New MGM Modes on Beacons


Andy G4JNT
 
Edited

With the introduction of new modes within the WSJT-X suite that use Gaussian shift MFSK, the old technique of switching between several switched tones using either a DDS or Fract-N PLL is no longer valid
 
Although not yet really applicable to microwaves, where JT4, JT65/Qxxx and PI4 will continue to reign, it is worth thinking about ways to implement FST4 and FT8/4 ready for if and when new GFSK modes appear that are suited to uWave usage.
 
This paper summarises the situation to date, and suggests a few ways forward.
 
 
 
Andy
 

Virus-free. www.avg.com


John Quarmby
 

Plenty of food for thought Andy, thank you very much. I'm thinking that upgrading existing beacons to retain low phase noise and spurs might be easiest by using a fixed RDDS for frequency control, with a baseband I/Q modulator driven from a PIC via DACs modulating the output of the RDDS to provide all the modes required.

The 13cm version of GB3MHZ (when it is working!) uses a Next Gen Beacon source running on 320.830MHz with FSK CW and PI4, mixed with 2GHz multiplied from the 1GHz Next Gen clock. Phase noise performance is quite respectable, and in band spurs are just about acceptable here 4km LoS from the beacon.

73

John G3XDY

On 16/03/2021 10:21, Andy G4JNT wrote:

[Edited Message Follows]
[Reason: A pile of silly typos corrected]

With the introduction of new modes within the WSJT-X suite that use Gaussian shift MFSK, the old technique of switching between several switched tones using either a DDS or Fract-N PLL is no longer valid
 
Although not yet really applicable to microwaves, where JT4, JT65/Qxxx and PI4 will continue to reign, it is worth thinking about ways to implement FST4 and FT8/4 ready for if and when new GFSK modes appear that are suited to uWave usage.
 
This paper summarises the situation to date, and suggests a few ways forward.
 
 
 
Andy
 

Virus-free. www.avg.com


alwyn.seeds1
 

Dear John and Andy,

Getting DDS spurs low enough for beacon use must be quite a challenge. At work, all of our DDS based signal generators, including the pricey ones, have quite a few spurs at the -80 dBc level- the nuisance is that they do not have easily predictable frequencies. We moan to the manufacturers but I have not seen much improvement over the last decade.

John, what sort of phase noise numbers do you get from the GB3MHZ beacon?

Regards,

Alwyn G8DOH
_____________________________________________________

Alwyn Seeds, Director
SynOptika Ltd.,
114 Beaufort Street,
London,
SW3 6BU,
England.


SynOptika Ltd., Registered in England and Wales: No. 04606737
Registered Office: 114 Beaufort Street, London, SW3 6BU, United Kingdom.
_____________________________________________________


John Quarmby
 

Alwyn

I have not got hard numbers to hand, but I can compare it with the 23cm beacon, where the beacon is approximately 95dB above noise in 2.5kHz bandwidth and results in a rise in the apparent band noise of 5db when I beam at it on 1296.200. My LO and the beacon oscillator are almost identical, so probably each contributes the same amount of phase noise. I have a crystal notch filter in the 28MHz transverter IF to reduce the beacon signal level by about 50dB going into the K3.

The 13cm beacon is about 10dB weaker, there is no change in band noise at 2320.200 when I beam at it. There are no major spurs in the SSB/CW part of the band, but they are apparent in the beacon band. When we have got it on air again I will make some proper measurements.

73

John G3XDY

On 16/03/2021 13:19, alwyn.seeds1 wrote:
Dear John and Andy,

Getting DDS spurs low enough for beacon use must be quite a challenge. At work, all of our DDS based signal generators, including the pricey ones, have quite a few spurs at the -80 dBc level- the nuisance is that they do not have easily predictable frequencies. We moan to the manufacturers but I have not seen much improvement over the last decade.

John, what sort of phase noise numbers do you get from the GB3MHZ beacon?

