Change storage Image rejection for mode to Image rejection for frequency

[PA3CCE]

I like the new IQ balance adjustment,
now I can turn the image completely into the noise !

In the menu, storing of the IQ balance settings is provided
for the different modes LSB, USB etc.

In my case, I find no- or neglectable differences in
settings between the modes, but quite some difference
in the settings over the entire frequency range.
If confirmed, I should like to have the IQ balance in the
menu per mode, changed to settings per frequency
(Low-, medium-, High Frequency, or even better per band).

Jos (PA3CCE)

[df8oe]

Hello Jos,

this is a very good idea. When I first adjusted IQ with the new functions I had to increase power because I could not measure unwanted signal! And so I stated that there is a dependency to power, too. If you adjust it @0.5W and signal is completely gone, you switch to 5W and adjust that signal again is completely gone, then switch back to 0.5W you will see there is a slight misadjustment.

But we are talking about things which before nobody was able to see :)

We will think about improved structure, freuency dependency included. It is not high priority because now its working for the first time, but "good things can be improved" of course :)

vy 73
Andreas

[PA3CCE]

Well, I don't find the present situation real "good", when set at mid frequency,
the image rejction at low- and high-frequencies are rather "poor"

73 de
Jos

[df8oe]

...but much better than all the years before :) So I call it 100% "good". Hundreds of users have lived without a working IQ-adjustment and nearly nobody has complained before. But now, if it works - it is not good ??

Sorry - I cannot understand.

vy 73
Andreas

[PA3CCE]

A change from not working at all to 100% working for just one frequency
is huge.
Working for the entire frequency range is much better.
I think that many users just calibrate for one frequency and think " OK
fine".
They seem not to be aware what happens at other frequencies.
Besides, I never heard you say before that the image rejection was not
working at all,
or did I miss that ?

Best 73 de
Jos

Op 8-4-2016 om 11:06 schreef eeDF8OE:

...but /much/ better than all ther years before :) So I call it 100%
"good". Hundreds of users have lived without a working IQ-adjustment
and nearly nobody has complained before. But now, if it works - it is
/not good/ ??

Sorry - I cannot understand.

vy 73
Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[df8oe]

Hello Jos,

haven't you read many posts in the past in Yahoo-NG that IQ-phase adjust does not work?? These posts go back to the time before I joined the project in May 2015... So the best adjustment in the past was not so good as if you adjust it with actual firmware and go to the band with the worst rejection. Because of it was not working properly before many users never have had adjusted their iq settings.

Additional problem may (again) be we reach edge of mcHF. The problem results of unsymmetrical capacities / inductivities at i and q LO signal line. I have already tested and they depend on many things like voltage of power supply, actual output etc. so it is possible that it is impossible to get all these dependencies included.

vy 73
Andreas

[PA3CCE]

Op 8-4-2016 om 16:06 schreef DF8OE:

Hello Jos,

haven't you read many posts in the past in Yahoo-NG that IQ-phase
adjust does not work?? These posts go back to the time before I joined
the project in May 2015... So the best adjustment in the past was not
so good as if you adjust it with actual firmware and go to the band
with the worst rejection. Because of it was not working properly
before many users never have had adjusted their iq settings.

I agree, I also complained in the past.
With me I could adjust the amplitude, but the phase adjustment did not
react at all.
However, this is'nt "not working at all".
To my opinion this is not to be used as a reference to declare the
present situation as "good".

Additional problem may (again) be we reach edge of mcHF. The problem
results of unsymmetrical capacities / inductivities at i and q LO
signal line. I have already tested and they depend on many things like
voltage of power supply, actual output etc. so it is possible that it
is impossible to get all these dependencies included.

Sure it will never be perfect, but what I proposed might be an
acceptable approach.
Do not underestimate the importance of the best image rejection in Tx-
as well in Rx-mode.

73 de
Jos

vy 73

Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[DD4WH]

Hi Jos, hi Andreas,
my impression is, we have all fully acknowledged that we need a smart solution for the dependencies of the phase adjustment in order to get the very best out of our hardware. But we should also be smart in using a slim solution, if it is feasible and appropriate. I am not sure whether a phase adjustment for every band would achieve this (and would be necessary), but we would probably need more data on that in order to be able to decide.

