Chasing Meteor Scatter on FT8, a fruitless business.

My last posting (finalised yesterday) was about the length and peak of the Sporadic E season. You can read it here, though it is just below this one on the blog.

This time I am mentioning a factor which makes 50MHz and 70MHz contacts at this time of the year difficult to classify. 

Difficult to work too.

There are often a lot of single decodes cropping up on my software just now. Some of them do carry on weakly to a second decode and the odd one makes it to a full QSO. Then there are sudden decodes which appear to come from nowhere, sometimes linking into distant propagation modes like TEP which seem like strange mini-openings. But after one strong decode they are gone. What is this thing which is teasing me?

The clue lies with meteor scatter. As I write we are in a meteor shower. When I look at dxMaps I see similar decodes on 144MHz, often over pretty well the same paths, on the 2m band. There are two differences - 1) the 2m ones don't often link into anything because those dx propagation modes are rare on 2m, and 2) on 2m they are listed as meteor scatter rather than Sporadic E.

Ah, I have at last found a reason why I watch 5 or 6 bands simultaneously. To find things like this, of course.

The maps of 6m and 2m propagation look very similar, but on 6m there are postings showing Sporadic E on FT8, whereas on 2m there are decodes as meteor scatter on MSK144. Look at the listings and you see that the 6m ones are just single decodes but the 2m ones are meteor scatter QSOs.

Meteor scatter propagation happens in the E layer when signals are scattered from ion tails left by meteors. At 2m these are fleeting short pings, but the same event produces longer periods of propagation on 6m. Thus it is possible to have good strength contacts on 6m using MSK144 and these show up as meteor scatter contacts.

Ionisation caused by meteors is short lived on higher frequencies but the effects last longer as we use lower frequencies. This longer ionisation period on 50MHz persists and can allow what looks like "standard" Sporadic E because FT8 can decode signals during bursts (at least briefly). This slightly longer period for 6m can permit a decode which is little affected by Doppler shift. So it is possible to have a 6m decode using FT8, and that tends to be posted as Sporadic E. If it happens to last long enough, a QSO can result using modes which normally do not respond well to meteor scatter.

Of course, it is ionisation on the E layer, it is short lived, so it is Sporadic E. But as it is directly caused by meteors it follows the general principals of the meteor scatter mode. Thus is tends to happen principally at the same time as meteor showers. Yesterday was the peak of the Lyrids shower, and these short decodes peaked then with me. It tends to peak at about 06:00 local time and decline during the day, as did this effect yesterday. You can get meteors at any time, but during the morning meteor shower peaks I see this type of propagation peak too.

Sometimes, during an intense shower you can get enough direct ionisation to form a small area in the E layer and get some prolonged propagation, but this is not so common. In a weaker form it can occur at any time too, but mostly I only notice it during shower peaks.

I first found out about this through a posting by another amateur who goes meteor scatter hunting on 28MHz during every meteor shower. On 10m the effect is more long lived than on 6m, and he can have quite a lot of success. I wish I could find that posting again. Anyway, once I knew what to look for it was easy to see. Here are the two maps for yesterday, 22 April at around 08:00:-

dxMaps of 144MHz on 22 April 2024

 

 

 

dxMaps of 50MHz on 22 April 2024 

 

At this time of year, and during meteor showers generally, you can turn on PSKReporter or DXMaps and see lots of 50MHz decodes reported, while there are almost no QSOs possible (try selecting "Only DX-cluster" from the DXMaps options during one of these events to see the difference).

So what difference does it make that this form of Es is created directly but briefly during a shower, when "standard" Es is just the same ionisation which has been organised into a layer (for details of that complex process see here)? The immediate effect is that it is very difficult to have a QSO on FT8 like this directly after a meteor burst. FT8 is not designed to handle it. MSK144 is a far better mode for this purpose. Also, the station at the other end may not be expecting to deal with single decodes in the way that most meteor scatter operators do. MSK operators will have long watchdog periods and keep transmitting for ages. Plus the key difference is that the MSK144 algorithm is far more sensitive to the weak signals as the level of ionisation tails off during the period of the ping.

