OK, so this means the 1/f noise is not from the LO. Otherwise it would finally appear around the LO offset frequency.
The noise arises in the baseband AFTER the mixing and therefore is not affected by the DSP magic. I think, I understand now. Erik ________________________________ Von: Nick Foster <bistrom...@gmail.com> Gesendet: Mittwoch, 5. Juni 2019 19:15 An: Erik Heinz Betreff: Re: [USRP-users] B210: 1/f noise and LO offset If you specify an LO offset, UHD will make sure that your center frequency still arrives at baseband, by using the CORDIC in the DDC to downconvert it. Nick On Wed, Jun 5, 2019 at 1:44 AM Erik Heinz via USRP-users <usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com>> wrote: Thank you all for the comments. Some summarized remarks: Regarding the external reference: I have no GPSDO installed and selected "Clock Source: external" in the GRC UHD blocks. I cannot tell definitely if the B210 is locked to the external reference, but there is an error message if I disconnect the external reference, so it probably is. Regarding the noise spectrum: Yes, I am simply capturing the output signal of the UHD USRP Source without an external signal. I noticed that below 40dB RX gain the noise floor is mostly independent from gain. Above 40dB the amplifier noise is added. So I measured at 40dB RX gain and assume this noise floor to be the absolute limit for the SNR when using the B210, regardless which LNAs are used. My guess was that the 1/f noise originates from semiconductor components, might be the mixer. LO phase noise would add to the noise floor as well, as I learned. For the actual device the origin of the noise floor is of course only of academic interest, since it cannot be changed, but some deeper understanding might help for future projects. What I really do not understand is why I can shift the spectrum using an external software modulator but not by using the LO offset. Best regards. Erik ________________________________ Von: USRP-users <usrp-users-boun...@lists.ettus.com<mailto:usrp-users-boun...@lists.ettus.com>> im Auftrag von Torell, Kent L via USRP-users <usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com>> Gesendet: Dienstag, 4. Juni 2019 18:44 An: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> Betreff: Re: [USRP-users] B210: 1/f noise and LO offset I think I’m misunderstanding what you are measuring. From my radio experience, I immediately assume you are measuring the phase noise of a synthesized tone. But the spectrum you see isn’t quite right for that. Re-reading your original post, are you simply capturing the noise of the USRP, without any signal input at all? In that case, you are plotting the noise spectra, which will have a quantized noise floor, thermal noise, and 1/f noise from the baseband amplifiers. The tuning of the synthesizer most likely won’t affect this. Kent Torell From: USRP-users <usrp-users-boun...@lists.ettus.com<mailto:usrp-users-boun...@lists.ettus.com>> On Behalf Of Lundberg, Daniel via USRP-users Sent: Tuesday, June 4, 2019 9:00 AM To: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> Subject: Re: [USRP-users] B210: 1/f noise and LO offset Just a couple of sanity checks on this…You have to make sure that 1) You don’t already have a GPSDO installed on your B210. If you do, the external reference won’t make it in. 2) Assuming you don’t have the GPSDO, you have to tell the B210 to use the external reference instead of its own. Also, from past experience, I have found that avoiding an LO offset that is an integer multiple of the sampling rate will avoid some pretty bad spurs. I think these are integer boundary spurs in the PLL. Dan Lundberg From: USRP-users <usrp-users-boun...@lists.ettus.com<mailto:usrp-users-boun...@lists.ettus.com>> On Behalf Of Erik Heinz via USRP-users Sent: Tuesday, June 4, 2019 10:10 AM To: Torell, Kent L <kent.tor...@gd-ms.com<mailto:kent.tor...@gd-ms.com>> Cc: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> Subject: Re: [USRP-users] B210: 1/f noise and LO offset Thank you for the explanation. I tried using an external reference clock (HP 58503A). Unexpectedly, this made no difference at all. The noise is still exactly the same. Erik Heinz -- ---------------------------------------- Supracon AG Dr. Erik Heinz An der Lehmgrube 11 07751 Jena Tel.: +49 3641 2328-165 Fax: +49 3641 2328-109 Internet: http://www.supracon.com ---------------------------------------- Handelsregister: Amtsgericht Gera HRB 208970 Umsatzsteuer-Id.: DE 216 111 685 Kaufm. Vorstand: Matthias Meyer Vorsitz Aufsichtsrat: Prof. Dr. Michael Siegel ________________________________ Von: Torell, Kent L <kent.tor...@gd-ms.com<mailto:kent.tor...@gd-ms.com>> Gesendet: Montag, 3. Juni 2019 17:10 An: Erik Heinz Cc: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> Betreff: RE: B210: 1/f noise and LO offset The phase noise comes from the B210 LO (RF synthesizer), and is present regardess of the frequency offset…you are misled by the log axis of the plot. The close-in noise can be improved by using a high quality external 10 MHz source. The control loop action of the synthesizer translates the reference phase noise to the LO phase noise. You may have a lab standard available; or you could purchase a GPSDO version for the B210 (Ettus part 783454-01). If this doesn’t meet your needs, you will need to shift the signal lower in frequency with an external mixer and a high quality microwave synthesizer (e.g. $20K Rhode/Keysight/etc.) that has the phase noise you require. The B210 uses an Analog Devices AD9361 chip, which generates the RF local oscillator signal with a high frequency phase locked loop, then divides it down to the requested frequency. 5 GHz is at the upper end of it’s range, so the division is small and the phase noise will be at it’s highest. Moving down several octaves will improve the phase noise, dropping 6 dB for every octave (e.g. 500 MHz would have 20 dB lower phase noise than 5 GHz). Kent Torell From: USRP-users <usrp-users-boun...@lists.ettus.com<mailto:usrp-users-boun...@lists.ettus.com>> On Behalf Of Erik Heinz via USRP-users Sent: Monday, June 3, 2019 3:06 AM To: usrp-users@lists.ettus.com<mailto:usrp-users@lists.ettus.com> Subject: [USRP-users] B210: 1/f noise and LO offset Hello everyone, preliminary remark: I am an physicist - so please excuse some possible knowledge gaps in radio engineering :-). I would like to use a B210 to measure the IQ response of superconducting resonators at about 5 GHz. This is done by feeding the resonators a signal from the transmitter with a frequency near the resonance, amplifying by LNAs, coupling to the receiver, demodulating to base band, and recording the resulting IQ signal. The signal bandwidth of interest will be small, in the kHz range. Since noise of the resonators has to be measured as well, I first had a look at the noise of the B210 output without an external signal. The result is plotted in figure 1 (vertical axis is scaled to the output range of the ADC). It seems, below 100kHz the B210 receiver has strong 1/f noise. I would guess this is hardware noise from the mixer. Anyone knows if this assumption is correct? So I thought it would be a good idea to demodulate not directly to base band, but to an IF of may be 100 kHz and do the rest in software. I did a quick test using gnu radio, feeding the output of the USRP source and a 100 kHz signal to a multiplier. This indeed has the desired effect (figure 2, blue curve vs. red curve). If I understand the concept of the B210 local oscillators correctly, the same should be possible directly on the B210, using an LO offset of 100 kHz. The result, however, is completely different (figure 2, green curve). Any ideas why? Thank you. Best regards, Erik _______________________________________________ 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
