Dynamic Noise Reduction is a nearly unachievable Holy Grail that means
different things to different users and designers. It can mean reduction
of circuit noise, natural atmospheric noise like lightning static, or
man made electrical noise like power line noise and ignition/injector
pulse noise. It can mean reduction of splatter from over driven
transmitters in the neighborhood (which may be a couple states away or
more) or the splatter from an over driven receiver front end and that
also may include transmitters a state or two away if the receiver
doesn't have enough high end dynamic range. Each of these noise sources
take a different technique and many leave behind a background that is
very unlike white noise that we have learned to discriminate against by ear.
Repetitive impulse noise like ignition noise has been handled by
hardware and DSP for decades. But with side effects. Like chopping
strong adjacent channel signals and restoring the noise over the weak
more desired signal. The impulses from injector drivers tend to be
longer than ignition noise so the hardware or logic that looks for short
peaks may not detect them without adjustment. The same thing roughly
applies to power line or neon sign noise and there have been hardware
noise blankers that handled all three with the known side effects. To
discriminate on pulse width requires the receiver first selectivity be
broad to get the pulses short and easily discriminated against. Once the
bandwidth is narrowed to the 2 or 3 kHz we plan to listen too, its
really hard to discriminate based on pulse width.
Circuit noise at HF is easily reduced by circuit design usually trading
off strong signal handling capability but except for very low efficiency
receiving antennas like a Beverage antenna noise is stronger than
circuit noise. At VHF, UHF, and microwave one often doesn't hear any
atmospheric noise (unless there are storms within a couple hundred
miles) so lowered circuit noise is a great benefit. To the point that
today, the earth surface puts out more noise than the sky or the preamp
at its input, so you can check the receiver performance by comparing
quiet sky to the warm earth surface or moon surface.
Reducing atmospheric noise is the tough one because its not nearly so
well defined or repetitive as power line or ignition noise. Most schemes
for noise reduction are based on averaging the signal and noise and
expecting the randomness of the noise to cancel out over several
averages. Or to look at the time or the spectrum of signal plus noise
for the coherent signals and to extract what is coherent from those that
constantly change. This is easiest with signals having constant
amplitude and frequency or predictable changes, like PSK-31. This is
tough with voice signals because we speakers aren't quite so coherent,
with sibilants, clicks, wind noise over the microphone, and coughs
definitely strong signals but not very coherent. Yet sibilants are
critical to easy understanding of voice. These coherent correlation
filters take time and we don't like time delay in our receivers, it
makes tuning them like tuning with a loose rubber band for the shaft
coupling. We stop tuning but the receiver keeps on moving for the time
delay of the digital signal processing. The what's left over isn't
nearly so random so we face hearing voices or CW signals against a
background that isn't white noise, it has tones and qualities that are
"not natural."
In urban areas of the world, much atmospheric noise comes from switches
being opened or closed, and hash from digital data devices, whether
local or distant, whether intentional radiators or not.
Some coherent vs none coherent processes seem to work on actively
adjusting a collection of bandpass filters to pass the parts of the
spectrum detected as coherent (e.g. still there from one Fast Fourier
Transform to the next). That's easy to do with the transformed signal.
Then a quick inverse transform makes it audio again. With delay, more
delay while comparing for coherence.
No doubt there are schemes made of combinations of these and more
elaborate but the fundamental is discrimination against random vs
coherent, and the narrower the hardware filter the less random what we
perceive as the noise. The key is identifying the wanted vs the wanted
which can be difficult.
Different designs work differently because many of these processes in
hardware and DSP have been patented and so to use them a maker has to
find a way around the patent to get the same or better results or pay
for the privilege of using the patented technique. So different
manufacturers often use different techniques with different results just
to avoid paying patent royalties.
There are other techniques for weak signal signaling that involve very
narrow signals with slow data rates like WSPR that takes 2 minutes to
send a call, a signal report, and a location, or WSJT for EME that uses
slow data rates, many repeats, and much extra data for error correction
at the receiver without having to wait for a repeat. But they aren't
copied by ear or generated by mouth or morse key.
73, Jerry, K0CQ
On 10/31/2010 3:23 PM, Ray Sills wrote:
Of course, that's the comment by -one- reviewer. Sometimes these
issues are really opinions.. one person may like, another may not.
Differences between the performance level of various radios will
certainly be related to the cost of the radio. I'm sure they are all
built to a price point. To add more capability, or performance, you
usually have to spend more. And you then get to the usual trade-
offs: fewer high performance radios vs. more lesser performing units,
but more market share.
The IC7700 costs 3 times the amount of an Eagle. Is the DNR 3 times
better? I doubt it. There is often a point of diminishing returns
for performance. Engineers make things "good enough"... for the price.
73 de Ray
K2ULR
On Oct 31, 2010, at 3:47 PM, John Rippey wrote:
I notice a comment by the first Eagle reviewer on eHam.net that the
DNR
feature was unimpressive.
Similar comments have been made about the DNR feature in the
Kenwood TS-590,
including it continues the (to me awful) SPAC
setup of the TS-570.
On the other hand, Icom's 7700 and 7600 seem to provide the best DNR
iteration so far, based on the comments I've seen. Yaesu, as usual,
seems to
lag behind in the effectiveness of its various DNR iterations. So,
why does
it seem so difficult for manufacturers (other than Icom) to get DNR
right in
a DSP environmnent? Is it the cost of the hardware, firmware, software
engineers, or what?
73,
John, W3ULS
_______________________________________________
|