It is Sporadic E, Jim, but not as we know it (with thanks to Gene Roddenberry, Bones, and the crew of the Starship Enterprise for that misquote).

To make full use of this effect, I suggest you switch to MSK144 and work meteor scatter on 6m - if you can find anyone else with the same idea. Or go down to 28MHz where the ionisation hangs around long enough to have an FT8 QSO (possibly). However, it only lasts for a day or two around a meteor shower peak.

Me, I just sit and watch as operators chase individual decodes. It must be frustrating. It was for me until I worked out why this happened at the meteor shower peaks.

73

Jim GM4FVM

Trying to estimate the length of the Sporadic E (Es) season and calculate the "lag".

I have recently been criticising some others for taking a few bits of evidence related to amateur radio and then building that up into a theory of Sporadic E. In my view this does not have credibility.

I am about to do something similar so this needs some qualification. I have been looking through my log book. This is obviously limited by loads of factors, but it still might be worth a look. There are 13859 entries in there, not the five data points which one of the other theorists used. Not that I have made 13859 contacts since I arrived at this location in 2009. Oh no, there are more than that. I am not legally required to log all my contacts, and a lot of local ones were not logged. That was in the olden days when there were local stations to work.

None of the missed log entries should matter for this purpose, as I have been very careful to log all my Es contacts. However, this is an early warning that we need to be aware of the many possible areas of bias in the data.

OK, I looked at the Es contacts at GM4FVM on 50MHz from 2011 to 2023. I arbitrarily excluded contacts from early December through to February to exclude "Winter Es" or "Christmas Es" or whatever you want to call that. This first measure was for the length of the season between the first contact and the last, not counting how many contacts resulted in any day or week. This method does not identify the peak of the season in terms of contacts, nor log periods with specific openings such as mini-seasonal paths to Japan or links to TEP. However, every Es contact in the log counted for this purpose including those specific ones. The data was aggregated up from activity days into activity weeks to make the maths practical.

I expected that the "lag" means that Es started later in the year and ended later in the year than a strict correspondence to the Midsummer Day (by which I mean 21 June -ish). I had expected to find that the seasons tend to be roughly the same length each year, just all shifted towards the end of the year.

Sporadic E contacts at GM4FVM 2011 to 2023 (1)

As usual, click if you wish to see an enlarged version of the images.

First of all, you can see the lag. It is about 4 weeks in the first period rising to about 5.5 weeks later. Es starts later than the Spring Equinox and now ends after the Autumn Equinox. So the data (as far as it goes) backs up my initial hunch.

Secondly, and unexpectedly, the Es season appears to be getting longer. The data seemed to fall into three distinct periods. That averages fell very close in each year within each four year period, and then there was a sharp step up to the later period each time. There were three distinct periods standing out from the statistics.

I really doubt that the period of enhanced propagation really is getting longer. I reckon it just looks like that. This finding took me back into my log for some explanation of what is happening. During the 2011 to 2014 period I was mostly using SSB ("pre data period"). During 2015 to 2018 I was using mostly PSK, JT65 and some JT9. Although FT8 was appearing not many people were using it (so this was the "PSK+JT period"). During 2019 to 2023 FT8 was in full swing and lots of people were using it ("FT8 period").

I may be wrong here, but I think that what appears to be a longer season is largely down to me switching from SSB via PSK and JT into FT8. During the SSB period I had to search out stations as they were spread throughout the band. During the PSK+JT period it was either the same spread (PSK) or I was using modes with long periods (JT65) or few followers (JT9). The sensitivity would have been higher than with SSB, but those modes did not have wide acceptance. Then during the FT8 period lots of stations appeared on the same frequency and the mode was even more sensitive. This is my theory about this; just my idea.

Working stations on Es on 50MHz has become easier over those years. More people are doing it, and that also contributes to what looks like a longer season. I doubt if the propagation has changed even though it looks like it has. It is easier to work people now because while at the start of my time on 6m stations had to have transverters, and build their own antennas, these days they have ready made equipment. Data modes have become much easier to use, computers have become cheaper and faster. Also we no longer need lossy data interfaces but we can now pass digital audio in our radios via USB plugs. As it gets easier, more stations come on looking for DX, and the season appears to extend as marginal propagation gets exploited.

