It looks like the RFzero could fit the bill, but there is one big problem and that is availability.  This situation will probably not resolve in the near future so this unit is out if you don't own one now.   I have been trying the QRP labs QDX, but the TX is not good for FST4w  for various reasons as the frequency spread is too much during tx.   It could be fixed from a thermal standpoint  and with a GPSDO, but it looks like the modulation method is also the culprit for the high  tx spread. To fix this would probably require minimum a software mod.  The code however (to hide it from being copied) is encrypted.  Hans is very busy right now filling orders and chasing components so it is safe to assume that any changes would be a very low priority for them. 

Now, what else is out there that could work but does not break the bank??

Early on in the FT8  introduction there was a number of Xcvrs coming out, but they were all DSB with a DC receiver.  Then came the FT8 Phaser by Dave Benson K1SWL.   The QRP guys also came out with their Mod III.  In looking at those I think the old phasing method of generating SSB is probably the most viable for FST4w.    I am thinking of  perhaps use the FT8 Phaser tx design  with a GPSDO.  Comments anyone? 

One very important question though and that goes back to my initial experiment with the QDX.  How stable is the WSJT-X as a source and generator  for FST4w?  

Rolf

K9DZT

 

Rolf,

I agree with your interest in a low cost FST4W-capable transmitter.  Gwyn and I also own and have measured the QDX's suitability for FST4W on HF.  As you describe, what we found is that even with a GPSDO fixing the transmit temperature related problem, though on receive it is a quite good, spreading due to the method Hans uses to follow and steer the Si5351 is not. It is so bad that it really isn't a candidate for this task.  We documented this and I've sent the results to Hans but he seems to have gone silent after initial interest.  Perhaps additional notes to him from others about the desirability of a fix might encourage him to raise the issue higher on what is certain to be a large to-do list.  Other than this group, I don't think many hams are trying to use the QDX for HF FST4W.  It certainly would be a nice solution since it has four bands and an SDR receiver which works well.

I wouldn't be surprised if the difficulty in component procurement isn't greatly hampering efforts for not only the RFzero but also QRP labs.  The 'usual suspects' like Digi-Key and Mouser have almost all of the low cost SkyWorks (used to be from Silicon Labs) synthesizer parts on 52 week or similar delivery.

A phasing I/Q detector or transmit modulator of the Tayloe variety is easily implemented using even the Si5351, which is what Hans' QDX does on receive, since the part can produce quadrature sin/cos LOs easily and GPSDO referencing its ~25 MHz oscillator or injecting an external GPSDO signal isn't difficult, but without parts its impossible. 

Recently I've phase locked my Icom IC7300 with a simple, passive, injection locking scheme that is showing on the order of 5 milli-Hz spread on 20m. This works fine but even on the used market this transceiver costs 10X as much as the QDX.   I wish there were a better way or product but I haven't found or thought of it yet and certainly don't want to go into the manufacturing business (for profit) in any case!  My hat is off to Hans and others who are willing to take on all the headaches that entails.

From other measurements I've found that the signal generated by typical PC sound cards (and their crystals) by WSJT-X is pretty good.  The (sole Windows) PC I have here shows  only 10 or 20 milli-Hertz error when generating a 1500 Hz tone. This as measured either by fldigi (which actually allows correction of yhe error) or WSJT-X's frequency measurement mode. Since it is a stable signal I don't think this gets in the way of either decoding or encoding even on -1800 FST4W sequences. So I think for low cost HF FST4W transmitters the difficulty lies in using a very clean LO along with a suitable modulation scheme. So far I don't know of one.    Maybe Hans will be able to fix the QDX.  I hope so.

Glenn n6gn

--

Well gang nothing is easy anymore these days. 

Glenn answered my question about WSJT-X  and I spent some time on the FT8 Phaser circuit.   I think it can be done with this circuit. Here is my thinking -  I am only looking at the tx chain.  First I would use a Bodnar mini GPSDO programmed for 4 x the tx freq (have adjustable gain and 3.3V square wave).  Then I would feed the 74ACT74 directly from this.   I would use WSJT-X to drive the audio channel  Anything like a Raspi would probably do as long as you use a high performance USB sound card.   

The FT8 Phaser is no longer available (go figure)  but I can get my hands on a 20 m one  for testing.  

Comments anyone?

73  Rolf  K9DZT

    

--

Cheap he? Maybe not cheap. What about a Red Pitaya STEMlab 125-10 (not the 14 bit or the SDRlab 16 bit but the entry level version) with a GPS module connected. Only needs an impedance transformer and a very small amp. Can do the decoding from 2 antenna's in parallel as well. But would need FST4W software to be made :-(

Happy I already have a RFZero, but no amp here.

Would the Hermes Lite V2 work ? Not in production currently.

A self generated FST4W message/signal would be preferred instead of having additional cpu+wsjtx+audio out+audio in for a single band non band switchable signal from WSJT-x

I agree with that if possible to use direct modulation of the SI5351A. However, it is a very complex to get this right and the same goes for how the GPS correction is applied to the system.  That is why my first thought was to use the Bodnar.  It is available and proven. The rest of the phasing circuit is what I like to test first. That way we only test one variable at the time. 

Rolf  K9DZT

Rolf
Your first post ended with the question, "How stable is the WSJT-X as a source and generator  for FST4w?" 
The attached table gives the answer for the best possible case from my measurements: Baseband audio from WSJT-X 'transmitting' FST4W from one computer to the audio input of another 'receiving' the FST4W, decoding and outputing its spectral width measurement (Macbook Pro and Macbook Air). These provide the 'gold standard' against which to test transmit-path-receive setups. Using GPSDOs each end, and over a line of sight path, Glenn, N6GN to N6GN/K is currently showing 31 out of 104 spots decoded with SW of 2 mHz, with a tail to 15 mHz.

