Hi everyone,
although there are some comparisons between the R820T and the E4000 already [1, 2], I also did some tests with another use-case in mind. I'm working on a thing similar to RTLSDR-Scanner [3]. I want to monitor a large part of the spectrum continuously. So I compared the R820T with the E4000 using RTLSDR-Scanner w/ and w/o an antenna. My results are here: https://docs.google.com/folder/d/0ByDAKwyEiyx_XzZ5ZnpRV1VZWDQ/edit?usp=shari... There's much more spurs with the E4000 than w/ the R820T. According to [1, 2] one also would expect a better overall sensivity compared to the E4000. However, the GSM900 signals for example seem to be way better with the E4000 according to the RTLSDR-Scanner. Tuning to a certain channel w/ OsmoSDR Source in GNUradio gives about the same SNR - contrary to the RTLSDR-Scanner output. Can anyone explain this? Also, the DVB-T channel at 502MHz is quite weak in the R820T RTLSDR-Scanner output when compared to the E4000. I had a closer look at the lower limit of the channel in gnuradio. This can be seen in the 502MHz_*.png pictures. The E4000 produces a nice +20dB step while one can hardly see the channel in the R820T spectrum. I don't understand this as well. Is this AGC-related? Manually setting a fixed gain didn't really help though...
Any explanations?
Thank you!
Best regards,
Hunz
[1] http://steve-m.de/projects/rtl-sdr/tuner_comparison/ [2] http://www.hamradioscience.com/rtl2832u-r820t-vs-rtl2832u-e4000/#more-1852 [3] https://github.com/EarToEarOak/RTLSDR-Scanner
Well, there can be lots of possible reasons... They are both direct conversion receivers, with local oscillator going from low vhf to high uhf band. We can expect lost of silicium bugs here : power level of the local oscillators frequency dependant, quadrature mismatch, mixers with conversion loss not constant etc. etc. honestly it is not easy to know; I guess this also depends on the part itself. My experience show that from one E4000 to another, the preamps can completely behave differently. On some units, when manually set to more than 17db, it works as an attenuator :-( the power level received decreases....
For me the most interesting plots were the "no antenna" ones, showing the LO leakage ;-) If I understood clearly your plots, you have at some places high power ghosts. Maybe you are close to powerful transmitters, but this is more probably a LO to RF isolation problem: the receiver receives himself...
I usually play with the E4000, but I am waiting for some 820T to try.
2013/3/15 Benedikt Heinz zn000h@gmail.com
Hi everyone,
although there are some comparisons between the R820T and the E4000 already [1, 2], I also did some tests with another use-case in mind. I'm working on a thing similar to RTLSDR-Scanner [3]. I want to monitor a large part of the spectrum continuously. So I compared the R820T with the E4000 using RTLSDR-Scanner w/ and w/o an antenna. My results are here:
https://docs.google.com/folder/d/0ByDAKwyEiyx_XzZ5ZnpRV1VZWDQ/edit?usp=shari... There's much more spurs with the E4000 than w/ the R820T. According to [1, 2] one also would expect a better overall sensivity compared to the E4000. However, the GSM900 signals for example seem to be way better with the E4000 according to the RTLSDR-Scanner. Tuning to a certain channel w/ OsmoSDR Source in GNUradio gives about the same SNR - contrary to the RTLSDR-Scanner output. Can anyone explain this? Also, the DVB-T channel at 502MHz is quite weak in the R820T RTLSDR-Scanner output when compared to the E4000. I had a closer look at the lower limit of the channel in gnuradio. This can be seen in the 502MHz_*.png pictures. The E4000 produces a nice +20dB step while one can hardly see the channel in the R820T spectrum. I don't understand this as well. Is this AGC-related? Manually setting a fixed gain didn't really help though...
Any explanations?
Thank you!