Regards,

Alwyn G8DOH
_____________________________________________________

Alwyn Seeds, Director
SynOptika Ltd.,
114 Beaufort Street,
London,
SW3 6BU,
England.


SynOptika Ltd., Registered in England and Wales: No. 04606737
Registered Office: 114 Beaufort Street, London, SW3 6BU, United Kingdom.
_____________________________________________________


Virus-free. www.avast.com


Robin Szemeti - G1YFG
 

" When the RF is passed through a frequency multiplier these spurii increase in relative amplitude as the square of the multiplication factor"

I found this "interesting" as I would have thought that if you had say 100MHz at 0dBm and 90MHz spur at -30dBm ... and then whacked the signal through a x3 multiplier, by driving some MMIC or similar hard into clipping ... I would have expected the output to be swamped by the 100MHz signal, and the "weak" 90MHz signal to be effectively blocked, and thus removed from the output,  and by the time it had been through the output filter to pick up the 3rd harmonic, I would expect it to look very clean indeed.  I had always assumed that multiplier chains cleaned up the spurs, not made them worse. I shall ponder this further ...


On Tue, 16 Mar 2021 at 16:57, John Quarmby via groups.io <g3xdy=btinternet.com@groups.io> wrote:

Alwyn

I have not got hard numbers to hand, but I can compare it with the 23cm beacon, where the beacon is approximately 95dB above noise in 2.5kHz bandwidth and results in a rise in the apparent band noise of 5db when I beam at it on 1296.200. My LO and the beacon oscillator are almost identical, so probably each contributes the same amount of phase noise. I have a crystal notch filter in the 28MHz transverter IF to reduce the beacon signal level by about 50dB going into the K3.

The 13cm beacon is about 10dB weaker, there is no change in band noise at 2320.200 when I beam at it. There are no major spurs in the SSB/CW part of the band, but they are apparent in the beacon band. When we have got it on air again I will make some proper measurements.

73

John G3XDY

On 16/03/2021 13:19, alwyn.seeds1 wrote:
Dear John and Andy,

Getting DDS spurs low enough for beacon use must be quite a challenge. At work, all of our DDS based signal generators, including the pricey ones, have quite a few spurs at the -80 dBc level- the nuisance is that they do not have easily predictable frequencies. We moan to the manufacturers but I have not seen much improvement over the last decade.

John, what sort of phase noise numbers do you get from the GB3MHZ beacon?

Regards,

Alwyn G8DOH
_____________________________________________________

Alwyn Seeds, Director
SynOptika Ltd.,
114 Beaufort Street,
London,
SW3 6BU,
England.


SynOptika Ltd., Registered in England and Wales: No. 04606737
Registered Office: 114 Beaufort Street, London, SW3 6BU, United Kingdom.
_____________________________________________________


Virus-free. www.avast.com


--
Robin Szemeti - G1YFG


Andy G4JNT
 

No, it doesn't work like that.   Visualise the multiplier as a hard limiter or zero crossing detector:

That -30dB 90MHz component adds to the 100MHz signal and periodically shifts the point at which the main 100MHz component crosses the zero threshold, resulting in a jitter on the limited output.  The resulting jitter then reappears as sidebands on the multiplied output.   You have to consider the total input, with spurii, in the time domain.

I can't supply the mathematical proof of the reason for squaring, or why it isn't any other order, but it's certainly there in practice and quite observable.

Andy


On Tue, 16 Mar 2021 at 17:46, Robin Szemeti - G1YFG <robin@...> wrote:
" When the RF is passed through a frequency multiplier these spurii increase in relative amplitude as the square of the multiplication factor"

I found this "interesting" as I would have thought that if you had say 100MHz at 0dBm and 90MHz spur at -30dBm ... and then whacked the signal through a x3 multiplier, by driving some MMIC or similar hard into clipping ... I would have expected the output to be swamped by the 100MHz signal, and the "weak" 90MHz signal to be effectively blocked, and thus removed from the output,  and by the time it had been through the output filter to pick up the 3rd harmonic, I would expect it to look very clean indeed.  I had always assumed that multiplier chains cleaned up the spurs, not made them worse. I shall ponder this further ...


