As we further refine the use of the Kiwi, at KFS I have been comparing frequencies and spreading reported by Bodnar-fed Kiwis with frequencies reported by standard GPS-aided Kiwis at KFS and nearby WD sites.

That investigation led me to realize that the .1 Hz frequency correction added by default to spot frequencies reported by a GPS-aided Kiwi should not be applied to GPSDO (e.g. Bodnar) Kiwis fed with  66.660000 Mhz.  So if you are using a Kiwi modified to use an external GPSDO, set it to  66.6600, enter that value on the Kiwi's Admin->Config-> External clock frequency field, and suppress WD's correction by adding this line to your WD.conf file:

I'm a little unclear about the precise frequency hard coded ito the Kiwi.  I've been using 66.66 MHz as that number though I've recently seen some cause for thinking it may acturally be 66.666 or 66.666666 MHz. The difference between that and 66.66 MHz has little effect in the baseband offset being corrected but maybe somebody here can tell me/us what the 'real' number for zero offset is.  Maybe just an academic question.

Glenn
  It is 66.660000 MHz because the Kiwi's audio decimation is fixed at (505*11) = 5555 and 66660000 divided by 5555 gives 12000 Hz exactly for the audio sample rate, from the calculation by jks at:

forum.kiwisdr.com/index.php?p=/discussion/comment/14977#Comment_14977

Gwyn

Hello Rob

Having added the line to my wsprdaemon.conf:

SPOT_FREQ_ADJ_HZ=+0.0

I then realised that it applies to all of the receivers that wd handles. Here it handles one Kiwi with a mini-Bodnar and one standard aided (primarily so I can monitor the behaviour of both clock sources), and so, in my case, a global value is unfortunately not appropriate? In mixed cases like mine, does the value need to be set for each receiver named in the list in the conf file?

regards and thanks as always

Gwyn G3ZIL 

--

Does this mean the info on the Kiwi itself is wrong ? 

Set when external 66.666600 MHz (nominal)
clock connected to J5 connector/pad.

I defer to Gwyn here but I think the baseband processing has 66.66 MHz hardcoded into its processing and does not use whatever value the user enters for the External ADC clock and the rest of the Kiwi. Thus the RF frequency can be moved around (over/under clocking is possible) but a baseband offset will be introduced if that external clock is not 66.66 MHz.  I think that offsert amounts to about 1.4 Hz if/when the external clock and its setting is 66.60 rather than 66.66 MHz.   But I may be wrong about this...

Glenn's explanation is essentially correct, and is a summary of the explanation set out by John Seamons at

forum.kiwisdr.com/index.php?p=/discussion/comment/14977#Comment_14977

but two of Glenn's numbers are not quite right (the 1.4 Hz should be 0.14 Hz at 1500 Hz baseband frequency, and the 66.60 should be 66.666600).

If you enter 66.666600 MHz as the external clock frequency and set your clock to the same value you will be emulating the nominal crystal clock frequency in the Kiwi itself. The RF section will produce the correct frequency - so the carrier frequency will be correct.

However, the original choice of  66.666600 MHz  for the crystal in the Kiwi was an approximation to the required frequency of 66.660000 MHz for the baseband processing to produce audio at an exact rate of 12000 Hz. The difference amounts to 0.14 Hz at 1500 Hz, the centre of interest for this community. This did not matter a hoot when only 1 Hz frequency resolution was available, but with 0.1 Hz resolution available in the wsprdaemon extended spots database, and the improvements since v1.557 in the Kiwi's own GPS-aided frequency stability when used without a browser the difference of 0.14 Hz became visible when compared with a Kiwi using an external clock set to 66.660000 MHz that provides spot-on audio frequencies.

As Rob wrote, the frequency stability of the standard aided Kiwi is now good enough to see phenomena such as the morning fall in the height of the ionosphere by observing the order 1 Hz Doppler shift. Following the advice in this post ensures that there is no offset between standard aided and GPSDO Kiwis thereby helping those investigations.

Gwyn G3ZIL 

I love these clear explanations ! Thank you all.
So look at the below graph, see the 1 Hz jump/drop. My KiwiSDr is now on 66.66 since a day or so, the site of LX1DQ has 4 Kiwi's with slightly different effects compared to my spots. This one has the smallest frequency difference.
Can the 1 Hz morning fall be seen when comparing 2 different stations in this way ? Or is sunshine falling on one of the Kiwi's creating a quick change in the frequency ?

The remaining small offset in your graph is down to the settings at LX1DQ. If you run the same graph against G3ZIL and receiver G3ZIL_4 that has a mini-Bodnar on 20 m then I am seeing no more than +/-0.1 Hz difference this morning.

As for the morning fall. You can probably see it between two receivers for spots from the same transmitter, but I have not looked at that case yet, just the whole population of spots received at two receivers. I've added a slide with the relevant plots and an extended caption to show the observations so far. 
What needs to be done with this example is to turn the Doppler shift of the JA-> California spots into velocity, integrate to get distance, and see if the result is anywhere near the distance the ionosphere has descended in those three hours with the rise then fall in the Doppler.

Gwyn G3ZIL

Gwyn,

Looking at OZ0RF's transmissions I hope I am correct in seeing 14.0097155 as the lower frequency?
There has always been discussion about exactly which frequency is quoted referring to WSPR and other modes.

