[RTTY] rtty

[Kok Chen]

On Oct 4, 2008, at 10/4    4:47 AM, Jay wrote:

> Its the usage of 3 or 2.4 Khz SSB filters that cause nearly 100% of  
> the
> problems between RTTY operations and PSK31 operations.

Hi Jay,

My humble 2 cents (my understanding at this point in time):

This is not a universal problem.  Some rigs have 90 dB to 100 dB of  
dynamic range all the way to where you tap a sound card into.  If  
your sound card also has that kind of dynamic range, you will have no  
trouble running 2.4 kHz passbands.

Good 16 bit sound cards have 96 dB of dynamic range and the best 24  
bit codec that you can buy practically (the particular Asahi one that  
is used in the FlexRadio 5000, for example) has 123 dB of dynamic range.

If your modem software had enough dynamic range (any software doing  
all of its arithmetic in floating point will easily outperform the  
dynamic range of the rig), then your software filters will also  
handle the dynamic range -- i.e., if you don't clip your sound card,  
and the sound card noise floor matches the noise floor of your rig,  
the filters in your software should be able to perform the narrow  
filtering for you.

Don't be so quick to condemn people like me who use wide bandwidths.   
If you set up noise floors nicely and use sound cards with adequate  
dynamic range, the limitation is at the receiver.   You don't hear me  
complaining about it and I have been using 2.4 kHz passbands for many  
years now, and with the K3 plus LP-PAN combo, I plan to use a 100 kHz  
passband once I get some software written for my non-Windows computer.

The FlexRadio, like many SDRs, taps the "sound card" right off the  
mixer without any high gain I.F. stages that is often the source of  
dynamic range woes of superhet receivers (even crystal roofing  
filters can reduce the dynamic range of a superhet)   You can think  
of these SDRs as direct conversion rigs, and you achieve "single  
signal" by phasing the unwanted sideband of a direct conversion rig  
in software by taking the signal from a quadrature mixer.

Absence of other higher order effect, the blocking dynamic range of a  
receiving system determines the largest and smallest signal that can  
be copied simultaneously.  But there is also the IM dynamic range of  
a receiver which determines how much two strong signals can combine  
to create a phantom signal that wipes out a third weak signal.

While the IM dynamic range of the best SDRs and the best supehets are  
pretty much equal, the problem with SDRs (and any one else using  
sound cards with digital modes, for that matter) is that, in addition  
to the IM dynamic range of a sound card, the blocking dynamic range  
of the receiving system is directly tied to the dynamic range of the  
sound card.  Some superhet rigs have blocking dynamic range of 140  
dBm or better, while the blocking dynamic range of the Flex 5000A for  
example, is limited by the 123 dB dynamic range of its sound card  
(although this can be improved if you are willing to use a narrower  
passband than 192 kHz).

The former (BDR) is probably important if you have a neighboring  
ham.  The latter (IMDR) is important in a contest or pileup situation  
where there are lots of S9+40 dB signals in the passband, but no one  
at S9+80 dB.

Until better codecs become available (and inexpensive enough), good  
quadrature mixer based SDRs will have a limited BDR (in the range of  
115 to 125 dB), with a very competitive IM dynamic range running in  
the upper 90s to mid 100s dB.

One of the nice things about SDRs is that the IM dynamic range is a  
constant whether you use a 250 Hz passband or a 192 kHz passband.

The K3 has one of the best BDR among amateur superhets today.  The K3  
has a BDR upwards of 140 dB and a IM dynamic range just at about 100 dB.

However, by applying decimation (digital filtering and then  
downsampling), you can achieve better blocking dynamic range from an  
SDR if you ware willing to use a narrower passband.

The decimation technique is in fact used by the so-called "direct  
SDRs" to get enough BDR.  The codecs in the "direct SDR" work at 70  
MHz or higher and they are sampling the signals directly without an  
analog mixer.  However, they typically start with only 80 dB of  
dynamic range.  But by the time they decimate the RF down to a 2.4  
KHz passband, the BDR will improve to better than 100 dB.  Increasing  
BDR this way does not usually improve IMDR, though.

During the next RTTY contest, take a note of the loudest signal which  
you can find.  Also, take a note of the weakest signal that you can  
copy with a narrow I.F. filter.  It wouldn't surprise me if most  
people don't see a range of signals with more than 100 dB (the  
difference between S1 to S9+52 dB).  If so, you should be able to use  
a 100 kHz waterfall in a good SDR.