Now that I had the basic information that it is indeed shifted to later in the year, I started thinking about the "shape" of the season. It feels to me that once the main season starts, usually about 4 weeks after the first contact, it suddenly bursts forth and reaches a quick peak, and then subsides gradually until the end.

So how to find where the peak is? My data measured so far is just contacts spread over time. It does not show when the main burst of activity happens. I could only do this on 50MHz by going over every contact individually - which I might do some day but not right now. There are thousands of them. So what might measure the timing of the peak? It dawned on me that my 2m Es contacts usually only happen at the peak.

I make the usual pre-condition that these are just my own data, and sometimes I am away, not watching 2m, or something similar. I spent a lot of my early years here thinking that 2m Es was rarely possible from this far north. Would there be enough data? Well, just like on 6m, the spread of data modes has provided me with a lot more data than I thought I had - 31 data points. There are bound to be fewer as these probably represent the peak of the Es propagation each year.

I then calculated those days from first to last contact and put them on the same graph:-

Sporadic E contacts at GM4FVM 2011 to 2023 (2)

The 2m contacts were almost symmetrically spaced on either side of midsummer. Unlike the 6m contacts where the lag was at least 4 weeks, the lag at 2m was almost zero (less than 12 hours) and not statistically significant.

This suggests to me that my second hunch that the Es peak comes in a rush towards the start of the season is also right at this station. However, just as on 6m, my figures just show when the 2m contacts happened not how big the opening might be on that day. Such figures on the "bigness" of an opening can be very misleading, as openings to my west are never big because there is just the Atlantic Ocean in that direction. 2m Es openings almost never include multi-hop Es, or not from here anyway, and nor have I noted them linking to TEP (not YET!!!). So the subjective quality of an opening is not something that I have tried to measure. But still, if the 2m contacts do indeed represent the peak of Es propagation then the peak here does not appear to lag behind the midsummer like the rest of the season does.

Now I want to go further into speculation. What might the shape of the Es season look like if I could work out the "bigness" of openings? Well, here is the FVM conjecture. I suggest that it looks like a gamma graph something like this one:-

Gamma distribution.

With apologies to whoever drew this, I cannot give an attribution as I cannot find where it came from via the internet. Let me know if it is yours; meanwhile I will try to find a better one. I am using it as x along the bottom axis looks like months of the Es season, and f(x) could be the "FVM-Es-bigness scale" [now you are getting carried away Jim].

Anyway, if this is roughly the shape of the season it would fit with the pre-data mode earlier years simply sampling the top part of the graph, say above f(x) = 0.3, and then as we got more sensitive modes and more activity we are sampling further down the vertical axis and finding the tail on the right. Thus the lag seems to be getting later and the season longer, which is what I found. 

Warning: This is just one guy's log history. It is not very reliable. I could do more and better statistical analysis but that would expose the unreliability of using any amateur's data. So it is not scientific, and no ham survey ever will be. Despite this I think it does provide something to think about.

When I was at college there was a lot of enthusiasm for using "qualitative" data. That is not gold standard statistically but it does show trends. What I have tried to do here is show what I have found in a qualitative way. It is up to you whether you believe it or not. For me, however, it has "face validity". It seems to explain what I have been seeing over the years.

First thing I can see is that the lag on 6m pretty well rules out Es linking into TEP in the Spring from here, but would suggest that is more likely in the Autumn. A quick check suggests in my case that Es linking is at least 9 times more likely to happen in Autumn. Secondly, I had better not rely on the lag to look for 2m Es contacts later in the year, as they seem to be equally spaced around midsummer. I will miss them if I wait. And thirdly, whatever the reason, it is worth hanging on looking for 6m Es right into the tail of the season, certainly in October and even later.

This works for me, and I hope it sheds some light on the Es season in some way for others.