73 Alan 4ZFQ

Hello Alan
The frequency you are seeing is correct (the '00' being a typo for '0').
From experience, the Kiwi G3ZIL_4 here with the mini-Bodnar external clock, shows the same frequency as the transmit station has entered.
From the 0.1 Hz resolution data in wsprdaemon_spots_s table, the frequencies for OZ0RF here this morning were as below.

Gwyn G3ZIL

        time         |    freq    

---------------------+------------

 2022-11-18 12:25:00 | 14.0971550

 2022-11-18 12:15:00 | 14.0971551

 2022-11-18 12:05:00 | 14.0971552

 2022-11-18 11:55:00 | 14.0971549

 2022-11-18 11:45:00 | 14.0971556

 2022-11-18 11:35:00 | 14.0971552

 2022-11-18 11:25:00 | 14.0971557

 2022-11-18 11:15:00 | 14.0971554

 2022-11-18 11:05:00 | 14.0971551

 2022-11-18 10:55:00 | 14.0971552

 2022-11-18 10:45:00 | 14.0971556

 2022-11-18 10:35:00 | 14.0971552

 2022-11-18 10:25:00 | 14.0971552

 2022-11-18 10:15:00 | 14.0971552

 2022-11-18 10:05:00 | 14.0971553

 2022-11-18 09:55:00 | 14.0971550

 2022-11-18 09:25:00 | 14.0971551

 2022-11-18 09:15:00 | 14.0971555

 2022-11-18 09:05:00 | 14.0971554

 2022-11-18 08:55:00 | 14.0971555

 2022-11-18 08:45:00 | 14.0971554

Hello

The RFzero WSPR frequency is correct in the library since release 0.8.6 from 2019-12-03. This follows an investigation done by Warren, WA8TOD, and response from Steven, K9AN, forwarding an email from Joe, K1JT, saying in short:

"The case for WSPR is rather different, though. The four tones are closely spaced and have equal weighting. It doesn't make much sense to use one convention for the frequency of received signals (the center of the four tones) and another convention for the transmitted signal (that of the lowest tone). It seems that I should move the transmitted tones lower by 1.5*12000/8192 = 2.197 Hz. I will do so."

However, if FST4W uses the same way, I don't know, but I have implemented it in the same way.

While we speak of RFzero the last of the missing component is to arrive next week. So I am still confident that units will be produced in the beginning of December.

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

On 18/11/2022 12:44, Gwyn Griffiths via groups.io wrote:

The frequency you are seeing is correct (the '00' being a typo for '0').

Gwyn, Bo,

Thanks, I was surprised to see the '00'.

I was playing with Spectrum laboratory and my AOR7030 with a cheap u-blox GPS for a marker. Unless I am wrong I see the lowest tone at the frequency that is reported. 14.097155

I'm sorry but I still found Bo's quote of Joe's comment "It seems that I should move the transmitted tones lower" confusing.

I noticed about 0.5 Hz rise from 15.15 to 15.40 on the 17th, not too easy to see when mixed with WSPR. I have not yet found how to access that on the the wsprdaemon database.

I see a gap in reception this morning 06.40-07.40, almost looks as if propagation had continued the shift would have been higher than the 1Hz at 07.45

73 Alan G4ZFQ

The WSPR math/spec is:

tone0 = f - 1,5 * sample rate / samples
tone1 = f - 0,5 * sample rate / samples
tone2 = f + 0,5 * sample rate / samples
tone3 = f + 1,5 * sample rate / samples

e.g. using the same spec for FST4W-300, where the sample rate is 12 kHz and the samples are 21 504:

tone0 = f - 1,5 * 12000 / 21504 = f - 0,837 Hz

for WSPR tone 0 would be = f - 2,197 Hz.

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

Hello Alan

Please do send me a private email to see if I can help you with what you'd like to get from the database / Grafana.
"I have not yet found how to access that on the the wsprdaemon database".

I could understand Bo's quote from Joe, "It seems that I should move the transmitted tones lower" if there was a typo and 'transmitted' should read 'received'.

Gwyn G3ZIL

https://sourceforge.net/p/wsjt/mailman/message/34135506/

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

On 19/11/2022 12:38, Bo, OZ2M wrote:

https://sourceforge.net/p/wsjt/mailman/message/34135506/

Bo,

I am still confused, quite normal:-)
Can we agree that the normal uploaded frequency from wd is that of the lowest RF frequency transmitted?
14.09755 without any shift?

Gwyn, thanks, I'll be in touch.

73 Alan G4ZFQ

Hello Alan

Well, the intended nominal transmitting frequency of OZ0RF is 7 040 156 Hz and 14 097 156 Hz this the respective tone0 are

40 m: 7 040 155,163 Hz
20 m: 14 097 155,163 Hz

unless there is a bug somewhere I haven't found.

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

Bo, Alan

Over the last day I make the average frequency from OZ0RF as follows, using same notation and with a mini-Bodnar Kiwi:

20 m: 14 097 155,224 Hz (difference from Bo's expectation is + 0.061 Hz)

I do not have a Bodnar on the Kiwi that covers 40 m.

Gwyn