Could you two state the amount of sideband suppression that is lost when you adjust phase & amplitude perfectly for one band and then change to the band with the worst suppression? Is it in the range of 2-5dB or in the range of 10-20dB? And please also state whether your measurement is for TX or RX.

Theoretically, (for TX) if we can adjust sideband suppression to an optimum of 55dB for one band and this gets worse for another band in an order of 6dB, well, then don´t let us talk about that again . . ., that´s still 49dB of suppression (10W -49dB equals 126µW, which is acceptable in my view).

For RX, with listening to DX signals and a local 100W station at 12kHz of your frequency (if you use the 12k IF), it could be nice to have 6dB more suppresion . . . however, the best hardware phasing receivers hardly ever achieve more than 45dB of sideband suppression (in the best band!).

Let us look into the HARDWARE signal I&Q chain, this is different for TX and RX:
RX:
C68/C69 (22nF)--> they should be matched! I did that when I built the mcHF and I do not have such a big improvement with the new phase adjustment as others, maybe because of the matching, dont know!?
C71/C73 --> must be 0R !
C26/C31 (470nF) --> maybe matching helps? Have not tried yet
C18/C19 (1µF) --> maybe matching helps? Have not tried yet
--> For RX, it should be easy to match the three pairs and have very good sideband suppression with the hardware alone

TX:
C7a/C8a (470nF)
C7/C8 (470nF)
C109-C112 (10µF)
C114-C120 (100µF)
C84-C87 (22nF) --> at least these must be matched!
--> There are not so many capacitors and no inductor in the I&Q signal chain. So, I do not see that it would not be possible to improve a lot on the hardware side by matching the capacitors in the I&Q TX line.

Question 1: Maybe this also helps to decrease the frequency dependence of the phase adjustment?

Question 2: Is it necessary to distinguish between phase adjustment for LSB & USB, AM & FM?
--> At least theoretically, I do not see any reason for this: All the demodulation steps occur in software. So, if we adjust I& Q phase and amplitude before demodulation (as is the case in the moment), this adjustment should be the same and should hold for ALL demodulation modes.
(It would be good, however, if that could be supported by your observations!)

This could free up EEPROM and menu space for frequency dependent adjustments!? --> Just a thought . . .

added later:
As comparisons of the mcHF and the KX3 seem fashionable at the moment (;-)), it is worth noting that the KX3 also has separate IQ phase adjustments entries for EACH band.

73 de Frank DD4WH

[PA3CCE]

Hi Frank,

First of all I must tell you that at this moment I cannot perform any
test, since
today my mchf stopt working. Something is wrong in my LO.
At times the frequency display is in red color and the spetrum display is
interrupted in abt 1 sec rate.
When I change frequency, the spectrum display stops inmediately.
Could not find the cause so far.

I also can tell you that I have made all modifications as they are in
the circuits
of IZ6MAF, so all unnecessary capacitors are shorted etc.
I did the tests in Rx mode only.
Further more I had matched the appropriate components in the demod circuit.

I can give you the adjustment figures below :

Freq LSB USB
ampl/phase ampl/phase

3MHz 5/2 5/1
5MHz 11/5 11/4
7MHz 17/8 18/7
10MHz 19/14 20/13
14MHz 30/22 28/22
21MHz 50/38 54/37
24MHz 58/52 58/50
28MHz 63/65 63/65

These are the settings I found in the menu (all positive).
You see that the difference between the demod modes are neglectable,
this was the reason I did not test the other modes.
You also can see large differences per frequency, in phase as well in
amplitude.
I did not write down the differences in image rejection but, since you
know the
change per step, you can have an idea.