73 Jim

GM4FVM

Bernd Wilde, DL7APV, SK.

I learned via the EI7GL blog site (see link in sidebar) that Bernd, DL7APV, has become silent key.

Bernd was a fixture on 432MHz before he constructed his massive antenna in 2018, and after that he introduced many to the potential of the band. His behemoth array has 128 11 element antennas formed into a vast box with an overall gain of 33.7 dBd. Not only did I work him via Earth-Moon-Earth using this monster, I also worked him easily on tropo at a distance of over 1000km.

Bernd, DL7APV, now sadly SK, and his 70cm antenna

I will miss Bernd's gentle reminders about upcoming contests. He had a delightful way of expressing his wish for me to come on and give him a contact via moonbounce. He would send me an email and point out the date of the forthcoming event. He would then say that if it could be fun for me to have a contact with him, it would also be fun for him to have a contact with me during the contest. He did not say "come on and work me", but rather he correctly pointed out that I could if I would enjoy it. He was right: I did enjoy it.

Actually there was no need for a formal sked, I could hear him very clearly and reply to his CQ call. My antenna only has about 14 dBd gain, so he was doing all the work. No doubt he sent similar E mails to large group of operators who, like me, had enjoyed early EME success working Bernd on 70cm. 

I shall miss Bernd and I send his surviving family and friends my condolences.

Where will we ever find such an enthusiast again?

73 Jim 

GM4FVM

Amateurs - don't do as I say !!!

Amateurs - don't do as I say ... do what you want to do!

Please, I implore you - find your own way. Do not follow anyone's insistence, but learn from the many voices out there. Also, I encourage you to learn from what you do, and to progress in your own path.

And I know, that is me telling people what to do.

Irony, eh? Who needs it?

I recently posted an encouragement to VHF+ amateurs to try Q65, a splendidly efficient data mode devised by Joe Taylor and his merry tribe of creators. I did not say "give up your old modes". 

Here is a screenshot taken by Phil, EI9KP of our 2m QSO on Q65. Not bad at all - 479km on a flat band. We went on using Q65 successfully on 70cm. Eventually we tried 23cm, and although we did not complete a QSO there no doubt we will in due course. Q65 is good for contacts like this as it benefits from scatter enhancement such as that from aircraft. 

I think that it is best to use the best mode for the purpose, so for meteor scatter I use MSK144, for HF I would use FT8, and so on. Those were a great set of QSOs with EI9KP. You do not have to follow suit, but you can see how it can be done. As usual, click on the images to enlarge if you wish.

144MHz Q65-30B contact between EI9KP and GM4FVM on 25 March 2024 (Screenshot EI9KP)

Then I went on the 2m/70cm KST chat room [that is where you went wrong, Jim]. 

On KST someone asked for a meteor scatter contact and I explained that I can use MSK144. He then told me that I was wasting my time with MSK144 and I should install MSHV and use the JTMS mode instead. He said that there is "scientific evidence" from many hams to prove that JTMS was more efficient. 

What is the difference between me encouraging people to use Q65, and him telling me to install MSHV and use JTMS not MSK144. Is it just that I cannot take instruction, or learn from scientific evidence?

Well, I think not. I sincerely hope that everybody who reads this blog takes my ideas as representing what works for me. You can read about what I do, go away to your own blogs, and say the opposite. Or go to the local club and say that that old codger GM4FVM is talking nonsense. Sometimes I do talk nonsense. I can be wrong and I often am (source of ionisation for Es, for example). I can learn as I go along, and so can everybody else. I can even admit to being wrong.

Returning to the recent KST barney. I simply said that I use MSK144 on meteor scatter. This seemed to make things worse. After a blast in return I said that I respected his choice of mode and I hoped that he would respect mine. And then things calmed down a bit. I had some support from others on KST which I appreciate.

As for the scientific evidence, he did not say what he was talking about. Data modes for meteor scatter work on the principle that the meteor pings are short duration, thus the data rate is set so that the message fits into a very short timeframe. It is then repeated over and over again for 30 seconds in the hope that some of it will be transmitted at the same moment as a meteorite provides a tiny moment of propagation between the two stations. Different modes use different periods but from what I recall MSK144 repeats the message every 72 milliseconds. That would suggest that your message is repeated more than 415 times in your 30 second period, and almost all of those repeats is not detected by anyone.

