Message	Date	From	Subject
757	2007-05-11 17:23:49	Kevin Purcell	ITU-R P.372-8 expected ambient noise level recommendation and "good
To answer the question "what's a sufficient receiver noise figure?" when designing a receiver you need to specify where you are located and which frequency you are interested but where do you get the numbers for the noise figure? The ITU recommendation ITU-R P.372-8 Expected ambient noise level is a simplified model that can help estimate the expected noise for some typical (temperate) locations. The problem is it's difficult to get hold of a copy. VK1OD has extracted the data from the ITU-R P.372-8 Expected ambient noise level recommendation for HF and posted the data and some very useful plots to his web site. He includes both the noise model data in tabular form and some reference lines for S meter reading (assuming a HF S meter calibrated to IARU R.1 with S9 == -73dBm == 50uV in 50 ohms and 6dB S units). <http://www.vk1od.net/fsm/FSAmbientNoise.htm> Or all in one handy PDF (with tables of data). <http://www.vk1od.net/fsm/FSAmbientNoise.pdf> One interesting thing to note on figure 2 is the relative separation of the business, residential and rural curves (about equally separated) and the seperation of "quiet rural" (i.e in the wilderness with no roads/people/electricity). "Quiet rural" is a lot quieter than "rural". This seems to match an observation made by W7WOI and company in EMRFD that I can't find. The received power data from VK1OD is in a 2000Hz bandwidth. It's perhaps more useful if you convert it to an equivalent noise figure (as noise figures are bandwidth independent). That's easy enough MDS = -174 + 10 log(2000) + NF or Atmospheric NF = Received power + 141 So a suburban site might expect to see -99.4dBm in 2khz bandwidth on 20m or an equivalent noise figure of 41.6dB. I think this model supports the point that EMRFD makes on 6.11: > Many argue that noise figure is rarely a significant parameter, > especially for HF reception. A NF of 10 to 12dB at 28MHz, with > higher numbers at lower frequencies will provide as much > sensitivity as one can use. A practical receiver test is very > simple: while listening to background noise on the band disconnect > the antenna. If the noise drops significantly then the receiver is > as sensetive as it needs to be. Which seems to imply a 3dB NF margin or so should be fine. You probably don't want more as that means you are "wasting" the high end of your receiver's dynamic range. It's also a good argument for variable attenuators in front of most of receivers. Of course these numbers vary with time of day (it gets quiter when people tun out the lights), phase of the solar cycle (low sunspot numbers brings lower atmospheric noise from tropical thunderstorms) and so on but they make a nice design average. Another experimental project would be to measure the ambient noise at your favorite sites (see other VK1OD pages for some suggestions for calibrating your antenna factor to make absolute measurements and making measurements without instrumentation receivers). The other curious thing is the quiet rural is quieter than the galactic noise curve. Ideas on why that might be so are welcome (I've asked VK1OD). I'm betting on alien abductions in the Boonies. -- Kevin Purcell kevinpurcell@pobox.com -- N7WIM / G8UDP Kevin Purcell kevinpurcell@pobox.com
760	2007-05-11 19:54:49	Kevin Purcell	Re: ITU-R P.372-8 expected ambient noise level recommendation and "
VK1OD clarified an item with me: these are estimates of man-made noise (ambient noise) only. So the galactic noise contribution is shown separately and should be added (in watts not dB) to get the contribution from both sources. > Kevin Purcell wrote: > >> One curiosity I noticed in your "translation" of ITU-R P.372-8 >> <http://www.vk1od.net/fsm/FSAmbientNoise.pdf> >> is the quiet rural (wilderness, I suspect in the US or deep in the >> Outback in Australia) has lower noise figures than the galactic >> curve. >> > > Hi Kevin, > > Perhaps it is my shorthand in labelling the graphs. > > The curves except for galactic are for man made noise. The expected > total noise at a place will be the sum of galactic, man made and > atmospheric. It is rare that galactic noise is not present. Man-made noise is the dominant noise source (for the majority of locations) so this information is still the best for the sort of estimate I outlined. Atmospheric noise is much more complicated and is also modeled in ITU- R P.372-8. Of course as atmospheric noise is mostly generated by lightning and propagates from the lightening strike to your location it depends on the density of lighting strikes and propagation (on a given frequency for a given time on a given point in the sunspot cycle). The best (open) source for complete noise data is Greg Hand, a now retired Fedral employee, and author of ICEPAC, VOACAP, and REC533. He continues to enhance <http://www.greg-hand.com/hf.html> NTIA Report 87-212: An Updated Noise Model for Use in IONCAP > Abstract: This report presents an updated and improved nosie model > designed for use in the HF propagation prediction program, IONCAP. > The model has, however, much more general applicability, since the > frequency range 10 kHz to 30 MHz is covered. The report give the > hisotry, as near as can be determined, of the existing noise > routines, and then develops the updated model based on current > information. The three noise sources – atmospheric, man-made, and > galactic and treated and a more appropriate means of combing these > three sources is developed. Examples of the use of the improved > model in IONCAP are included and comparisons made with the existing > model. <http://www.its.bldrdoc.gov/pub/ntia-rpt/87-212/> and a very recent update to the noise model <http://www.greg-hand.com/noise/> > I have also developed a program that calculates results of > combining radio noise from the three sources (Atmospheric, > Galactic, Man-Made) that affect HF communication. The outputs of > this program are both single location calculations and contours of > any of the noise parameters plotted on a world map. These contour > maps can be used to reproduce the Fam 1 MHz maps found in ITU-R P. > 372-8. including a program to do the calculation using the new model of all noise sources.

