About 'internal AGC'

Leif Asbrink leif at sm5bsz.com
Sun Aug 11 20:48:51 UTC 2013

Hi Charlie,

The controls for AGC do not affect the peculiar wideband LNA AGC
in the R820T chip. The RTL1090 uses the same rtlsdr.dll as is
used by SDR#. I do not know how to evaluate noise levels and signal
levels in RTL1090, but with SDR# one can see funny things
directly on the waterfall.

The LNA is followed by a VERY sensitive power detector that
is somehow followed by filters and amplifiers. It is arranged
in a way to not react on narrowband signals, but already a 3 dB
increase of the noise floor causes a loss of gain through AGC

Most striking is this experiment:

Connect a combiner to the input of the dongle and use it to 
combine a signal generator and a noise source. The noise
source needs a filter that assures that it does not add any noise
on the frequency of the desired signal. It is OK to use a 
T-connector if you do not have a wideband combiner.

I used a 100 MHz low pass filter connected to a vacuum diode
noise source cabable of delivering 17 dB excess noise combined
with a signal generator on 144 MHz by use of a T connector.

I used SDR# to look at the spectrum around 144 MHz.

Without "RTL AGC" and without "Tuner AGC" the noise floor does not
change when the noise source is switched on or off. That is expected
because the noise source can not send any noise through the 100 MHz
low pass filter. That is true at modest gain settings, but if
the gain is set at maximum (49.6 dB) the noise floor increases by 3 dB 
when the noise source is turned on. A small but unexpected effect.

The signal however is attenuated by 23 dB for a total loss at max
gain of S/N of 26 dB!!!!! Please note that the true S/N is not affected 
at all. There is no noise added at 144 MHz.

If I switch on "RTL AGC" or "Tuner AGC" or both, S/N still changes 
the same way. Depending on the signal level of the 144 MHz signal
one can see the signal go down or the noise go up. Or both.

The way sensitivity is lost due to out-of-band noise is invisible 
to the user. There is no warning about overload.

The noise power from 0 to 100 MHz is -174+17+80 = -77 dBm
( -174 dBm/Hz = room temp)
( +17 dB is excess noise)
( +80dB is 100/MHz/1Hz)
There is some filter loss and the dongle presumably has a high pass
filter so one can assume that the noise power is -80 dBm RMS.

I have tried to activate the LNA AGC by use of narrowband 
signals in the 50 to 100 MHz range. Even two signals at
-30 dBm each do not have any effect regardless of the frequency
spacing. It seems the "intelligent" power detector of the LNA
AGC can reject narrowband signals even if they are much
stronger than the noise floor.

The wideband LNA AGC in the R820T may cause problems when
an up-converter is used in front of the dongle. The noise floor
of the up-converted HF spectrum may cause unexpected loss of
sensitivity in the upper part of the HF spectrum where the
noise floor is low.

Adding a filter for the desired HF band. With some gain to
ensure that the noise floor is higher in the desired frequency 
range than elsewhere could perhaps make the R820T dongles
behave much better.


Leif  /  SM5BSZ

> I use a R820T chipped dongle for receiving aircraft transponder signals
> through a specialised application called RTL1090 from jetvision, the control
> panel of which allows the tuner and device AGCs to be independently toggled
> on or off. I have not however used this with SDR-Radio in the past so cannot
> comment on the optimum settings.
> 73
> Charlie
> www.G4EST.me.uk
> From: sdr-radio-com at yahoogroups.com [mailto:sdr-radio-com at yahoogroups.com]
> On Behalf Of Leif Asbrink
> Sent: 10 August 2013 22:30
> To: sdr-radio-com at yahoogroups.com
> Subject: [sdr-radio-com] Re: About 'internal AGC'
> Hello Patrick and All,
> The R820T chip has an advanced AGC function that I do not 
> think can be disabled. The chip detects the noise floor 
> in a wide bandwidth and adjusts the gain to keep the noise floor
> constant.
> The wideband AGC has surprising effects. If one tries to measure
> the noise figure with a noise source that is manually switched 
> on and off one finds a really bad NF. That result is false however,
> if one measures S/N of a weak signal one finds the true NF which 
> is quite good.
> To verify the finding one can inject a weak signal together
> with the signal from a noise source. What happens when the
> noise is turned on is that the signal becomes weaker while
> the noise floor does not change.
> A 500 kHz wide filter in front of the R820T chip converts the
> noise from the noise source to a narrowband signal which will
> not affect the wideband AGC.
> I made some effort to switch this feature off but failed.
> The behaviour is probably quite clever for reception of digital TV
> but I find it very disturbing in a general purpose SDR. I did not
> take notes and I did not investigate in detail what types of signals
> will affect the AGC and what types will not. That would be
> a big investigation and I see no reason to do it because there
> are other chips.
> The R820T gives good signals many times but I do not like the feeling
> of not knowing what I am doing.....
> The "internal AGC" option is another thing as far as I understand.
> The chip has RF AGC as well as IF AGC.
> 73
> Leif / SM5BSZ
> > Hello Group,
> > 
> > When using a DVB-T type dongle (mine is a R820T), do you tick the
> "internal AGC" option or not ?
> > 
> > I tried both "internal AGC" desactivated (with more gain) and AGC
> activated (less gain to avoid spurs from my local FM TX) ... I can't tell
> which one is better. Even on weakish sigs it's about the same.
> > 
> > What about your experience ?
> > 
> > Regards,
> > Patrick
> > 

More information about the osmocom-sdr mailing list