AGC on the RFX900 board

GNU Radio
1

Hi,

I have a few questions about the (analog) AGC on the AD8347 mixer chip
on
the Flex900 daughterboard.

  1. Given a signal received with some known power, is it possible to read
    the
    programmed AGC gains (and thereby find the mixer’s noise figure)?
  2. Is it possible to lock these gain settings? If not, can the AGC
    functionality be disabled?
  3. How fast does the AGC track changes in signal power?
  4. If (1), (2) are not possible, is it still possible to know the RX
    noise
    figure?

I’d appreciate some inputs :slight_smile: If I understand correctly, to calibrate
the
receive SNR, we need to know the RX noise floor (seen in software). For
this
we need to know the RX noise figure…

To provide some context, the following is our immediate application: we
have
a coded 16QAM system implemented on GNURadio. We want to measure the
Frame
Error Rate vs. Es/No (“per-symbol SNR”) for this system. The TX and RX
have
separate USRP boards, each with a RFX900 daughterboard. A coaxial cable
connects the TX and RX antenna ports. With this setup, we vary the
transmit
power and measure the frame error rate. To compare these results with a
Matlab model, we need to know the received Es/No: for that we need the
RX
noise figure…and hence my question above.

I’d appreciate your help.

Thanks in advance,
Regards
Sundar

2

On 07/03/2011 08:05 PM, Sundaram Vanka wrote:

  1. How fast does the AGC track changes in signal power?
    system. The TX and RX have separate USRP boards, each with a RFX900
    Sundar

Discuss-gnuradio mailing list
[email protected]
Discuss-gnuradio Info Page

The gain control on the RFX900 is all purely manual, unless you modify
the hardware for AGC.

Figure 19 on the AD8347 datasheet shows noise figure vs gain settings
for the mixer, but on the
RFX900, there’s an LNA in front of it, which at least for higher mixer
gain settings very much
offsets the conversion noise. The very best noise figure on the mixer
is at highest gain, giving roughly
5dB for the mixer. Dividing this by the 13.2dB LNA gain and adding
the noise of the MGA82563 LNA
(2.2dB) gives a best-case noise figure of something on the order of
3dB. In reality, you’re likely
looking at closer to 3.8dB noise figure, due to connector and trace
losses in front of the LNA, etc.

The gain setting in UHD/Gnu Radio should correspond fairly-precisely to
the gain setting on the
RFX900 mixer, which provides something like 69dB of total gain range.
In most cases in RF systems
like this, gain-control is implemented using variable attenuators in
front of the gain element, so
effective noise figure becomes worse as you lower the gain. I believe
that UHD attempts to optimize
its gain distribution for noise figure, but you can also manually set
the individual gains that
are available. On the USRP, there’s a PGA in front of/part-of the RX
ADC, in addition to the gain control
on the AD8347 mixer chip, and UHD will distribute gain settings over
these elements.

3

Thanks, Marcus, for your prompt and detailed response. If I understand
you
correctly:

  1. The best case noise figure of the RFX900 daughterboard is about 3-3.8
    dB.
    This occurs when the mixer amplifiers are set to their highest gain
    setting.
  2. These gain settings can be fixed via software (via UHD).
  3. UHD distributes the single programmed gain value over the mixer and
    the
    PGA to optimize for the noise figure.

I looked up the UHD app. notes. For RFX boards (as you probably already
know
:)) there is a single gain setting: PGA0. I found this at:
http://www.ettus.com/uhd_docs/manual/html/dboards.html#rfx-series

So if I want the best case noise figure, would it be sufficient to set
PGA0
to 70 dB?

As a side note, we found that for a given USRP motherboard (and hardware
settings), the received power as seen in software changes from one
RFX900
board to another by as much as 10 dB. Is this expected?

Thanks again,

Regards
Sundar