As soon as I have mcHF working again I can profide you more info.
I hope this helps.

best 73 de
Jos

Op 8-4-2016 om 19:43 schreef DD4WH:

Hi Jos, hi Andreas,
my impression is, we have all fully acknowledged that we need a smart
solution for the dependencies of the phase adjustment in order to get
the very best out of our hardware. But we should also be smart in
using a slim solution, if it is feasible and appropriate. I am not
sure whether a phase adjustment for every band would achieve this (and
would be necessary), but we would probably need more data on that in
order to be able to decide.

Could you two state the amount of sideband suppression that is lost
when you adjust phase & amplitude perfectly for one band and then
change to the band with the worst suppression? Is it in the range of
2-5dB or in the range of 10-20dB? And please also state whether your
measurement is for TX or RX.

Theoretically, (for TX) if we can adjust sideband suppression to an
optimum of 55dB for one band and this gets worse for another band in
an order of 6dB, well, then don´t let us talk about that again . . .,
that´s still 49dB of suppression (10W -49dB equals 126µW, which is
acceptable in my view).

For RX, with listening to DX signals and a local 100W station at 12kHz
of your frequency (if you use the 12k IF), it could be nice to have
6dB more suppresion . . . however, the best hardware phasing receivers
hardly ever achieve more than 45dB of sideband suppression (in the
best band!).

Let us look into the HARDWARE signal I&Q chain, this is different for
TX and RX:
RX:
C68/C69 (22nF)--> they should be matched! I did that when I built the
mcHF and I do not have such a big improvement with the new phase
adjustment as others, maybe because of the matching, dont know!?

C71/C73 --> must be 0R !
C26/C31 (470nF) --> maybe matching helps? Have not tried yet
C18/C19 (1µF) --> maybe matching helps? Have not tried yet
--> For RX, it should be easy to match the three pairs and have very
good sideband suppression with the hardware alone

TX:
C7a/C8a (470nF)
C7/C8 (470nF)
C109-C112 (10µF)
C114-C120 (100µF)
C84-C87 (22nF) --> at least these must be matched!
--> There are not so many capacitors and no inductor in the I&Q signal
chain. So, I do not see that it would not be possible to improve a lot
on the hardware side by matching the capacitors in the I&Q TX line.

Question 1: Maybe this also helps to decrease the frequency dependence
of the phase adjustment?

Question 2: Is it necessary to distinguish between phase adjustment
for LSB & USB, AM & FM?
--> At least theoretically, I do not see any reason for this: All the
demodulation steps occur in software. So, if we adjust I& Q phase and
amplitude before demodulation (as is the case in the moment), this
adjustment should be the same and should hold for ALL demodulation modes.
(It would be good, however, if that could be supported by your
observations!)

This could free up EEPROM and menu space for frequency dependent
adjustments!? --> Just a thought . . .

73 de Frank DD4WH

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[DD4WH]

Hi Jos,

thanks a lot for your data. That helps, also it would be very good to also have settings translated into dB of suppression ;-) I altered the setting of the finetuning a few versions of firmware ago, so a difference of 5 to 63 is not so big anymore as it was in former times ;-)

I had the same problem with red digits and it returns from time to time. Check soldering of the Si570 and around that, very likely that the problem is there, but Andreas probably knows more.
From your data, I would conclude, we can stop distinguishing between modes in the phase and amplitude adjustment! Also, if you have matched capacitors in the Rx chain, it would have to be concluded that this does not help against the frequency dependance of the adjustments! Very interesting and valuable results. I will think about that and wait for more info from other users.

Thanks a lot for your info and good luck in fixing your rig! If it is ok, I would appreciate very much more info, eg. from TX mode and other measurements!

73 de Frank

[df8oe]

Hello Jos,

that is very interesting! My settings differ only +/- 2 (both phase and gain) between optimum and worst band.

Why does yours differ so much? Now I understnad your opinion - but I think, you do have a hardware problem. All of my 4 mcHFs differ only +/- 2...

y 73
Andreas

[PA3CCE]

Hello Andreas

+/- 2 difference over the entire frequency range, are you sure ?
That is almost ideal, I can hardly believe that.
I hope my mcHF will be operable soon, I ordered a new RF board,
the old one is, due to many experiments, worn out.

best 73 de
Jos

8-4-2016 om 23:58 schreef DF8OE:

Hello Jos,

that is very interesting! My settings differ only +/- 2 (both phase
and gain) between optimum and worst band.

