I have just been doing some more tests on 30m. Clock errors might well 
have been part of the problem. Initially on 30m the few decodes I was 
seeing were showing a DT of over 3s, so I had a play with the dsec 
setting, and setting that to -3.5 brought me rather more spots with DTs 
around 0. I noticed at that time that my computer clock was indeed 
around 3secs out relative to MSF, but that doesn't tie in with when I 
was listening to Jim earlier when it was no further than 0.5s 
different, even after an internet resynch. Surely my PC clock didn't 
drift 3s in a couple of hours....

As to strong signals failing to decode, thanks for the information. I 
am using the built in sound chip on the motherboard, and as you know 
these are always a bit ropey, so it may be the reason. I have another 
problem at the moment, in that I have just fitted a fixed level audio 
output connection to the K2 and I think I have got its level a bit high 
- I have to set the input level in WSPR to virtually zero to get around 
0dB input. I shall be turning it down a bit next time I have to go 
inside the K2.

It does look as if some aspects of WSPR are a bit finnicky though, not 
too impressed myself but if it means you are spotted in AA1A land it is 
obviously serving a good purpose. But for working Hatfield from 
Bracknell I guess there are quicker ways...

73 Dave G3YMC

On 5 Sep 2009 at 12:17, James Moritz wrote:

> Dear Daves, LF Group,
> 
> Thanks for the overnight reports - the beacon was shut down at 0925utc
> here. I see several spots in the database from AA1A - the first time
> M0BMU has been copied in the US this season, thanks Dave.
> 
> In answer to YMC Dave's questions:
> 1) I don't know of an actual figure for required frequency stability,
> but more than several Hz drift during a 2 minute period would be a
> problem. But a drift of 1Hz in 2min is small enough. Some of the HF
> people have problems with rigs heating and cooling between transmit and
> receive cycles and the resulting cyclic drift. Using a converter and
> receiver with seperate oscillators will certainly tend to multiply drift
> problems.
> 
> 2) I have not so far had much success with internet based clock 
> synchronisation utilities, so at the moment I am just setting my PC
> clock against an MSF clock at the start of each session - this is
> adequate, but my PC does drift a few seconds overnight. Having said
> that, I was being received OK by a number of other stations this
> morning. The transmission is supposed to start on each even minute, and
> have a duration of 110.6 seconds.
> 
> 3) WSPR seems to have good dynamic range, provided the RX and sound card
> gains are adjusted so that overload does not occur. Certainly, it is
> possible to receive rather weak signals giving just a faint trace in the
> noise on the waterfall display at the same time as strong "red" traces
> occur - a range of 40dB or more. Even if a strong signal does cause
> overload, the randomised transmit periods mean that only a fraction of
> the weak signal time slots will be affected. I have not noticed any
> problem with decoding strong WSPR signals here, although there are not
> really any "very" strong signals.
> 
> The "false decodes" occur from time to time, apparently more when 
> narrow-band QRM is present. If you read W1JT's description of the
> various JT modes, he has made a trade-off between excluding decoded weak
> signals that contain a lot of noise but could be correct, and falsely
> decoding noise that by chance resembles a signal. In practice, they are
> not really an issue, due to the fact that the callsign is encoded with
> other information (for WSPR the locator and power output) and these
> would have to be consistent with the callsign for the signal to be
> accepted as valid. Even if the falsely decoded callsign is plausible, it
> is extremely unlikely that the locator would be correct, and
> astronomically unlikely that the same decode would be generated from
> noise at multiple receiving stations on multiple occasions. Using these
> types of criteria, false decodes are periodically stripped out of the
> WSPR database.
> 
> A number of people have had problems with strong signals failing to
> decode, on HF as well as 500kHz. We did some investigation of this last
> winter; there may be more than one cause, but one issue is that some
> sound cards/operating systems appear to generate significant timing
> errors, which over the relatively long transmission period of WSPR
> accumulate to mess things up. This results in some receivers failing to
> decode some transmitters. One way to test this is to receive a strong
> WSPR signal, but then after about 40s or a minute from the start of a
> transmission disconnect the audio to the PC or turn the gain down to
> zero. There is enough data in the shortened signal to decode, but less
> time for timing errors to accumulate. If this results in successful
> decoding, then a timing error of this sort seems likely, which could be
> at either TX or RX. Weak signals may decode fine, because parts of the
> signal are "missing" due to noise and fading. So another possible test
> is to reduce gain until the strong signal is just a weak trace on the
> waterfall in the receiver noise, and see if correct decoding occurs
> then.
> 
> Cheers, Jim Moritz
> 73 de M0BMU
> 
> 


http://www.davesergeant.com