and now for something completely different? 😁


I was stimulated to think about ways to generate multiple-band FST4W and other precise digital modes with low spreading and (relatively) economic hardware for use during the upcoming US solar eclipse studies by HamSci.

It occurs to me that we may already have many of the pieces working well and it might (almost) be possible to have working HW in time for the event.

How about if we take a [10m] QDX, join it with the existing N6GN reference, modified only to use the B version of the Si5351 (which is available for around US$5) and add to those passive mixers, a bit of filtering and some 5W PA modules that Clint, KA7OEI has found, or else something similar if those prove to be too expensive.   Filtering would be nearly free and the main additional cost would be for the mixers and PA modules.  It would have to be mounted on a heat sink capable of dissipating 50W - not too different from a VHF PA.

It seems to me that we could have simultaneous (or not) transmission on any LF-6m bands that the PAs could handle - each band transmitting identical QDX content.

$80 QDX + ~$50Reference + ($20 conversion IF )   + $50?(SMAconnectors, fab PCB+assembly + shipping) + (#bands*$(mixer+PA))

The last part cost still TBD but perhaps the whole think on the order of the original KiwiSDR cost.

It could generate any band through 6m and any digital mode precisely and at WSPR-like power so might appeal to different goals and uses.

How useful would this be and is that too expensive to consider?

File attached but I don't know if it will get through to the group so I'll resend it if this attachement doesn't work.


Glenn n6gn

Sorry for the previous pdf. I didn't mean to include a mangled block of the Kiwi 0-2 GHz converter as page 2.  Attached is a somewhat cleaner version of the thought I was trying to convey.The attached may be a little clearer.

Based on my experiments last week I would like to offer 2 alternatives to this approach.

1. Use a commercial transceiver for the event. Normally we would not like to use our main equipment for running a beacon, but for a few days only why not. I have shown that my Yaesu FT817nd is perfectly capable of delivering FST4W transmissions with very low spread on all bands and up to 5 W. Extra cost ? No.

2. Buy a Hermes Lite v2. I have shown that the Hermes is perfectly capable of delivering FST4W transmissions with very low spread on all HF bands and up to 5 W. Cost US$ 269 + 52,7 + 16,9 + shipping + power supply. 

A Home build parallel transmitter would be interesting though as I have 5 AD9850 boards lying around that could be used as a cheapo si5153 but with a lower max frequency?

An interesting software setup for transmitting WSPR or FST4W (not multi frequency-parallel) I found :

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

Erwin

From a brief scan of the code, unfortunately this seems to be only a FLL not a PLL,

//***********************************************************************
// This interrupt is used for Si5351 25MHz crystal frequency calibration
// Called every second by from the GPS 1PPS to Nano pin D2
//***********************************************************************

It's only frequency correction applied once/second so the oscillator must free-run in between.  This means that the short term performance and spreading performance for FST4W will be that of the XO rather than phaselocked to the GPS constellation and will not provide results as  good as fully PLL solutions.

Which would be the same "GPS-aided" solution as the KiwiSDR is using for reception. 

The Si5351 could be fed from an external reference clock to improve things. Instead of using this GPS-aiding code.

Like Harry from Zachtek explains here :
https://www.zachtek.com/post/2019/02/14/using-the-si5351-directly-with-a-10mhz-ocxo-for-increased-stability

I happen to have the QRPlabs OCXO and of course your Reference board (and a whole bunch of QRPlabs bandpassfilters for the output stage)

TBC.

Hi

Clocking the Si5351A beyond 25 MHz to 27 MHz means worse spectrum performance. On the other hand most Si5351A fans don't care about anything else than the fundamental frequency and the harmonics. I know that I am a bit of a hardliner, when it comes to the spectrum: https://tinyurl.com/smw6kyt9 but spurious in- and out-band and phase noise do matter, and they are much more difficult to do reduce than the harmonics are. Those who program Raspberry Pies for RF use should only be allowed to use a keyboard wearing boxing gloves.

Using an OCXO and an Si5351A seems unbalanced to me. A bit like the IC-706 syndrome: simple and cheap initially but adding the accessories that most do, will result in a cost that is higher than a transceiver with it all built in from the beginning, e.g. IC-756.