Why does yours differ so much? Now I understnad your opinion - but I
think, you do have a hardware problem. All of my 4 mcHFs differ only
+/- 2...

y 73
Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[df8oe]

Yes. I have one mcHF built as RX-only and three mcHFs built as transceivers.

but I have made a modification on all of these which is not documented yet - maybe that is the reason??!!

I kicked off drivers U12 and U13 on rf-board and swapped R22...R25 with 0R. Drivers are producing more phase shifting at higher frequencies and in my eyes are not neccessary (and maybe they do disturb more than they are useful). But I have not made intensive measuring. Of course this can be the reason for the great differences between your mcHF and mine!

Must be investigated... But I am only one person and have to make business, too , so I do not have much time for all of this. McHF takes already big part of my time :)

vy 73
Andreas

[DD4WH]

Very impressive differences! Hmmm . . .

Andreas, is that in RX or TX or both? Are the differences between modes also so small?

Would love to try your mod, did you solder out the SN74LVC2G17s or did you just solder two wires on top of them to bridge the buffers?

73 de Frank

P.S.: Unfortunately, I am not able to measure myself, because I have no stable RF generator with such a high drive . . .

[df8oe]

You must desolder little drivers and connect in/out (2x at each driver) with wire. My settings are in a small range at RX and TX.

vy 73
Andreas

[PA3CCE]

@ Andreas, Frank,

OK Andreas that is most interesting information, thank you.
As a matter of fact, the new RF board I ordered is of version 0.5.
In this version these drivers are rejected indeed and the traces
of the LO are matched for impedance, as described by Chris.
As soon I have the results, I will let you both know.

Have a nice weekend,
Jos

Op 9-4-2016 om 14:18 schreef DF8OE:

Yes. I have one mcHF built as RX-only and three mcHFs built as
transceivers.

but I have made a modification on all of these which is not documented
yet - maybe that is the reason??!!

I kicked off drivers U12 and U13 on rf-board and swapped R22...R25
with 0R. Drivers are producing more phase shifting at higher
frequencies and in my eyes are not neccessary (and maybe they do
disturb more than they are useful). But I have not made intensive
measuring. Of course this can be the reason for the great differences
between your mcHF and mine!

Must be investigated... But I am only one person and have to make
business, too , so I do not have much time for all of this. McHF takes
already big part of my time :)

vy 73
Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[df8oe]

Hello Jos,

thank you for this information. That will be an explanation... Additional I have used 7474 manufactured by "Potatoe Chips" which has anupper working frequency of 1GHz. This will become eccessary when you will add the rf-mod-pcb so that 6m and 4m are available, too. Maybe timing at higher frequencies is more precise like on normal logic IC...

Your feedback after getting new pcbs is welcome!

Have a nice weekend, too
Andreas

[PA3CCE]

Hi Andreas

Thanks again, I must have some "Potatoe Chips" circuits somewhere too,
so I will see.
I hope and expect theat I can keep using the UI version 0.4, for I just
ordered the RF board.

Cheers
Jos

Op 9-4-2016 om 16:11 schreef DF8OE:

Hello Jos,

thank you for this information. That will be an explanation...
Additional I have used 7474 manufactured by "Potatoe Chips" which has
anupper working frequency of 1GHz. This will become eccessary when you
will add the rf-mod-pcb so that 6m and 4m are available, too. Maybe
timing at higher frequencies is more precise like on normal logic IC...

Your feedback after getting new pcbs is welcome!

Have a nice weekend, too
Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[DD4WH]

Very interesting, these 7474 chips! Did you also use the faster potatoe version of the 3253 as substitutes for the 3253 Muxer? The potatoe version seems to have much larger on resistance (17-22ohms vs 3-6ohms), could potentially be a problem!

