Re: [wsjt-devel] Dial calibration errors

[Bill S. <g4...@cl...>]

On 19/05/2015 05:18, Michael Black wrote:

Hi Mike & All
>
> Thanks Bill….yup…forgot to remove the RR73 from my other 
> experimentation.  Removed the public API too.  So patch is definitely 
> smaller now.
>
I have implemented a change based on your patch. I had to change it a 
bit to make it more robust and also corrected the math errors ;)

It does behave a bit strangely on rigs with less than 1 Hz resolution 
but there's not a great deal we can do that is better for them, at least 
it is benign if the calibration constants are zero.

Currently the intercept error is +/- 10 kHz and the slope factor is +/- 
1000 ppm which is probably enough range.
>
> 73
>
> Mike W9MDB
>
73
Bill
G4WJS.
>
> *From:*Bill Somerville [mailto:g4...@cl...]
> *Sent:* Monday, May 18, 2015 4:34 PM
> *To:* wsj...@li...
> *Subject:* Re: [wsjt-devel] Dial calibration errors
>
> On 18/05/2015 20:10, Michael Black wrote:
> Hi Mike,
>
>     If you mean can you compute it yourself…sure…
>     http://www.miniwebtool.com/slope-intercept-form-calculator/ Just
>     put the frequencies expected in X and what you measure in Y.
>
>     Attached is a patch against 1.6.1 that adds intercept and slope to
>     the Frequencies tab.  You do have to change band after setting as
>     I couldn't quite figure out how to force an update and exactly
>     when that should be done (on exiting dialog?  After entering
>     either number?).
>
>     Might also need some mention that you do either this or the
>     offsets in the table…both are possible but not sure that makes any
>     sense to do.
>
> You can simplify that patch considerably by not adding the linear 
> correction constants to the Configuration public interface. They are 
> only needed within the Configuration class implementation.
>
> Also you probably want to take out the "RR73" change as it is 
> unrelated and probably not intended.
>
> 73
>
> Mike W9MDB
>
> 73
> Bill
> G4WJS.
>
> *From:*Edson W. R. Pereira [mailto:ewp...@gm...]
> *Sent:* Monday, May 18, 2015 2:03 PM
> *To:* WSJT software development
> *Subject:* Re: [wsjt-devel] Dial calibration errors
>
> Hi Mike,
>
> You are correct. I should have mentioned that the same measurement 
> should be done on a different band in order to have the second set of 
> measurements. With band hopping and one calibrated station it may be 
> possible?
>
> 73, Edson PY2SDR
>
>
> ---
>
> - We humans have the capability to do amazing things if we work together.
>
> - Nós seres humanos temos a capacidade de fazer coisas incríveis se 
> trabalharmos juntos.
>
> On Mon, May 18, 2015 at 2:21 PM, Michael Black <mdb...@ya... 
> <mailto:mdb...@ya...>> wrote:
>
> You have to have samples along the entire frequency band to get a fit 
> for slope an intercept.
>
> One freq doesn't do it.  You really only need 2 for a linear fit but 
> more samples gets a bit more accurate.
>
> I've just about got this thing coded up…and you COULD just put a fixed 
> value in A if all you do is one band.   So in your case you could 
> stick a -9 in A and perhaps that's all you need.
>
> Mike W9MDB
>
> *From:*Edson W. R. Pereira [mailto:ewp...@gm... 
> <mailto:ewp...@gm...>]
> *Sent:* Monday, May 18, 2015 12:14 PM
> *To:* WSJT software development
> *Subject:* Re: [wsjt-devel] Dial calibration errors
>
> Hello Joe,
>
> I just checked some of the common received spots between my station 
> and Steve's on 15m and I am 9 Hz off (down). This makes me think that 
> having a standard station in WSPR like Steve's could allow an 
> automatic calibration by performing a query on the wsprnet database 
> for our station and a standard one and compare the results The 
> difference could be converted into the A and B values. Could this work?
>
> 73, Edson PY2SDR
>
>
> ---
>
> - We humans have the capability to do amazing things if we work together.
