RTL SDR Dongles: Anomalous Frequency Domain Response

I have been experimenting with a cheap-and-cheerful SDR receiver containing an R802T2 front end and RTL2832U demodulator. My ultimate aim is to make a low-cost network and spectrum analyser, but before that it is necessary to understand the receiver’s strengths and limitations.

In an ideal world the receiver’s power-vs-frequency response would be flat, but inevitably that won’t be the case with such a low-cost device: it will need calibration. That calibration has revealed some strange quirks in the R802T2 RF front end frequency response, discussed below. While the anomalies don’t kill the idea of using the SDR dongle as a spectrum/network analyser, I would feel happier if their cause (and preferably cure) was understood.

The signal source is a wideband white noise source “BG7TBL 2014-08-20” which, by definition, contains equal power at all frequencies. The physics of such wideband noise sources is simple and predictable: use a zener diode and amplify the AC component of its voltage. The power is measured using Kyle Keen’s program “rtl_power” tool, and post-processed in a way derived from the associated “flatten.py” script. The rtl_power arguments specify a range from 30MHz to 1500MHz with 1MHz resolution, integrating for 60s, a fixed gain of 0dB and crop 50%.

There are four components of interest, all of which are robust: they are not related to signal strength, nor to cable length, nor to sweep speed, nor to FFT bucket size, nor to cropping:

  • a fairly flat on average response up to 300MHz
  • a progressively reduced sensitivity from 300MHz to 1.5GHz
  • a “wiggle” between 1050MHz and 1150MHz
  • very sharp transitions in the region below 300MHz, e.g. a +5dB jump at 246.3MHz

The progressively reduced sensitivity is presumably an artefact of the R820T2’s internal operation, or of a rolloff in the noise source’s power, or both. I would need a spectrum analyser to investigate those possibilities.

The 1.1GHz “wiggle” is at least superficially similar to anti-resonance effects associated with lead inductance and input capacitance. I initially suspected that it might be related to BAV99 ESD protection diodes, but the above response is without those diodes. Alternatively the wiggle might be an artefact of the noise generator, but I would need a spectrum analyser to investigate that.

I initially suspected the sharp transitions at, for example, 246.3MHz might be a software error in rtl_power, but they now appear to be an artefact associated with the R802T2’s  RF filter and mixer. The transitions occur between successive FFT bins even when the bins are only 1kHz wide. The R802T2’s mixer output power detector (pdet2, pin 5) shows a sudden voltage transition at that frequency – and at the 1.1GHz wiggles.

The SDR dongle is a “Newsky RTL2832U + R820T2 RTL-SDR”, and the noise source a “BG7TBL 2014-08-20”, both purchased on EBay.

Update: another dongle’s measurements show interesting differences and similarities. There are the same abrupt transitions below 300MHz, but I measure larger amplitude variations. The 15dB rolloff to 1.5GHz is similar, but I don’t see the knee at about 880MHz. The absence of the 1.1GHz wiggle makes me wonder if it is an artefact of my noise source – I have no way of telling.

Update: checking the noise generator’s output with a Tektronix 492AP spectrum analyser shows:

  • the 1.1GHz wiggle is indeed an artefact of the noise source. It is significantly more pronounced when touching any part of the circuit between the noise diode and the input to the first amplifier
  • the noise source power output drops 4dB between 300MHz and 1.5GHz, whereas the SDR dongle indicates a 14dB drop. Hence the SDR dongle’s response is 10dB down at 1.5GHz relative to 300MHz


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s