73 de Frank

[DD4WH]

I could now reactivate my Dipmeter and an RF amp and an attenuator, so I had an RF source at hand which produced a huge RF signal with -17dBm. Here are my settings for optimum IQ amplitude and phase adjustment setting for specific frequencies:
(frequency, (ampl., phase for LSB), (ampl., phase for USB))

2.5MHz -14/-20 LSB -14/-20 USB
4MHz -13/-19 LSB -12/-19 USB
8MHz -13/-20 LSB -13/-20 USB
12MHz -20/-34 LSB -23/-23 USB
16MHz -16/-30 LSB -16/-27 USB
21MHz -24/-31 LSB -25/-29 USB
24MHz -34/-49 LSB -35/-44 USB
28MHz -37/-55 LSB -38/-46 USB

side observation: at 24MHz and 28MHz the nulling was dependent on where I put my fingers on the mcHF (which has no shielding or metal case in my mcHF)!!! That could also be one of the reasons for your varying results, Jos!?

My conclusions:
1.) we do not need to distinguish between LSB, USB, AM, SAM, FM in the amplitude & phase adjustments --> software can be simplified
2.) the adjustment is more or less frequency dependent (depends on your build/your measurement situation) --> whether we need a band-specific possibility to adjust, is not clear yet, because in my case (see above), it would really be overkill . . . but in others (Jos!) it could be nice to have.
3.) The frequency dependance is probably 100% the result of the buffer ICs U9, U12, and U13 (if you compare Andreas´ results with Jos´ and my results). The buffer ICs U9, U12, U13 and corresponding Rs R19, R22-R25 should be eliminated as a suggested modification.

Comments and more measurement data very welcome!

73 de Frank

[df8oe]

I do not have swapped mixers (due to quite different Ron what will be a problem). But I have tested nd they are usable up to 100MHz - they will work with 6m and 4m od, too.

vy73
Andreas

[PA3CCE]

Thanks, see below.
Jos

Op 10-4-2016 om 13:09 schreef DD4WH:

I could now reactivate my Dipmeter and an RF amp and an attenuator, so
I had an RF source at hand which produced a huge RF signal with
-17dBm. Here are my settings for optimum IQ amplitude and phase
adjustment setting for specific frequencies:
(frequency, (ampl., phase for LSB), (ampl., phase for USB))

2.5MHz -14/-20 LSB -14/-20 USB
4MHz -13/-19 LSB -12/-19 USB
8MHz -13/-20 LSB -13/-20 USB
12MHz -20/-34 LSB -23/-23 USB
16MHz -16/-30 LSB -16/-27 USB
21MHz -24/-31 LSB -25/-29 USB
24MHz -34/-49 LSB -35/-44 USB
28MHz -37/-55 LSB -38/-46 USB

side observation: at 24MHz and 28MHz the nulling was dependent on
where I put my fingers on the mcHF (which has no shielding or metal
case in my mcHF)!!! That could also be one of the reasons for your
varying results, Jos!?

Jos :
I had a shield between RF board and UI board, but no metal case.
Observed NO sensitivity to fingers placed on the mcHF, measurements were
stable did not look for instabilities.
My measurements were executed at approx. -40 dBm at antenna in.
Had a RF preamp with PGA103+.

My conclusions:
1.) we do not need to distinguish between LSB, USB, AM, SAM, FM in the
amplitude & phase adjustments --> software can be simplified

Jos:
Agreed

2.) the adjustment is more or less frequency dependent (depends on
your build/your measurement situation) --> whether we need a
band-specific possibility to adjust, is not clear yet, because in my
case (see above), it would really be overkill . . . but in others
(Jos!) it could be nice to have.

Jos:
Overkill ? , I don't know, what was your difference in image rejection
between lowest/highest band when adjusted at mid band?
Sorry I did not note down , as said I could adjust deep into the noise,
so rejection was >>-60 dB.
I prefer to have highest rejection as possible to avoid all kind of
interference.

3.) The frequency dependance is probably 100% the result of the buffer
ICs U9, U12, and U13 (if you compare Andreas´ results with Jos´ and my
results). The buffer ICs U9, U12, U13 and corresponding Rs R19,
R22-R25 should be eliminated as a suggested modification.