To answer the question "what's a sufficient receiver noise figure?"
when designing a receiver you need to specify where you are located
and which frequency you are interested but where do you get the
numbers for the noise figure?

The ITU recommendation ITU-R P.372-8 Expected ambient noise level is
a simplified model that can help estimate the expected noise for some
typical (temperate) locations. The problem is it's difficult to get
hold of a copy.

VK1OD has extracted the data from the ITU-R P.372-8 Expected ambient
noise level recommendation for HF and posted the data and some very
useful plots to his web site. He includes both the noise model data
in tabular form and some reference lines for S meter reading
(assuming a HF S meter calibrated to IARU R.1 with S9 == -73dBm ==
50uV in 50 ohms and 6dB S units).

<http://www.vk1od.net/fsm/FSAmbientNoise.htm>

Or all in one handy PDF (with tables of data).

<http://www.vk1od.net/fsm/FSAmbientNoise.pdf>

One interesting thing to note on figure 2 is the relative separation
of the business, residential and rural curves (about equally
separated) and the seperation of "quiet rural" (i.e in the wilderness
with no roads/people/electricity). "Quiet rural" is a lot quieter
than "rural". This seems to match an observation made by W7WOI and
company in EMRFD that I can't find.

The received power data from VK1OD is in a 2000Hz bandwidth. It's
perhaps more useful if you convert it to an equivalent noise figure
(as noise figures are bandwidth independent). That's easy enough

MDS = -174 + 10 log(2000) + NF

or

Atmospheric NF = Received power + 141

So a suburban site might expect to see -99.4dBm in 2khz bandwidth on
20m or an equivalent noise figure of 41.6dB.

I think this model supports the point that EMRFD makes on 6.11:

> Many argue that noise figure is rarely a significant parameter,
> especially for HF reception. A NF of 10 to 12dB at 28MHz, with
> higher numbers at lower frequencies will provide as much
> sensitivity as one can use. A practical receiver test is very
> simple: while listening to background noise on the band disconnect
> the antenna. If the noise drops significantly then the receiver is
> as sensetive as it needs to be.

Which seems to imply a 3dB NF margin or so should be fine. You
probably don't want more as that means you are "wasting" the high end
of your receiver's dynamic range. It's also a good argument for
variable attenuators in front of most of receivers.

Of course these numbers vary with time of day (it gets quiter when
people tun out the lights), phase of the solar cycle (low sunspot
numbers brings lower atmospheric noise from tropical thunderstorms)
and so on but they make a nice design average.

Another experimental project would be to measure the ambient noise at
your favorite sites (see other VK1OD pages for some suggestions for
calibrating your antenna factor to make absolute measurements and
making measurements without instrumentation receivers).

The other curious thing is the quiet rural is quieter than the
galactic noise curve. Ideas on why that might be so are welcome (I've
asked VK1OD). I'm betting on alien abductions in the Boonies.
--
Kevin Purcell
kevinpurcell@pobox.com

--
N7WIM / G8UDP
Kevin Purcell
kevinpurcell@pobox.com