On Jan 30, 2009, at 9:15 PM, Rick Ruhl wrote:
> I just use TNC to TNC or TNC to soundcard to do my testing. You can
> set the
> level of the KAM XL or KAM 98 to emulate weak signal RTTY using the
> XMITLVL
> command.
By doing that, you will only be testing the AWGN case in VE3NEA's
examples
http://www.dxatlas.com/RttyCompare/
Well, it won't actually be exactly the AWGN case if the input noise of
the modem under test does not have Gaussian statistics (i.e., if the
noise is predominantly flicker noise or something else).
If you want realistic testing, you will need an HF Channel Simulator
(like AE4JY's PathSim) in between the two modems. That will add
Doppler spreading and Doppler shifting which cause QSB and flutter
effects in Rayleigh channels (what we roughly know of as "HF
Propagation"). Most Channel simulators are also capable of adding
noise that is very close to Gaussian; see figures 4-2 and 4-3 here,
for example
http://homepage.mac.com/chen/w7ay/cocoaPath/Contents/technical.html
In general, what determines a modem's performance is not whether the
signal is weak but instead, the performance depends on the Signal-to-
Noise Ratio, so long as the signal is 10 dB or so above the A/D
converter noise floor for digital modems -- i.e., when the predominant
noise is no longer quantization noise from the A/D converter.
(Many people are amazed to copy weak RTTY signals when they should not
be. If the noise level is also low, weak signals are easy to copy as
long as you adjust the sound card correctly. I.e., you always want
the noise floor of the sound card to be lower than the noise floor
coming from the receiver -- as long as nothing is clipping the sound
card. Very often, you cannot satisfy both (noise floor and clipping)
conditions, and that is when we say that the receiving system
(receiver + sound card + modem) does not have enough dynamic range.
RTTY copy is dependent on SNR, not on how strong or weak a signal is
on the S meter. Again, look at the curves at the VE3NEA site.)
For comparing between modulation modes, textbooks actually use Eb/No
-- ratio of energy per bit to spectral noise density -- instead of
SNR, since Eb/No also takes into account the baud rate of the signal
and the effective bandwidth of the signal. So it is a fairer test
between modulation modes. For testing different modems that use the
same mode (e.g. FSK), it is perfectly fine to use SNR as the
criterion, as Alex VE3NEA has done, since energy per bit and
bandwidths are fixed. But you will often not see plots that are based
on SNR, but instead based on Eb/No. They are equally useful -- there
is a simple relationship between Eb/No and SNR, as explained here:
http://www.sss-mag.com/ebn0.html
By the way, the figure in this last web page shows why PSK can get
through with lower power than FSK. But note that it is for AWGN
(notice that the shape of the curves are similar to the shape of the
AWGN curve at VE3NEA's web site).
Once you add enough multipath however, the PSK curve "blows up" and
PSK no longer prints even when you use enormous amount of power, while
FSK will produce printable output if you raise the power enough to
have a good enough SNR. Again, if you refer to VE3NEA's plot, a quite
band with no QSB (AWGN) needs a SNR of -8 dB (in a 3 kHz passband) to
print well. Whereas, when you have flutter (go to his flutter graph)
you need +7 dB SNR to print well.
VE3NEA uses 3 kHz bandwidth because that appears to be the standard
among ham HF Channel simulators (the ITU specs appear to prefer the
use of 4 kHz and 1.24 kHz). If the signal is narrower than 3 kHz, it
is easy to derive the SNR with a narrower filer. If you are using a
300 Hz filter instead of 3 kHz, you would just add 10 dB to the SNR
scale in Alex' graphs. If you use 600 Hz filters, add 7 dB to Alex's
SNR, etc.
73
Chen, W7AY
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|