Jos:
U12 and U13 were still in, corresponding Rs were out.
I'm anxious to see the results of version 0.5 RF board.
However, results may be dependent of many variables.

Comments and more measurement data very welcome!

Jos:
Will do so when new version 0.5 RF board has arrived and built.
When the LO modification solves this problem (more or less),
there still must be a solution for version 0.4 users.

73 de Frank

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[DD4WH]

I just soldered out U12 and U13 and put in 0R for R22-R25. So these buffers are now out of the I&Q path:
Did the same measurements again for RX:

2.5MHz -10/-22 LSB -7/-22 USB
4MHz -11/-24 LSB -4/-23 USB
8MHz -3/-29 LSB 2/-28 USB
12MHz 2/-33 LSB 5/-33 USB
16MHz 11/-31 LSB 14/-31 USB
21MHz 4/-44 LSB 7/-43 USB
24MHz 5/-72 LSB 7/-72 USB
28MHz 33/-56 LSB 35/-56 USB

So, what would you conclude from this?
Comments welcome! ;-)
73 de Frank

[DD4WH]

I would conclude that my measurements at 24/28MHz probably are not reliable, because the (very strong) signal of -17dBm seems to enter the mcHF over additional paths . . .
However the other measurements seem to indicate the following:

@2.) we need a frequency dependent I&Q amplitude and phase adjustment. I fully agree with you, Jos, now with those measurements. A difference in adjustment of +-3 can result in a difference in mirror rejection of >20dB (not measured, but a sophisticated guess by me). Sometimes +2 steps turns absolute silence to a rather loud annoying heterodyne.

@3.) my hypothesis must be rejected: the buffers are not the problem! New hypothesis: the 74LCX74M (U11) is the problem, that chip seems to have a huge frequency dependence in handling amplitudes and phase. I will order the "wonder chip" (7474) by Potatoe chips in order to validate that.
Andreas, could you confirm, it is this one, please:
http://www.ebay.com/itm/7474-G-Series-GHz-TTL-CMOS-logic-IC-14pin-SOIC-QTY-1-/330551715157

73 de Frank

[df8oe]

YES - that is the one I am using.

vy 73
Andreas

[PA3CCE]

Op 10-4-2016 om 14:39 schreef DD4WH:

I would conclude that my measurements at 24/28MHz probably are not
reliable, because the (very strong) signal of -17dBm seems to enter
the mcHF over additional paths . . .

Might be, but see my findings also.
Can't you make a simple attenuator of say 20dB or more ?
OK, you are using a dipmeter, may radiate a lot :(
1 turn loop on a long cable ?

However the other measurements seem to indicate the following:

@2 https://github.com/2.) we need a frequency dependent I&Q
amplitude and phase adjustment. I fully agree with you, Jos, now with
those measurements. A difference in adjustment of +-3 can result in a
difference in mirror rejection of >20dB (not measured, but a
sophisticated guess by me). Sometimes +2 steps turns absolute silence
to a rather loud annoying heterodyne.

This matches my findings, 1 step makes a difference of no tone to a
noticeable tone.
I also could make use of the S-meter, no signal was approx. S3 of noise.
Signal input was S9 + 30 dB or so.
I could calculate the approx. rejection 1 S-unit = 6dB

@3 https://github.com/3.) my hypothesis must be rejected: the
buffers are not the problem!

Looks worse than better, but as said there are probable more variables,
it's tricky.

New hypothesis: the 74LCX74M (U11) is the problem, that chip seems to
have a huge frequency dependence in handling amplitudes and phase. I
will order the "wonder chip" (7474) by Potatoe chips in order to
validate that.