Gwyn G3ZIL

[ not to list, only to yourself]

Rolf,

Perhaps I wasn't clear. I wasn't suggesting direct modulation of the synthesizer. I was only pointing out that a Si5351, the low cost version of the line but fairly un-obtainium as are the more expensive and more capable parts, is capable of directly generating quadrature signals and that 4X LO and dividers was unnecessary. 

I guess I'm still unclear as to your goal, whether a single FST4W transmitter or a design for quantity production. Relatedly I'm also unclear what role you think cost plays in this.  Since you are suggesting a Bodnar, already this is not an inexpensive solution, even if the transmitter is free.

I was trying to point out that generation of quadrature LO could be done more easily and I would add that generation of high accuracy quadrature audio is also very easy. ~50 dB of accuracy is possible over a decade of BW with passive 6-tier RC networks of this sort, more if less range, e.g. 500-2500 Hz for FT8 or even less for FST4W:

This is relevant when one contrast this approach with the "Phaser" you showed. That has the problem that even with 1% parts probably no more than 40 dB of balance and sideband suppression can be achieved. That may be OK for QRP but really isn't very good amateur practice.

Simple generation of quadrature combined with simple mixers allows multi-band SSB generation with low cost even after LPF and PA is added. That leaves the GPSDO question. To solve that, I have designed and built a "poor man's Bodnar". The goal of this is something nearly as good as a Bodnar and capable of FST4 performance through 23 cm.  I may not have *quite* achieved this, the Si5351 itself may be a limiting factor but at ~$2 instead of 10x that it seemed a good choice. Total parts cost in small quantities can be  $25-$50 depending on quantity and availability, as mentioned. It uses an IoT33/WiFi (an Arduino) and provides web server interface. It's small and nearly all that would be necessary to make an SSB up-converter, quadrature audio and mixer perhaps being added:

Here's one of them presently locking my IC7300 which as Gwyn as posted in other mail produces ~5 mHz of total spread when measured over a ~20 kM  non-ionospheric path with a Bodnar/Kiwi as receiver

In this design there still are two available clock outputs which could be operated at the same frequency but in quadrature which could be used from LF through 2m as mentioned above. Thus I think much lower system cost is possible than other approaches mentioned.

BUT the parts aren't readily available and who wants to go into a small business that requires customer support and doesn't offer much upside other than the goodwill of the small group of users?  As I said elsewhere, my hat is off to Hans Summer and QRP Labs.  I don't want to do it !

Also OTOH, if building a very small quantity or single transmitter is what's fun, by all means do that.

Best

Glenn n6gn

Phooey, I didn't mean to burden the list with this, Reply-To v. From fooled me once again.

It was still interesting.

On the continuing subject of availability of Si5351A chips, Amazon has lots of them available in generator boards in the $8 to $12 range. Could these be modified to be useful?

Steve KD2OM

Sent from my iPhone.

Yes. I have been buying them from Amazon and AdaFruit and taking them and the crystal off the board to re-use.  Costs about double but at least there is still availability.  Hard to say if these will run out before production starts again.

The same doesn't seem to be true for the more capable parts, such as the Si5328 that Bodnar uses in the Mini and standard GPSDOs. He ran out of stock for a while but recently indicated stock had returned. Maybe he found a source somewhere.

I'm very interested if anyone sees a reliable source with a reasonably short lead time.

I took a peek on Amazon and they besides selling the original Adafruit SI5351A board with the onboard  clock xtal they are also selling a Chinese knock off  "Hi Let go" brand and some others.  They all look similar.
I have used the Adafruit one, but  it is still using a free running clock generator.  W3PM had a write up in QEX some years ago using the Adafruit one for a GPS corrected VFO.   As far a the Chinese knock off  goes  only testing them will tell.

As far as the Adafruit unit goes the Xtal can come off and be replaced with a high quality TXCO and/or inject the SI5351A chip with a clock  that is temperature and frequency stabilized preferably by a GPS Source if you have FST4w in mind.  

Rolf K9DZT     

Rolf,

I've used the knock offs and they seem to be OK.

The Reference I built makes a VCXO out of the Si5351/xtal which is then lockedto either the GPS constellation by way of the uBlox timepulse oor a local 10 MHz reference from OCXO or Rubidium standard. See the block diagram

I think trying to use a 'high quality TCXO' is probably a bad idea, at least on longer FST4W modes, unless perhaps only for receive. The QDX does very well this way - until one starts transmitting and thermally cycling the board with PA heating.  Trying to free-run any TCXO on 20m for FST4W transmitting is probably not a very good idea, IMO.

Unless others prefer it, let's take these details off-list.  I'm feeling bad about filling everyone's mailbox with details they might rather not see here on WD group.

Glenn n6gn

--

https://github.com/W3PM/Auto-Calibrated-GPS-RTC-Si5351A-FST4W-and-WSPR-MEPT/blob/main/w3pm_GPS_FST4W_WSPR_V2.ino

(cross posted in other subject also)

Erwin

On Tue, Oct 11, 2022 at 08:38 AM, Rolf Ekstrand wrote:

I agree with that if possible to use direct modulation of the SI5351A. However, it is a very complex to get this right

Rolf, what complexity are you thinking of?  I've written FT8 and WSPR generation code for the Si5351A that implements Gaussian FSK by updating the Si5351 frequency control at a 2.4 KHz rate (which drops the sidebands significantly):


The only remaining issue is the start and end of transmission key-clicks, but that's not generally a problem.

And there's no need to have such a fast update rate for FST4W, I just wanted to see how far I could push it.