locked Performance of FST4-120 vs that of WSPR #FST4 #WSPR


Mike M0AGP
 

* Not sure if this is the right forum, but here goes! *
I've been trying to make a QSO over a difficult path on 160m: G to ZL, and have tried FST4 a number of times.
I wonder if I am doing something wrong, as I have been copied in ZL on WSPR numerous times but never on FST4, even on FST4-300.
What is going on?
Has anyone else out there compared FST4 with WSPR over the air (as compared to using a source of white noise)?

We are trying this after UK sunset and just prior to ZL sunrise, as is customary for this path.

I guess we can't post screen shots so here are the settings:
Freq: set to desired frequency like 1839 for 160m (I tend to use a lower frequency to avoid stepping into the FT8 band).
Tx 1000Hz
Rx 1000Hz
F Low 900
F High 1100
T/R 300 s
Auto Seq (ticked)
Call 1st (Ticked)
Tx even/1st [Obviously one of us is on evens and one is on odds when trying for a QSO...]
Hold Tx freq (ticked)
NB 0% - I read something in an early document that some people set this to 10-15% in some cases - is there much of an effect here?

I have used FST4 several times in tests with a buddy in the states, and have had no trouble, but we didn't compare to WSPR.
It would be especially interesting to know if anyone has compared FST4-120 and WSPR over the air.

One other question in case an experienced FST4 user shows up: in making a sked, is there any advantage to narrowing the "F Low to F High bandwidth"?
Any good tricks pls?

Thanks and 73
Mike M0AGP


Jim Brown
 

On 2/4/2022 9:30 AM, Mike M0AGP wrote:
I've been trying to make a QSO over a difficult path on 160m: G to ZL, and have tried FST4 a number of times.
I wonder if I am doing something wrong, as I have been copied in ZL on WSPR numerous times but never on FST4, even on FST4-300.
What is going on?
Are you trying a sked with a specific station, or just calling CQ? I haven't looked this season, but last season there was almost no random FST4 activity on 160M. Likewise for FST4W. I tried a FST4 sked with a VK, and had to go to 120 sec and legal limit for him to hear me. He was pretty loud here. That might have been a seasonal problem -- summer there with lots of QRN, winter here with much less.

73, Jim K9YC


Mike M0AGP
 

Hi Jim
It was a sked, having exchanged exact settings including screenshots of settings, etc.
Good that you managed to get a FST4 sked to work, and also with a VK!
I tried with a VK and it failed. A few days later we had freakishly good prop and managed it on FT8 - just barely.
I'm wondering if FST4-120 is actually more sensitive than WSPR in real life situations, not just with white noise testing.
Maybe it is, and I was unlucky in that propagation was always worse when I used FST4.
73
Mike M0AGP


N1BUG
 

Hi Mike,

FST4W-120 is about 1.4 dB more sensitive than WSPR. I was part of the alpha test team for FST4 and FST4W. Extensive comparisons on 2200 meters under real world conditions showed that FST4W got more decodes than WSPR. FST4 is 1.5 dB less sensitive than FST4W, so that puts it about on par with WSPR. Note that AP decoding offers up to 4.7 dB advantage over FST4 without AP. This means that for messages which can take advantage of AP decoding, FST4 is considerably more sensitive than WSPR.

I am not aware of any controlled, real world comparison of FST4W to WSPR on 630m or 160m. Results there may vary. Please note that these modes were developed for the LF (2200m) and MF (meaning 630m) bands primarily. Achieving the sensitivities noted above requires that frequency drifts and path-induced doppler shifts must be less than the tone spacing over the entire sequence length. These requirements are easily met for the 120 second submodes at LF and MF. Path-induced doppler may be greater at 160m but it is my belief that 120 second FST4 and FST4W submodes should do well there, assuming equipment used has suitably low drift.

Submodes with longer sequence length use narrower tone spacing and thus require greater frequency stability and lower path-induced doppler shifts. Submodes all the way down to 1800 seconds have proven extremely effective at 2200m but the longer 900 and 1800 second variants have been somewhat less so on 630m. This may be due to several factors. Clearly some of the stations trying 300 and 900 second submodes on 630m did not have sufficient transmitter stability. As far as I know, only limited testing of 900 second transmissions has been done on 160m with reasonable success, but I am not aware of testing over paths of great length such as what you are trying to do.