Andreas, could you confirm, it is this one, please:
http://www.ebay.com/itm/7474-G-Series-GHz-TTL-CMOS-logic-IC-14pin-SOIC-QTY-1-/330551715157

73 de Frank

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[DD4WH]

After some time for thinking, I have some new findings:

1.) With the buffers in my mcHF, there were no differences between LSB/USB. After taking out the buffer ICs, measuring, and after that additionally resoldering my solder bridges and measuring everything again, there are still significant differences between USB & LSB, but only in the amplitude adjustments, not in the phase adjustment! Thus, the buffers U12/U13 seem to be useful for eliminating frequency effects on the IQ amplitude differences. So, we do need software diiferentiation between LSB/USB adjustment, because RF board 0.5 has no more buffers.

2.) & 3.) Even IF the potatoe chip 7474 eliminates frequency dependancy, probably not every user will order it and build it in. So do we want a software adjustment? If yes, I would suggest: 12 different adjustments with the following low corner frequencies which I have chosen for ham radio use AND SWL in mind:
100kHz - 2200m
450kHz - 630m
1.75MHz - 160m
3MHz - 80m
5.5MHz - 40m
8MHz - 30m
11MHz - 20m
14.5MHz - BC 19m
16MHz - 17m
19MHz - 15m
22MHz - 12m
25MHz - 10m

12 adjustments for amplitude & phase for USB/LSB separately means 48 adjustments (and also 48 EEPROM memory places)!

73 de Frank

[df8oe]

After we got a working IQ-adjustment -and first at this point - we consider that there is huge hardware problem sleeping in design of PCB. Because of IQ-lines are not designed @50 ohms there are frequency dependencies which never were recognized before. Capacities of mixer inputs and output impedance of drivers (or of 7474) are involved, too.

I think before we start a huge storing orgy we first have to accumulate informations. Maybe it is possible to find a formula how better scalings can be calculated. But therefore we need much more samples of actually best IQ adjustments...

vy 73
Andreas

[DD4WH]

Yes, you are right, the new software now makes the hardest hardware tests possible for the first time!

And the few results are still not leading into one direction and hardware effects are complex.

So let us do more tests and gather more data under the diverse configurations.

I am really curious whether the 7474 chip from Potatoe chips will solve all that . . .

73 de Frank

[PA3CCE]

@andreas, Frank

Just my penny :

1. I do NOT see the necessity of making difference in adjustment in
   demodulation modes.
   dependency of other factors are many many times bigger.
   Besides, I have never seen this in other projects either.

2. I do NOT believe in one satisfactory adjustment for all frequencies
   (and on all existing mcHF variants) ,
   there are too many dependencies involved ( trace impedances, stray
   capacitances, stray inductances,
   component variations, etc. etc.) .

Best compromise :
Be realistic and make the best (simplest) solution for everyone, provide
for an adjustment for each
frequency band (maybe more for the highest frequencies), NOT regarding
modulation mode.

Best 73
Jos

Op 10-4-2016 om 16:55 schreef DF8OE:

After we got a working IQ-adjustment -and first at this point - we
consider that there is huge hardware problem sleeping in design of
PCB. Because of IQ-lines are not designed @50 https://github.com/50
ohms there are frequency dependencies which never were recognized
before. Capacities of mixer inputs and output impedance of drivers (or
of 7474) are involved, too.

I think before we start a huge storing orgy we first have to
accumulate informations. Maybe it is possible to find a formula how
better scalings can be calculated. But therefore we need much more
samples of actually best IQ adjustments...

vy 73
Andreas

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
#233 (comment)

[df8oe]

It is not trivial to implement so much storing (especially for users who don't have fitted serial EEPROM because in actual virtual EEPROM there is a cut at 383 words).

So let us "be patient" and find a solution.

Again:
all these problems came up a few days ago when IQ adjustment works as fine as is is now....

I don't think it is a problem in waiting some more days (or weeks) for finding best compromise - problem exists from the birth of mcHF on!!!!

We are pointed on this and working on it to get best solution.

vy 73
Andreas

[s53dz]

If I may, I would propose:

1. to use S-meter scale for testing the results of using lin. interpolation. The main purpose being to have all measurements on the same ruler. S meter scale is not big (may be not accurate-calibrated), however it allows for measurements which to my opinion are enough for this testing of method. As the difference can be quite high. The frequencies at which it would be measured don't matter as it can be easy drawn on a scatter chart. The chart would than show the deviation from the optimal balance at two points (measured and stored at the beginning).