On Tue, 16 Mar 2021 at 16:57, John Quarmby via groups.io <g3xdy=btinternet.com@groups.io> wrote:

Alwyn

I have not got hard numbers to hand, but I can compare it with the 23cm beacon, where the beacon is approximately 95dB above noise in 2.5kHz bandwidth and results in a rise in the apparent band noise of 5db when I beam at it on 1296.200. My LO and the beacon oscillator are almost identical, so probably each contributes the same amount of phase noise. I have a crystal notch filter in the 28MHz transverter IF to reduce the beacon signal level by about 50dB going into the K3.

The 13cm beacon is about 10dB weaker, there is no change in band noise at 2320.200 when I beam at it. There are no major spurs in the SSB/CW part of the band, but they are apparent in the beacon band. When we have got it on air again I will make some proper measurements.

73

John G3XDY

On 16/03/2021 13:19, alwyn.seeds1 wrote:
Dear John and Andy,

Getting DDS spurs low enough for beacon use must be quite a challenge. At work, all of our DDS based signal generators, including the pricey ones, have quite a few spurs at the -80 dBc level- the nuisance is that they do not have easily predictable frequencies. We moan to the manufacturers but I have not seen much improvement over the last decade.

John, what sort of phase noise numbers do you get from the GB3MHZ beacon?

Regards,

Alwyn G8DOH
_____________________________________________________

Alwyn Seeds, Director
SynOptika Ltd.,
114 Beaufort Street,
London,
SW3 6BU,
England.


SynOptika Ltd., Registered in England and Wales: No. 04606737
Registered Office: 114 Beaufort Street, London, SW3 6BU, United Kingdom.
_____________________________________________________


Virus-free. www.avast.com


--
Robin Szemeti - G1YFG


Dominique Dehays
 

Hi all ,


Yes my experiments show a 20 log N increase of spurs.
73
Dom


G8DQX list
 

If you think of the spurii as FM sidebands, then on multiplication by 2, for example, the modulation index (assuming it's very small) has doubled, thus a 6dB increase in level, and so forth.

HTH, 73, Stay Safe,

Robin, G8DQX

On 16/03/2021 17:53, Andy G4JNT wrote:
No, it doesn't work like that.   Visualise the multiplier as a hard limiter or zero crossing detector:

That -30dB 90MHz component adds to the 100MHz signal and periodically shifts the point at which the main 100MHz component crosses the zero threshold, resulting in a jitter on the limited output.  The resulting jitter then reappears as sidebands on the multiplied output.   You have to consider the total input, with spurii, in the time domain.

I can't supply the mathematical proof of the reason for squaring, or why it isn't any other order, but it's certainly there in practice and quite observable.

Andy


On Tue, 16 Mar 2021 at 17:46, Robin Szemeti - G1YFG <robin@...> wrote:
" When the RF is passed through a frequency multiplier these spurii increase in relative amplitude as the square of the multiplication factor"

I found this "interesting" as I would have thought that if you had say 100MHz at 0dBm and 90MHz spur at -30dBm ... and then whacked the signal through a x3 multiplier, by driving some MMIC or similar hard into clipping ... I would have expected the output to be swamped by the 100MHz signal, and the "weak" 90MHz signal to be effectively blocked, and thus removed from the output,  and by the time it had been through the output filter to pick up the 3rd harmonic, I would expect it to look very clean indeed.  I had always assumed that multiplier chains cleaned up the spurs, not made them worse. I shall ponder this further ...