> Since PSK31 only
> requires a very narrow bandwidth as does RTTY then it makes PSK31  
> ops angry
> when the RTTY operator (contesting and using 250 Hz fitlers doesn't  
> hear
> nor is within say 500 Hz of the PSK31 Operator that all of hate and
> disconent messages get sent and answered.

The AGC of the rig and the dynamic range of the waterfall display are  
usually the culprits here.  A large signal often makes the waterfall  
go "clean" and you can no longer see a weak signal (i.e., although  
the signal is available to be decoded, you can see it visually in the  
waterfall when a strong signal comes up.  However, you will notice  
that unless the large signal has clipped the sound card, good  
demodulators will continue to copy a signal which you are already  
tuned to.  cocoaModem has a waterfall dynamic range selector for such  
occasions.  Using false colors in a waterfall can also extend the  
visual range, but they tend to produce a "busy" display -- hence I  
ended up allowing the op to change the dynamic range of the waterfall.

A wideband system does work (otherwise none of the Flex-5000 ops will  
ever be able to get on the air in a contest)  Neal K3NC and Bill W9OL  
are two RTTY contesters I know who use Flex rigs; you can ask them if  
large signals bother them in a contest.  My guess is that they can  
copy S1 signals in the midst of a few S9+40 to S9+50 signals.

> What SNR is needed with PSK31 using 250 Hz filters?

Lets see... for white noise, a 250 Hz noise bandwidth is about 10.8  
dB down from a reference 3 kHz noise bandwidth.  So a signal that  
appears at 0 dB SNR with a 3000 Hz passband will appear as +10.8 dB  
SNR with a 250 Hz passband (assuming that your signal is narrower  
than 250 Hz).

(When noise power is evenly distributed, a 3000 Hz passband will have  
12 times more power than a 250 Hz chunk that is taken from it.)

Listening with a 250 Hz filter, a PSK31 has to be quite loud (+11 dB  
SNR) relative to the noise before it can print.

Typical HF Channel Simulators can be set up for 1.24 kHz, 3 kHz and 4  
kHz noise bandwidths.  Specifying modems using a 3 kHz bandwidth is  
common in the amateur practice since the primary Windows based HF  
Channel simulator (PathSim) uses a 3 kHz noise bandwidth.  My  
cocoaPath program for the Mac allows the three noise bandwidths  
above, but I still mostly use 3 kHz in my experiments just so that I  
can compare the numbers from my experiments with other amateur articles.

http://homepage.mac.com/chen/w7ay/cocoaPath/Contents/users.html

A real world "250 Hz filter" could have a noise bandwidth anywhere  
between 200 Hz and 300 Hz (or worse :-).  You'll need to sweep a  
particular filter to really know.  However, as a rough guess, you can  
say that with a 250 Hz filter, a SNR of 11 dB  is needed to get "good  
print" with BPSK31.

The error rate curve can be very sharp -- in the theoretical Additive  
White Gaussian Noise (AWGN) case that you find in textbooks (i.e., no  
Doppler spreading or Doppler shifting), both PSK31 and RTTY can go  
between pristine print to complete garbage when you change the SNR by  
just 6 dB.

As you introduce slow QSB, you need to maintain this SNR even when  
the QSB has dropped the strength of your signal -- hence the need to  
use more power as there is more QSB.  As you further introduce  
further deterioration (such as multipath, or Doppler spreading)  
however, PSK31 drops out completely while RTTY will work if you have  
enough power to handle the depths of the QSB and flutter.

Under "CCIR Good Conditions" (also known as the "Mid-latitude Quiet  
Conditions" in the ITU 1487 standard), the range of SNR between  
pristine print and mostly garbled print for PSK31 increases to around  
15 dB.

Propagation will eventually kill RTTY too -- for example when flutter  
starts spreading a data bit.  If you look on a scope, you can often  
see the Space signal overlapping well into the Mark signal of a  
character as propagation gets worse.  The ST-8000 has a circuit (just  
a couple of gates) that makes sure mark and space don't overlap.   
Good software can actually "equalize" the mark-space delay and they  
handle the overlap better than the good old workhorse.

73
Chen, W7AY