Yes, same here when using the show_rawsyms3a utility. The peak in the skyblue spectrum-of-squares  will conveniently guide you to the correct frequency and time offset if Eb/N0 is above 4 dB or so. This is for rate 1/8, with more code spread (e.g. r 1/16) the "squaring loss" will be a bit worse.  

 

Here are the Eb/N0 results of successful decodes this afternoon:

 

 UT    dB
1500  3.1
1600 11.8
1700  9.0
1730  8.1
1800  4.4
1830  6.4  ***
1900  2.7
1930  7.9
2000  4.0
2030  1.5
2100  6.3

 

Best 73,

Markus

 

 Markus wrote: 

  

  >  I would't rule out propagational Doppler. 

  

 I see what you mean, the carrier looks quite ragged on 

 there too.   I guess I'm used to the nice sharp lines 

 we get a VLF.   These super high frequencies are a bit 

 different! 

  

 But I certainly have a problem with the timing system 

 here.  I'm capturing through the audio input of a 

 BT878 video chip sampling at 448k/second and discipling 

 the rate by centroid measurements of a pulse-per-second 

 with the LF mixed in, sharing the mono input.  It's 

 pushing things a bit to say the least. 

  

 I've altered some magic numbers in vttime to get some 

 long term stability at the cost of extra short term jitter. 

 In a few minutes we'll see how the 20:00 message looks. 

  

 Nice decode from Andy.   Eb/N0 from the symbol errors 

 is 3dB below the carrier measurement.  I'm not sure 

 what that implies.   This is all very new territory to 

 explore, and great fun! 

  

 The 20:00 just finished.  Decoded! 

  

 found rank 0: GE PAUL TNX RPRT  ps [ 15 -120 -120 -150 -150] 

 re-encode 290/1040 ber 2.79e-01  Es/N0=-7.6 Eb/N0=2.7 effrate=1/10.83 

  

 You'll notice that with weak signals, say less than about 

 Eb/N0 = +3dB, it becomes difficult to estimate the reference 

 phase by squaring the signal (or equivalently, costas looping). 

  

 If the initial guess of the reference phase is good, the message 

 will decode almost immediately with reference phase pattern 

 0 0 0 0 or 180 180 180 180.  But you can see from mine an 

 Andy's decodes that the initial guess based on squaring was 

 poor, and the program really has to do some trial and error 

 to get the best decode, which is coming out some 60 deg 

 adrift from the initial guess.   When you get down to around 

 zero Eb/N0 the initial guess is pretty worthless.  This 

 is one of the key features that sets EbNaut apart from 

 other coherent modes: it doesn't rely at all on trying to 

 extract a reference phase.  This is what allows it to 

 go well below 3dB. 

  

 -- 

 Paul Nicholson 

 --