On Tue, 16 Mar 2021 at 16:57, John Quarmby via groups.io <g3xdy=btinternet.com@groups.io> wrote:

Alwyn

I have not got hard numbers to hand, but I can compare it with the 23cm beacon, where the beacon is approximately 95dB above noise in 2.5kHz bandwidth and results in a rise in the apparent band noise of 5db when I beam at it on 1296.200. My LO and the beacon oscillator are almost identical, so probably each contributes the same amount of phase noise. I have a crystal notch filter in the 28MHz transverter IF to reduce the beacon signal level by about 50dB going into the K3.

The 13cm beacon is about 10dB weaker, there is no change in band noise at 2320.200 when I beam at it. There are no major spurs in the SSB/CW part of the band, but they are apparent in the beacon band. When we have got it on air again I will make some proper measurements.

73

John G3XDY

On 16/03/2021 13:19, alwyn.seeds1 wrote:
Dear John and Andy,

Getting DDS spurs low enough for beacon use must be quite a challenge. At work, all of our DDS based signal generators, including the pricey ones, have quite a few spurs at the -80 dBc level- the nuisance is that they do not have easily predictable frequencies. We moan to the manufacturers but I have not seen much improvement over the last decade.

John, what sort of phase noise numbers do you get from the GB3MHZ beacon?

Regards,

Alwyn G8DOH
_____________________________________________________

Alwyn Seeds, Director
SynOptika Ltd.,
114 Beaufort Street,
London,
SW3 6BU,
England.


SynOptika Ltd., Registered in England and Wales: No. 04606737
Registered Office: 114 Beaufort Street, London, SW3 6BU, United Kingdom.
_____________________________________________________


Virus-free. www.avast.com

--
Robin Szemeti - G1YFG


Mike Lavelle K6ML
 

A few comments:

- not all of the new WSJT modes use gaussian shaped keying...  when I asked, K1JT replied that the modes with tone counts greater than 4 did not benefit from gaussian shaped keying (so, FST4, yes, Q65, no; the earlier JT4 (which has unfortunately been declared 'obsolete' and removed from the latest release candidate), JT65 and QRA64 (also removed) do/did not use gaussian shaping).

- gaussian shaping mainly helps cut down the transmitter sidebands (a good thing on crowded HF bands, and probably for a beacon, too).  it actually makes the decoder's task more difficult (more intersymbol interference from the 'smoothing' is the price for less 'splatter').   I suspect that the decoders for a gaussian keyed mode would be happier (if they had emotions :) with hard keyed modulation and probably are not (can't be) doing anything special to listen to gaussian keyers.

So, other than VLF and WSPR beacons, I suspect the gaussian keying is not providing benefit, nor being used in many modes.  I also believe that at microwave and above, there is no need to worry about it.


Andy G4JNT
 

JT4 is still in the latest (a couple of days ago) 2.4.0 RC3 version.   It had better remain, as it's in quite widespread use on microwave beacons, being one of teh few modes capable of coping with really widespread scattering such as that from rainscatter on 10 and 24GHz

When I asked Joe about it a while back, he assured me that JT4 and JT65 would be kept for this purpose.   Beacon modulations can't be changed on the whim.



On Wed, 17 Mar 2021 at 02:00, Mike Lavelle K6ML <miclevel@...> wrote:
A few comments:

...; the earlier JT4 (which has unfortunately been declared 'obsolete' and removed from the latest release candidate), JT65 and QRA64 (also removed) do/did not use gaussian shaping).

- .


Virus-free. www.avg.com


Mike Lavelle K6ML
 

Andy,

Sorry, that was my mistake.   I mixed up ISCAT (which is unfortunately being deleted... it’s still being used for microwave aircraft scatter) with JT4 (which is alive and well).    In the last part quoted, I was not writing that JT64 was retired, just QRA64 (the parens only applied to QRA64).  

While I’m at it, my comment about which modes use hard vs Gaussian keying (and my quote from Joe that modes over 4 bits were hard keyed) was focused on microwave modes, so ignored FT8, which, as your writeup mentions does use Gaussian. 