73,
Paul N1BUG

On 2/4/22 18:10, Mike M0AGP wrote:
Hi Jim
It was a sked, having exchanged exact settings including screenshots of settings, etc.
Good that you managed to get a FST4 sked to work, and also with a VK!
I tried with a VK and it failed. A few days later we had freakishly good prop and managed it on FT8 - just barely.
I'm wondering if FST4-120 is actually more sensitive than WSPR in real life situations, not just with white noise testing.
Maybe it is, and I was unlucky in that propagation was always worse when I used FST4.
73
Mike M0AGP


Bob K4RCG
 

Great summary Paul N1BUG. Tnx for your contributions to WSJT-X I also do
not know of any real-world *controlled* comparisons between WSPR-2 and
FST4/W(x) .

I did participate along with several GM amateurs in a
non-empirical comparison of FST4W/120/300/900 to WSPR-2 at 160 through
40m.. It was a fun *test* to compare WSPR-2 to FTS4W/120/300/900, but I'd
be very hesitant to say we drew any conclusions from the
observations/data. Anecdotally, I would offer that we did observe FST4W(x)
decodes down to -34dB (perhaps lower, but I don't recall specifics and
don't have the test data any longer)....but -34db is a level not often seen
on WSPR-2. As I recall, our general "findings" were consistent with the
published info on FST4W sensitivity and that sub-mode usefulness above
160m. We also noted the above mentioned path induced doppler effects
became much more pronounced above 160m.

This question would be great for further analysis and perhaps lead to WSJT
decoder enhancements in the future, How weak can we decode??? :-)

73 de Bob
K4RCG

On Sat, Feb 5, 2022 at 3:47 AM N1BUG <paul@...> wrote:

Hi Mike,

FST4W-120 is about 1.4 dB more sensitive than WSPR. I was part of the
alpha test team for FST4 and FST4W. Extensive comparisons on 2200 meters
under real world conditions showed that FST4W got more decodes than
WSPR. FST4 is 1.5 dB less sensitive than FST4W, so that puts it about on
par with WSPR. Note that AP decoding offers up to 4.7 dB advantage over
FST4 without AP. This means that for messages which can take advantage
of AP decoding, FST4 is considerably more sensitive than WSPR.

I am not aware of any controlled, real world comparison of FST4W to WSPR
on 630m or 160m. Results there may vary. Please note that these modes
were developed for the LF (2200m) and MF (meaning 630m) bands primarily.
Achieving the sensitivities noted above requires that frequency drifts
and path-induced doppler shifts must be less than the tone spacing over
the entire sequence length. These requirements are easily met for the
120 second submodes at LF and MF. Path-induced doppler may be greater at
160m but it is my belief that 120 second FST4 and FST4W submodes should
do well there, assuming equipment used has suitably low drift.

Submodes with longer sequence length use narrower tone spacing and thus
require greater frequency stability and lower path-induced doppler
shifts. Submodes all the way down to 1800 seconds have proven extremely
effective at 2200m but the longer 900 and 1800 second variants have been
somewhat less so on 630m. This may be due to several factors. Clearly
some of the stations trying 300 and 900 second submodes on 630m did not
have sufficient transmitter stability. As far as I know, only limited
testing of 900 second transmissions has been done on 160m with
reasonable success, but I am not aware of testing over paths of great
length such as what you are trying to do.

73,
Paul N1BUG



On 2/4/22 18:10, Mike M0AGP wrote:
Hi Jim
It was a sked, having exchanged exact settings including screenshots of
settings, etc.
Good that you managed to get a FST4 sked to work, and also with a VK!
I tried with a VK and it failed. A few days later we had freakishly good
prop and managed it on FT8 - just barely.
I'm wondering if FST4-120 is actually more sensitive than WSPR in real
life situations, not just with white noise testing.
Maybe it is, and I was unlucky in that propagation was always worse when
I used FST4.
73
Mike M0AGP





Mike M0AGP
 

Paul I echo Bob's thanks for your contribution to WSJT-X. Coming from physics and with a high-level understanding of information theory, I appreciate the amazing achievements here. I especially was thrilled to hear that a physicist (JT) was behind much of this (note I was QRT from 1981-2020 - so this is still new and amazing)!

When I read that the FST4 mode was optimized/developed for MF and LF I was thinking of 160m as MF. But from what you say it sounds like it was optimised for 603/2200m.
So it was not optimized for 160m in that case?

I have no idea about path-induced doppler shifts for the the G to ZL path. I guess such effects come from reflection off of moving ionized layers, so the longer the path the larger the effect. If such things are approximately random, maybe we should expect the effect to increase as the square root of path length?

Would WSPR mode also be vulnerable to the same path-induced doppler shifts? I'm interested as to what situation might make WSPR "work" better than FST4-120.

I will have a look at how widely used FST4W is - I've never tried it!

All very cool stuff!
Thanks
Mike M0AGP