The argument has been put that because the pings are of shorter duration at higher frequencies, then the time of repeat needs to be shorter too. That is a fair enough argument except that MSK is also up to about 8dB more sensitive than some other modes, and that affects the timing issue too. When you increase the sensitivity the pings are received for longer. This is because as the ping proceeds the amplitude declines, and with more sensitivity you can receive it further down the "tail" as it tails off. This means that the decode can lasts longer. Thus MSK does not need such a short timeframe. This is a benefit which some people use to argue against MSK, even though it is actually a benefit.

MSK144 is tailored for the 2m band and is probably over-engineered for 4m and 6m (so of course it works even better there). Maybe it would not be quite so good on higher frequency bands. The WSJT-X notes say it is designed for VHF bands so it would not be my choice for 70cm.

Anyway, I use MSK144 for another reason too. MSK144 has powerful error correction which other modes used for meteor scatter lack. I think that the reason why some people prefer other modes is because they produce less well corrected results which they interpret their own way. I do not criticise this, but I choose not to do it myself. I used older modes for years and I have had enough of looking at screens full of gibberish. Back then I quickly saw that I could interpret the results they produced in hundreds of different ways. To me that means anyone can interpret them in ways that suit them.

If you need some low-down on this you might find this from K5ND interesting:-

https://k5nd.net/2020/10/msk144-vs-fsk441-meteor-scatter-modes-my-scattered-compilation-of-data-points/

Before MSK, most modes used much weaker error correction, leaving operators to make sense of bits and pieces of decodes. The temptation to find evidence of success blinded quite a few to the fact that they were actually piecing together bits of decode provided by random noise.

70MHz FSK441 screen from 4 January 2016 showing partial decodes.

The example shown above is actually from a fairly long and successful period of reception on  4m - a band where results are better than on 2m. It looked like EA2BCJ was calling me, and this eventually proved to be the case. Maybe he was calling GM4F6M, which was also a possibility. Despite all the favourable factors, no single ping contained both correct callsigns together. In many cases there were screens full of garbage from which amateurs of the time would piece together the evidence they thought constituted a contact. Bill Somerville, G4WJS, is quoted in the K5ND piece in the link above commenting on those operators "who prefer to pick out a few relevant characters from a jumbled stream of uncorrected errors and decide that a QSO has been completed." Of course they can do that, but I do not want to.

Sadly Bill is now silent key and we miss his wisdom.

My simple decision was to use only MSK144 on meteor scatter. I think that I am free to decide that. Others seem happy to only use CW on Top Band, some only use 20 metres to have chats around Europe, still others just solder together circuits they never use. Fine business old men. You have that choice. I am not wasting my time doing what I want to do just because you do something else. I might say here that Q65 is going begging, but I don't tell people that they are wasting their time doing anything else. It is up to them.

We all think that our own set-up is the best. The only better one is the one we plan to buy when we can afford it. Most of us think that everybody else is barking up the wrong tree. Fine, but please don't preach to me.

I did not have the heart to tell this guy on KST that I already have MSHV installed here. I could run JTMS if I wanted to. I choose not to. As I said, mode is my choice, just as much as it is his. I could run some other meteor scatter mode, but give me proper error correction and lots of sensitivity every time.

Here is some advice from someone with years of experience - do not accept advice from people with years of experience. By all means listen to them, but form your own ideas yourself and go your own way.

Even if you go the wrong way, it is still your way. And, you can always learn some more and change later.

Finally, an 11 minute QSO from the good old days before strong error correction. Is this complete?

70MHz FSK441 contact between GM4FVM and SQ85PZK on 6 April 2015

Who decided that GM5JVM was wrongly decoded but GM4FVM was correctly decoded? Is this my QSO at all? I would rather leave decisions like that to the algorithms in WSJT-X.

73

Jim GM4FVM