Wednesday, May 20, 2020

What WSPR signal to noise ratio is needed for sideband to be intelligible?

The end-fed antenna is working pretty well on 40m and I'd love to have a contact with Sydney friends. John, VK2ASU, kindly agreed to run a WSPR beacon so I could find the best times to receive him.

Over a 24 hour period, with him transmitting just 10mW, the best SNR was -21dB here.

My question, to the brains trust hopefully reading, is what WSPR signal to noise ratio would be enough such that a higher power sideband call would be audible?

I have read that a signal to noise ratio of 6dB is needed to be able to copy SSB.

John is transmitting 10mW which is 10dBm but he could run 100W on sideband or 50dBm. An extra 40dB.

If the bandwidths were the same then -21dB + 40dB = 19dB of signal to noise but while SSB is about 2.5kHz wide, WSPR is much narrower, perhaps only a few Hz?

Here is the SNR between VK2ASU and VK2TPM (portable VK3).


Spots from all stations to me looks like this:


Any insights or pointers gratefully received.

Kevin, VK2KB, commented: "My understanding is the noise power doubles every time the bandwidth is doubled while the power level of the received signal stays constant. I've verified this on my SA by changing the bandwidth of the SA response and noting the noise floor, you can see the floor decrease by 3db every time I halve the  SA  window but the amplitude of the signal I'm watching stays the same.  With WSPR you need to know what the effective bandwidth is. I had a look at this site:
https://en.wikipedia.org/wiki/WSPR_(amateur_radio_software)

They say the effective bit rate is about 1.5 baud  and its possible to recover a wspr signal at -34 db  on a 2500Hz bandwith rx.

Say that you are using an effective bandwidth of 1.5 Hz to recover the digital signal the ratio of 1.5/2500 =  .0006 or -32db  Because the wspr signal is coherent I think they get a few more dbs in the demodulation process.

The difficulty with this is however the power bandwidth of the wspr signal vs the power bandwidth of the ssb signal.  Anyway something to spend time on."

Update October 2023

Reader Math Pantus, PA3BUL, wrote an interesting response which he gave me permission to post here.

WSPR was made to be detectable 12dB below a
CW signal. 

So that we obtain an indication of
a propagation path that started to open up.

[about 2 S-units below]

CW vs SSB difference in readability is just
about 30dB.

> So 1Watt of CW is worth about 1000Watts in USB
for readability.

Here are the reasons for CW to be so readable:

1) CW 1 Watt (Key-Down) is 1 Watt
   USB 1 Watt is just .25 to .35 Watts due
   to the Peak-to-Peak versus average voice
   power. Screeming AAAAAAAH will give you
   .3 Watt and with the best compressor you
   will get .35 Watts. At .4 Watts readability
   suffers a lot.

2) Bandwidth of CW is just about 0.8 * WPM.
   So over 15 WPM confusion increases much.

   The bandwidth of your CW IF filter is
   enhanced with your brain expertise that
   brings the readability very close to the
   0.8 * WPM limit.
   This because of psycho-acoustic effects.
   
   The same effects that can compress music
   into MP3 format with less that 10% disc-space.

 NOTE:  U wide CW roofing filter can be just as good
 as a narrow one. Just more fatigue will be the result.  

For a voice to be readable we need just 20dB of
Signal to Noise margin.
Using the NATO words for spelling, we can creep by
with 15 to 16 dB of SNR.
(As the use of NATO words is a form of FEC,
Forward-Error-Compression).

>So 1Watt of WSPR is just as good as 16 Watts of CW and 16 KWatts of SSB.

NOTE: K1JT refers all WSPR signals to 2500Hz of receiver bandwidth.
            this makes it easy to compare.

My try at answering your question is:

If WSPR signals come in at + 42dB or more, you could pick-up the Mike and give it an SSB try.

NOTE: WSPR encoding uses: Data compression, redundancy, FEC, pseudo-random time slots, DSP
      narrow-band filtering synchronous detection.
      All in order to enhance de-code ability and selectivity.
      Also reducing the effects of QSB, QRN, and QRM.  
 
>The compression is like a PKZIP compression of the call sign plus QRA-locator.
> After compression, to the compressed data, redundancy is added.
> During the transmission the receiver and transmitter hop-skip and jump through the
   same time-slot. It is like encryption where the key is determined mathematically at both
   sides. This method avoids data clashes and corruption by QRM, QRN and QSB.
   Therefore WSPR can be used in cities with lots of QRM.

NOTE: WSPR lets a poor antenna look good.
            WSPR lets a good antenna look sublime. (Look at VK3MO   wspr results)
      He uses just 5 Watts in a 20dB gain antenna. So 500Watts ERP.
      
Last NOTE: I am trying to create a procedure to use the WSPR modulation quality to
have 2-way contacts.

3 comments:

Unknown said...

Hi Peter,
https://www.qsl.net/kp4md/wsprmodes.htm
contains some interesting comparisons of WSPR vs other modes.

Don't forget the VK2RWI CW beacon on 3699 is a good Sydney beacon.
VK3ASE is over 59+20dB on AM into Sydney tonight on 3670!

Hope the move went well and you're re-acclimatizing to those cooler temps!
- Jules VK2EXT

Peter Marks said...

Thanks Jules, that's a great reference.

Chris VK2CJB said...

I've been told by another WSPR Tragic that there's a 500x/27dB advantage over SSB-- not that I've been able to rouse other VHF operators during above-average conditions. So.. during a WSPR 10mW/-21dB, 10W might get +6dB SSB.

(A while back I gave John my own code that cycles through the Si5351's power settings. Using that, I've managed a 2mW ping on 40 metres with a station in the States.)