8 May 2013

WSPR - how long to get a decode?

A question for my esteemed readers please.

Using WSPR (WSPR2 that is) the transmission burst is nearly 2 minutes long, but I believe strong signals can be decoded when only a part of the burst is received. My question is, how short a burst can be decoded?  Presumably it depends on how strong the signal is.

LB9YE was coming through quite well this morning but there was no decode of a very strong (I assume MS) burst at 1048 which lasted about 30 seconds. Of course with MS there may also be some Doppler shift which can mess things up a bit.

If you know about WSPR and can help with the answer, I'd appreciate it. I am particularly curious to know how WSPR might do with short MS bursts which can be strong but only last a few seconds. It also begs another question: there is now a longer version of WSPR (WSPR15) with better decode S/N levels. I wonder if a shorter version of WSPR has been considered? Then again I think there may be better modes for MS.

5 comments:

g4fre said...

wspr takes 110.6 seconds to transmit fully (see http://physics.princeton.edu/pulsar/K1JT/WSPR_2.0_User.pdf) so it will not deal with short bursts. FSK441 is the mode used for meteor bursts with great success. Conversely FSK441 does an awful job decoding long lasting signals.

Roger G3XBM said...

Thanks G4FRE (sorry don't know your name).

So, are you saying that it will not decode AT ALL with a shorter period of reception? I thought there was some coding redundancy such that bits could be lost, to some degree, without problems as long as the signal quality was high.

Anonymous said...

Indeed, WSPR does require too long to decode to be of much use with MS.

As G4FRE mentioned FSK441 is one of Joe Taylor's creation specifically for MS as is JT6M (specifically for 6M)

Last summer for a while I monitored 6M WSPR here in North America. Very little activity but there was some success. In this respect, WSPR is much better suited to ES rather than MS.

I have a QRSS grabber set up for 10m and have been running it continuously for quite some time. It is interesting to watch signals fade in around then fade out as the conditions permit. Interestingly, when 10 has been active and signals have faded, about an hour later the signals will fade in once for about 30 to 45 minutes.

On my grabber I will see from time to time what I refer to as 'squiggles' short horizontal S shaped signals of about 10 to 15 seconds duration. Indications are that there are being caused by MS but the exact source of the signal being so reflected is anyone's guess.

Apparently there is some MS activity on 10m but so far I have not made any concerted effort to locate and attempt to decode.

QRSS as you know is able to detect extremely weak signals through long term integration but decoding and reporting is entirely manual. the advantages of the likes of WSPR is their ability to record, report, and collate reception reports automagically.

WSPR on 10m ES OK, MS not likely.

The trouble I am finding is that there are too many modes, too many "watering holes" etc, etc, that it is sometime hard to know if you are just not hearing something or your equipment is at fault.

Lots of fun either way; the journey is what is important, not the destination.

cheers, Graham ve3gtc

Sverre Holm said...

A WSPR message consists of 50 bits and is encoded into 162 bits with a convolutional code. I believe that that means that under good conditions 50/162 or about 31% of the message is enough for a decode.

At least this is the case for the RS-coding of JT65 where it is claimed that 10-15 sec of the 48 sec transmission is all that is needed under good conditions. This is not far from my experience also. I believe that can be estimated from 72 informaton bits that are coded into 63 6-bit symbols each, so 72/(63*6) = 19% is the message. 19% of the transmission length of 48 sec is about 9.2 sec.

Sverre Holm said...

I've looked at this again. It is only the Reed-Solomon coding of the JT65 that can manage with a partial code as in the example above.

This doesn't apply, it seems, to convolutional coding, i.e. WSPR and JT9.