then the DDC in the FPGA downconverts so that your desired center frequency is still at DC in the digital sample stream, but the analog DC is offset (usually to somewhere outside your passband). This is conceptually very similar to a normla superhet approach with an actual real IF, rather than a zero IF.

I encourage you to try it. Make your lo_off be half your passband bandwidth, which will place "analog DC" at the edge of your passband.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

All I can get, i.e. 25 MHz 16 bit or 50 MHz 8 bit.

I am not sure what you mean. Do you mean that with a certain wanted analog carrier (center) frequency, fc, which is what I normally use but then also with this "offset tuning", I still receive/transmit the whole bandwidth centered around fc in RF (analog domain in the air) but have just (circularly) shifted the whole spectrum after sampling in the digital domain which I can then just undo in digital domain by the DDC with the uhd? Or maybe I got this totally wrong…

Is "offset tuning" equivalent to the "lo_off" parameter to use with the "uhd.tune_request(fc, lo_off)" ?

I never really understood that parameter until perhaps now… I will try that tomorrow!

On May 29, 2013, at 6:39 PM, Marcus Leech <mleech@ripnet.com> wrote:

How much bandwidth are you using?

With offset tuning, your desired center frequency is still slap-dab in the middle, but UHD has arranged to tune the analog hardware differently, and then use the DDC to move DC to somewhere else, and still have your desired Fc exactly where you want it.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

OK, thanks! However, I want to utilize all the bandwidth I can get so the passband need to stay in the centre. As a consequence then also the 1/f noise peak has a relatively small total power, except when the received signal is very weak which leads me to another related question I have...
On May 29, 2013, at 6:26 PM, Marcus Leech <mleech@ripnet.com> wrote:

The SBX is a direct-conversion receiver, so you'll see some 1/f noise. You can use offset tuning to move DC out of your passband, and as a happy consequence, move the 1/f noise out of your passband as well.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

Hi USRP users,

I have a (rather strong ?) 1/f shaped noise floor with the SBX daughterboards (rev 4 I think, acquired end of 2012).
Running the uhd-calibration utilities does not give any visible difference, see links to pics below.
As a comparison, the noise floor on the RFX2400 db is rather flat. Maybe generally a little higher but without the strong 1/f peak.

Q1: Is this a normal and expected (1/f colored) noise floor with the SBX db?
If I pump up the uhd gain (>0 dB) the 1/f shaped noise floor remains and gets amplified correspondingly.
I know 1/f noise is a common notorious feature in electronics so perhaps it is included with the SBX too.

Q2: Should I see (or not) a visible improvement with the uhd-calibration enabled (i.e., with the calibration values in .uhd/cal/) ?
With the calibration enabled the gr-application tells me that the appropriate cal values are "loaded", which sounds reassuring… but maybe I do something wrong.

Link to picture of PSD without calibration:
https://www.dropbox.com/s/vl7x8c8i7qn3ueh/UHD_uncalibrated_SBX.png

Link to picture of PSD with calibration (its identical to my eyes):
https://www.dropbox.com/s/ngc98l71ugifqs4/UHD_calibrated_SBX.png

Thanks for any comments,
Rickard
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Try setting your RF gain to somewhere mid-range, and see how that changes things.

-104dB FS is a very, very, very tiny DC offset, likely due to a slight numerical quantization bias somewhere in the chain. With any kind of actual signal, at any kind of reasonable gain levels, this very narrow spike at DC wil have effectively-zero effect on anything--it's a tiny amount of additional noise added to your signal
.

on May 30, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

Works great - it was easy to move the annoying 1/f noise out of my passband!

However, I still have a fairly high DC offset left - a strong spike exactly at zero digital freq (almost 20 dB above the noise floor).

https://www.dropbox.com/s/1lfhvr73405zvht/UHD_lo-offset.png

This DC-offset peak stays independent of the chosen offset tuning (lo_off) and center frequency, bandwidth, etc.

Where does this come from and it is somehow possible to get rid of? (Not that it seem to hurt as much as the 1/f noise)

Using the UHD calibrated values, or not, does not result in any difference. Why?

Rickard

On May 29, 2013, at 7:06 PM, Marcus Leech <mleech@ripnet.com> wrote:

Yes, the lo_off parameter basically tells UHD to program the analog hardware for an offset (by lo_off) tuned frequency, and then the DDC in the FPGA downconverts so that your desired center frequency is still at DC in the digital sample stream, but the analog DC is offset (usually to somewhere outside your passband). This is conceptually very similar to a normla superhet approach with an actual real IF, rather than a zero IF.

I encourage you to try it. Make your lo_off be half your passband bandwidth, which will place "analog DC" at the edge of your passband.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

All I can get, i.e. 25 MHz 16 bit or 50 MHz 8 bit.

I am not sure what you mean. Do you mean that with a certain wanted analog carrier (center) frequency, fc, which is what I normally use but then also with this "offset tuning", I still receive/transmit the whole bandwidth centered around fc in RF (analog domain in the air) but have just (circularly) shifted the whole spectrum after sampling in the digital domain which I can then just undo in digital domain by the DDC with the uhd? Or maybe I got this totally wrong…

Is "offset tuning" equivalent to the "lo_off" parameter to use with the "uhd.tune_request(fc, lo_off)" ?

I never really understood that parameter until perhaps now… I will try that tomorrow!

On May 29, 2013, at 6:39 PM, Marcus Leech <mleech@ripnet.com> wrote:

How much bandwidth are you using?

With offset tuning, your desired center frequency is still slap-dab in the middle, but UHD has arranged to tune the analog hardware differently, and then use the DDC to move DC to somewhere else, and still have your desired Fc exactly where you want it.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

OK, thanks! However, I want to utilize all the bandwidth I can get so the passband need to stay in the centre. As a consequence then also the 1/f noise peak has a relatively small total power, except when the received signal is very weak which leads me to another related question I have...
On May 29, 2013, at 6:26 PM, Marcus Leech <mleech@ripnet.com> wrote:

The SBX is a direct-conversion receiver, so you'll see some 1/f noise. You can use offset tuning to move DC out of your passband, and as a happy consequence, move the 1/f noise out of your passband as well.

on May 29, 2013, Rickard Radio <rickardradio@gmail.com> wrote:

Hi USRP users,

I have a (rather strong ?) 1/f shaped noise floor with the SBX daughterboards (rev 4 I think, acquired end of 2012).
Running the uhd-calibration utilities does not give any visible difference, see links to pics below.
As a comparison, the noise floor on the RFX2400 db is rather flat. Maybe generally a little higher but without the strong 1/f peak.

Q1: Is this a normal and expected (1/f colored) noise floor with the SBX db?
If I pump up the uhd gain (>0 dB) the 1/f shaped noise floor remains and gets amplified correspondingly.
I know 1/f noise is a common notorious feature in electronics so perhaps it is included with the SBX too.

Q2: Should I see (or not) a visible improvement with the uhd-calibration enabled (i.e., with the calibration values in .uhd/cal/) ?
With the calibration enabled the gr-application tells me that the appropriate cal values are "loaded", which sounds reassuring… but maybe I do something wrong.

Link to picture of PSD without calibration:
https://www.dropbox.com/s/vl7x8c8i7qn3ueh/UHD_uncalibrated_SBX.png

Link to picture of PSD with calibration (its identical to my eyes):
https://www.dropbox.com/s/ngc98l71ugifqs4/UHD_calibrated_SBX.png

Thanks for any comments,
Rickard
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com