>
> - Nós seres humanos temos a capacidade de fazer coisas incríveis se 
> trabalharmos juntos.
>
> On Mon, May 18, 2015 at 11:09 AM, Joe Taylor <jo...@pr... 
> <mailto:jo...@pr...>> wrote:
>
> A few more thoughts about dial calibration errors.
>
> A couple of days ago I put my TS-2000 radio through the calibration
> procedure described in Appendix C of the WSPR 4.0 User's Guide
> http://physics.princeton.edu/pulsar/K1JT/WSPR_4.0_User.pdf
>
> The procedure takes about half an hour, end to end, yielding the values
> of two constants A and B.  These constants appear in the equation
>
>    d = A + B*f
>
> where d is the radio's dial error in Hz and f is the received frequency
> in MHz.  In my case the constants are A = 2.14 Hz and B = 1.6254.  I
> note that the value of A has remained constant, but several years ago B
> was somewhat smaller (B = 1.2885 in 2011), so the master oscillator in
> my TS-2000 has aged a bit.
>
>  From the values of A and B I computed the dial error for each amateur
> band and entered those values (expressed in MHz) in the "Offset" column
> of the WSJT-X Settings | Frequencies tab. Frequencies reported for my
> WSPR decodes now agree with those reported by Steve, K9AN, to within 1
> Hz.  (Steve's receiver uses GPS-disciplined oscillators, so his WSPR
> reports are a good standard for comparison.)
>
> It might be handy to permit a user of WSJT-X to enter values for A and B
> and have the program calculate the resulting "Offset" values for the
> Frequencies tab.  The resulting system behavior is very sensible, in my
> opinion.  When WSPRing on 20 meters, for example, my TS-2000 dial now
> reads 14.09562 MHz.  WSJT-X intentionally sets the radio about 24 Hz
> high on  this band, to compensate for its dial error.
>
> The dial frequency displayed on the WSJT-X main window is the corrected
> value, 14.095 600 MHz.
>
>         -- 73, Joe, K1JT
>
>
> On 5/13/2015 10:48 AM, Joe Taylor wrote:
> > Hi all,
> >
> > One of the fun things about WSPR is the frequency accuracies that
> > are involved.  Having WSPR mode in WSJT-X motivates some serious thought
> > about how best to handle frequency calibration errors in transceivers.
> >
> > Typical dial readout errors in modern radios are a few parts per million
> > -- for example, a 20 Hz error at 14 MHz.  For JT65 or JT9 such
> > discrepancies are not very important.  But the WSPR sub-bands in
> > conventional use since 2008 are only 200 Hz wide, and we'd like to use
> > all of that range effectively.  If my transceiver's dial reads 20 Hz
> > low, and yours reads 20 Hz high, and we both set our dials to the
> > conventional 14.0956 MHz for 20 meters, after setting our WSPR Tx
> > frequencies at random within the 200 Hz sub-band there's something like
> > a 20% chance that we won't decode one another.
> >
> > Earlier production versions of WSPR have handled these issues in a
> > rather sophisticated way.  The User's Guide includes detailed
> > instructions for determining calibration constants for your transceiver
> > using over-the-air signals (see Appendix C of
> > http://physics.princeton.edu/pulsar/K1JT/WSPR_3.0_User.pdf ).  The
> > resulting accuracies can be better than 1 Hz.
> >
> > If CAT control is in use and  "Enable frequency correction" is ticked on
> > WSPR's "Advanced Setup" window, frequencies sent to the radio are
> > adjusted so as to compensate for the dial errors.  For example, if
> > 14.0956 MHz has been requested, the command for 14095620 Hz may be sent
> > to the radio.
> >
> > I picture this being implemented in WSJT-X in a similar way.  In this
> > example, the radio dial would read 14.096520.  I'm suggesting that the
> > frequency readout on the WSJT-X screen would read 14.095600, the
> > supposedly "true" frequency.
> >
> > Comments and suggestions would be welcome.
> >
> >       -- 73, Joe, K1JT
>