I also agree with your statements about frequency multiplication amplifying spurs by 20 dB per decade (same as phase noise and FM noise) being a serious issue by at microwaves and even worse at millimeter waves.  Working at 122 GHz, with a 12,200x multiplication factor, I have managed to produce some awful spectra with serious spurs, FM and PN.   John Miles (KE5FX) suggested on his excellent website that a great way to find out how truely awful a VHF source can be is to use it as the reference input to a microwave ‘brick’ and then look at the multiplied spectrum at microwaves.   ‘Clean’ oscillatros can shown to be quite ‘dirty’ using this technique.        

It looks like translation loops (PLLs with mixers) may be a good solution to spurs, etc, at SHF/EHF.    

I don’t see a problem with using frac-N with MFSK ... the requirement is CPFSK, which PLLs do meet... the VCO waveform is not discontinuous, just not elegantly shaped FM.... and if the PLL loop bandwidth is low (but still at at least 10-20x the baud rate), the low pass filtering produces smooth keying.


Mike K6ML

Mike K6ML


Bo, OZ2M
 

Good article! I think it was Bill, G4WJS, who wrote somewhere that the driver for the Gaussian keying was the crowed segments on the HF bands and use of spectrum. On VUSHF this is probably not an issue. The use of Gaussian keying requires a slightly wider FFT BW causing a S/N loss of a few 1/10 dB.

Back in spring 2018 Rolf, DK7FU, contacted me about the key click from DB0HRF, 2 m, that disturbed the DR9A contest group 172 km away. I made and installed a new S/W on the AD9912 driver that is able to perform linear intermediate steps when shifting from one tone to the next. We eventually settled for four intermediate steps. The steps reduced the key clicks by ~12 dB. More steps did not provide any significant difference. With 100 steps the signal sounded like a slide guitar straight out of Nashville. 100 steps also fooled PI-RX so this was not feasible at all.
BTW: if you ever get near Frankfurt a visit to the site of DB0HRF is worthwhile: https://taunus.info/angebote/natur-und-erlebnis/berge-und-felsen/grosser-feldberg/905/ The beacon on 2 m uses the "small" red-white omni-directional antenna.

I doubt that K = 1 modes will be successful on high VHF and USHF.

Bo
www.rudius.net/oz2m :: www.rfzero.net


Andy G4JNT
 

Yes, my thought behind doing this was FST4/FST4W  as a replacement for a WSPR beacon.



Virus-free. www.avg.com


On Sat, 20 Mar 2021 at 09:26, Bo, OZ2M <groups.io@...> wrote:
Good article! I think it was Bill, G4WJS, who wrote somewhere that the driver for the Gaussian keying was the crowed segments on the HF bands and use of spectrum. On VUSHF this is probably not an issue. The use of Gaussian keying requires a slightly wider FFT BW causing a S/N loss of a few 1/10 dB.

Back in spring 2018 Rolf, DK7FU, contacted me about the key click from DB0HRF, 2 m, that disturbed the DR9A contest group 172 km away. I made and installed a new S/W on the AD9912 driver that is able to perform linear intermediate steps when shifting from one tone to the next. We eventually settled for four intermediate steps. The steps reduced the key clicks by ~12 dB. More steps did not provide any significant difference. With 100 steps the signal sounded like a slide guitar straight out of Nashville. 100 steps also fooled PI-RX so this was not feasible at all.
BTW: if you ever get near Frankfurt a visit to the site of DB0HRF is worthwhile: https://taunus.info/angebote/natur-und-erlebnis/berge-und-felsen/grosser-feldberg/905/ The beacon on 2 m uses the "small" red-white omni-directional antenna.

I doubt that K = 1 modes will be successful on high VHF and USHF.

Bo
www.rudius.net/oz2m :: www.rfzero.net