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

I agree with you Bo, though with perhaps not quite the same emphases.  For the interesting investigations that seem to be springing forth from HF FST4W operation and analysis, I think there are several characteristics that need to be addressed which other amateur radio products and applications  might not see as so important

• much better than 1 ppb long term frequency accuracy
• excellent close-to-the-carrier spectral purity to produce system spreading numbers that are insignificant compared to any measurement. Probably small mHz/symbol time LF-HF
• in-amateur--band purity that at a 5W tx level puts total unwanted in a SSB bw a long way below carrier. I don't know what that number should be but 1 microwatt would  have it 50 dB down or more.
• Out-of-band purity at perhaps 60 dB below carrier, one can argue about this value but it should be small enough not to be noticed by anyone.
• low system cost. If this is going to be attractive to the large base of  users cost is a big deal. This is a constant tension with performance.
• turn key operation both of HW and SW involved in operating it, not more difficult than WSJT-X currently is. Users shouldn't have to understand the innermost details of what the system is doing.
  
• support, parts obsolescence management
•  ...

1 pps GPS-aided solutions aren't adequate. I think GPS constellation or precision shack references are necessary.

The cost driver makes me think that including an OCXO with a Si5351 solution is a mismatch as you suggest GPS-aiding probably doesn't have enough performance with inexpensive TCXOs for some of these measurements/uses.  It *is* possible to get the kind of performance necessary from the GPS constellation and inexpensive GNSS modules with Si products. Bodnar generally does it with the Si5328 and the reference I made is well on the way, though I think I might have to sign an NDA with SkyWorks about the Si5351 to get all the inner secrets that Leo has appropriated.  At present, my solution is not as good as I'd like when using GPSDO/PLL at 23cm as it is with local 10 MHz HP OCXO. With that external 10 MHz instead of GPS constellation I'm able to measure over-the-air ATSC TV pilot carriers from 100 km distance over very much non-LOS path to a few mHz at 186 MHz with good spectral purity and fairly low standard deviation. Surprising to me that although I live LOS so can see  the tower lights of the 7 NIST transmitters of WWV,  a distant VHF frequency reference might actually be as good!

All together, this is a challenge for an amateur radio solution, how to make available suitably high performance transmitters with good spectral characteristics at a low enough price point to meet these goals.  Picking the architecture, fabrication methods and distribution (marketing?) requires a balance of expertise and skill in many areas.  I don't claim to have all those skills!  It will be interesting to see if Gwyn and Rob's presentations yesterday raises interest in all these challenges within a wider audience.

Glenn n6gn

Bo, re the 10MHz vs 25MHz reference for the Si5351, is the spectrum performance due to the lower reference freq leading to greater jitter in the VCO?  I'm using 10 MHz for my '5351 projects and it's been good enough, but I haven't done the comparison.  I will be doing that soon though.

Glenn, what Si5351 secrets do you think there might be?

For an inexpensive 1 Hz FLL you might consider using one of the recently-available surplus 10MHZ OCXOs (https://www.ebay.com/itm/394139585347).  I just got two of these ($19 ea), and they could be easily modified to take an external tuning voltage.input.  Of course this isn't a solution for a long-term project, as we don't know how long these will be available and the form-factor may not be suitable.  But for hacking something together cheaply, they look pretty good.  On mine, I had to change a tuning resistor as the OCXO had aged beyond the available trim-range.  But now that they are dialed-in, they both look quite stable (relative to the Bodnar GPSDO).

But yes, using one of the GPS modules that has a higher-frequency output that 1 PPS is probably still part of a cheaper reference solution.

On Sat, Mar 18, 2023 at 10:42 AM, Paul WB6CXC (tech-blog: wb6cxc.com) wrote:

But yes, using one of the GPS modules that has a higher-frequency output that than 1 PPS is probably still part of a cheaper reference solution.

[typo corrected -- I wish we could edit after posting!]

I'm not sure if this thread is too low level for the group. If any find it annoying please speak up and we can take it off group.