Best regards,
Hunz
[1] http://steve-m.de/projects/rtl-sdr/tuner_comparison/ [2] http://www.hamradioscience.com/rtl2832u-r820t-vs-rtl2832u-e4000/#more-1852 [3] https://github.com/EarToEarOak/RTLSDR-Scanner
2013/3/15 Sylvain AZARIAN sylvain.azarian@gmail.com:
For me the most interesting plots were the "no antenna" ones, showing the LO leakage ;-) If I understood clearly your plots, you have at some places high power ghosts. Maybe you are close to powerful transmitters, but this is more probably a LO to RF isolation problem: the receiver receives himself...
I did two new measurements with the gain set to 20 in RTLSDR-Scanner. The files with the -gain20 in the filename [1] show the new results. There are more spikes from the R820T LOs now, but it's still less than with the E4000 I'd say. The good thing is that the DVB-T stations as well as some other signals can be found as expected now.
2013/3/16 Al al@eartoearoak.com:
I think you've run into a couple of issues, firstly it appears that the AGC isn't fully disabled on the R820T (if you remove the aerial the noise floor increases).
It looks like gain=0 just was too little for the R820T. gain=20 seems to be a good start. Maybe exposing the gain-value in the regular UI would make sense?
Secondly RTLSDR-Scanner averages 2 chunks of bandwidth either side of the DC point, these seem relatively quiet with an E4000 and FC0012 but I haven't had chance to check the R820T yet. This tuner may have a very different noise floor. I was wondering about adding a feature to allow the user to pick their own segments (by terminating the aerial and looking at the noise distribution).
Ah, that's interesting. I was indeed wondering how you avoid the DC spike and my pathon skills are close to zero ;-) According to [2], the R820T doesn't use a Zero-IF, so there should be no DC-spike at all.
Has anyone had a chance to test 2 different dongles with the same tuner? I'd be interested to know if it makes a difference to the noise floor, if there is little change, I could vary the scan based on the tuner.
I do have two E4000 and two R820T dongles, but I haven't yet compared the noise floor. I'm thinking about a per-dongle baseline file w/o an antenna for compensation. This approach should be more robust. Otherwise you need to shift the baseline spectrum according to the frequency error.
Best regards,
Hunz
[1] https://docs.google.com/folder/d/0ByDAKwyEiyx_XzZ5ZnpRV1VZWDQ/edit?usp=shari... [2] http://lists.osmocom.org/pipermail/osmocom-sdr/2012-September/000253.html
Hi All,
When measuring the NF of dongles using this chip with traditional methods one gets very discouraging results. I need to turn up the noise source for NF to 17 dB before I see the desired 3dB increase in received noise power.
Fortunately this is an artifact The NF is low, but there is some kind of noise controlled AGC somewhere.
If I send one or two carriers into the dongle everything seems normal, but if I send a single carrier plus white noise into it, the intelligent "something" applies automastic gain control to keep the noise floor at a nearly unchanged level. The signal which is at a fixed level seems to decrease while the noise seems unchanged. This is not what we normally want in SDR.
When looking at the code I just get confused. It is not obvious to me what gain control is involved. Presumably the RTL2832 chip has some clever algorithm for evaluating the noise floor and the output from that seems to be used for the R820T. I think that for SDR it would be better to disable all those things and allow the user to control gain manually.
One of the good usages of RTL-SDR is to monitor the wideband noise floor of microwave systems. Measuring performance by measuring sun noise for example. That is not possible with the current implementation of the R820T.
Is there anyone who could give a hint how to disable this feature?
Regards
Leif / SM5BSZ
not sure if anyone is interested: I had this IEEE Spectrum article on the RTL receiver lying on my desk for a while, for those who might want to read it http://sequanux.org/jmfriedt/t/ieee_spectrum.pdf
JM
On Mon, Sep 09, 2013 at 11:58:46AM -0400, friedtj wrote:
not sure if anyone is interested: I had this IEEE Spectrum article on the RTL receiver lying on my desk for a while, for those who might want to read it http://sequanux.org/jmfriedt/t/ieee_spectrum.pdf
This article[1] is also available in digital form for free.
Kind regards, -Alex
[1] http://spectrum.ieee.org/geek-life/hands-on/a-40-softwaredefined-radio
-
Alexander Huemer -
Benedikt Heinz -
friedtj -
Leif Asbrink -
Sylvain AZARIAN