Generally speaking uBlox GNSS modules have a 'magic number' related to their inner architecture for LTE systems and a 30.72 MHz global/system clock.  The RF synth and the DSP inside these things seems to run from two different clocks, which if not synchronous cause low rate glitches in the resulting timepulses, at least at higher rates which are of interest to full GPS PLL solutions like Bodnar or my board.  From my source code

const unsigned long UBLOX_MAGIC_FREQ = 1171875UL;   //  uBlox set from this for low-glitch GPS tp, 2500*(30.72^6/2^14)

and powers of 2 relative to that, 4.6875 MHz being one useful possibility that Bodnar uses often.

Also there is deeper mystery here.  Bodnar sends undocumented commands in set up of the Series 8 modules.  This is the class related to 'Firmware Update'  and may be some special code to further improve cleanliness of the timepulse. I haven't yet decoded or plumbed the depths of this magic. Without it applied, my output spectrumn at low microwave is not as clean as I'd like and probably will add ~100 mHz of spread at 1296.5 MHz.  At HF it is not significant.

Bodnar uses the Si5328 with more built-in PLL features than the Si5351. It also costs almost 20X as much and makes it a non-starter for my goals of total component cost for GPSDO (PLL) solution of US$25 or so.  Low end Chinese GNSS modules using uBlox chipsets are under $7 each and a Nano controller instead of the (web server capable) IoT33 I've been using is $2 while an RPI-ZeroW with Wifi is about $6 but takes a bunch of recoding and probably some circuit changes.

Putting an eBay (part availability problem in the future?) at almost the entire parts budget is a non-starter.  I built something like that a decade ago for VHF/UHF WSPR, also with external PLLs and eBay OCXO (pulls using a blow torch to get parts off of old cell site HW, I think) which can be seen on the WSPR 2m+ group page.  I don't want to go there again but would like to find a very inexpensive and sustainable solution. 

I think we should not even consider 1 PPS FLLs. 

Glenn

Coincidently with Pauls ebay buy, I ordered this from my friend Ali yesterday
To play with for Eur 16,-- including PP. Just doing archeology in Glenn's past ;-) it seems

I have the other version with second output with selectable output via divider.  Pulled pot off and provided voltage via PLL circuit. 

Erwin, I've seen two different schematics for that board, each with a slightly different configuration for the bias pot and surrounding resistors.  I had to change out one resistor for the trimmer to be able to pull the OCXO to frequency.  I can post both schematics if you like (they both say they are for board rev 002),

That would be very helpful Paul, please do. Thanks

@glenn
What would be candidates for cheap GPS constellation setups?
I found this https://www.e3s-conferences.org/articles/e3sconf/pdf/2021/86/e3sconf_spatialdata2021_03001.pdf
which might be helpful but does not hint at cost.
You mention  Low end Chinese GNSS modules using uBlox chipsets are under $7 each, at what detailed information should we look to find the useable ones ? There are ones claiming to have uBlox NEO-8m chips that turn out to be not being "real".
Maybe this firm could sponsor the (HamSci) effort : https://www.mulberrytec.com/gnss-gps-module/

Hans Summers discussing the Si5351 here : https://groups.io/g/BITX20/topic/si5351a_facts_and_myths/5430607
Long time ago but still instructional. On phase noise, higher frequencies etc.

I have a version of the reference board in fab right now that is laid out to accept either  the MAX form factor like that of the uBlox 10 series I've been using or alternatively  the larger GPS02-uBlox Chinese version that claims to have the timepulse capability and to use a uBlox chipset.  I expect that back from fab in 1-2 weeks. I have two of the <$7 versions in hand.

Glenn

Here are the two surplus OCXO schematics I've found.  The ones I have match the top schematic.  I replaced R3 (36K) with a 3.3K resistor to give me adequate trim range..

Built this this morning.  Output, 5, 10, 25 MHz. GPS top right under the board.  Used scrap copper and CNC to build the test up. 

My Ali OCXO was delivered yesterday. Measured a bit today. Works perfect. No need to trim based on the SA measurement. Will do a tinyPFA later as that is more precise.