2. If it turns out that for instance 10m band is deviating than we can move the upper point to say 12m instead to improve the method.

73 Bojan

[s53dz]

Hi all,
I just wanted to share my preliminary measurements after setting the two values for 80m and 10m and using interpolation.

The results show that the level of a mirror signal is from S3 to S5 compared to the worst case of S8 before.
The signal used was S9 +15dB = -58dBm.

Before: I have used the 20m values, where the mirror was faded out.
Now: I have faded out the mirror on 80m and 10m band.

In both cases the fade out was a deep one, down to S2 or less.
http://lea.hamradio.si/~s53dz/arhiv/x-razno/1-mcHF-70MHz/1-measurements/IQ_test1.pdf
Since the deep is really sharp I have observed also the influence of a temp drift.

73 Bojan

[s53dz]

And I added TX_IQ measurements still with a FW ver1.5.2!
http://lea.hamradio.si/~s53dz/arhiv/x-razno/1-mcHF-70MHz/1-measurements/IQ_test2.pdf

73 Bojan

[df8oe]

Very interesting to see phase and balance adjustments on different mcHFs. The influence of boards is enormous. But always linear iq calculation does a good job...

[s53dz]

I can comment only on my results (since I have not seen others):

• almost S5 (-97dBm) as a worst case on RX seems quite high to me
• -37dBc as a worst case on TX seems quite acceptable (1mW at 5W output)

73 Bojan S53DZ

[df8oe]

Other informations are scattered - not here on GitHub. We are starting collection of more informations now.

Linear interpolation is a big step again. Now we must see if there are other mathematical possibilities to get better proximation. Maybe there do exist some functions, between the user can select (depependent on mcHF PCB and/or modifications). Maybe there is no other possibility to add adjustment to much more datapoints (> 50). We don't know at the moment.

[PA3CCE]

My results of the new IQ calibration Rx.
mcHF version 0.5.
Latest daily Firmware 1.5.4 /BL: 1.0.0.
Potatoechips /4 in LO
LO and Rx mixer metal screened.
Signal generator (FA-VA3) via 26dB attenuator on antenna connector, showing abt -40dBm and S9+30 on the S-meter.
All dBm values read from mcHF screen.

11/12/2016
Rx IQ calibration :

3.600.001 MHz Balance - 9
Phase -12

28.100.001 MHz Balance -11
Phase 3
Result :

USB LSB Signal/Mirror

3.600/ 3.624 -40/-80 dBm
5.400/ 5.424 -38/-78 dBm
7.100/ 7.124 -40/-80 dBm
10.100/10.124 -38/-79 dBm
14.100/14.124 -42/-84 dBm
18.100/18.124 -40/-81 dBm
21.100/21.124 -40/-80 dBm
28.100/28.124 -42/-81 dBm

73 de
Jos (PA3CCE)

[s53dz]

Sorry,
My bad. Error at table copy.
This is the right attachment now.

Hi Jos,

Thanks for sharing your results. It seems that interpolation works just fine.
However, I think I do not understand completely. Could you please comment.

why the results on 80m and 10m (where the reference setting is done) are not better?
have you tried to measure mirror at somewhat lower input level?

I have recalculated my results to dBc. It shows some 17dB difference at 80m end.

73 Bojan S53DZ

IQ_test3_dBc.pdf

[PA3CCE]

Hi Bojan

Well, that the results at the calibration points were not better than the others stroke me also..
The only explanatation I can give is that I did not do the dBc measurements immediatly after calibration.
I also observed that when I did a recalibration the settings differed slightly sometimes (temperature ???)

I will do the calibration again and measure the dBc directly afterwards to see if the values are better at that moment.
That 17 dB difference is at the lower frequencies, above they are quite similar.
Will let you know my findings..

73 de
Jos

[S52CQ]

I have followed instructions and done calibration both Rx & Tx - now I can see/hear a big dip in the signal when changing calibration.

Great job!
73 de Jure, S52CQ

[PA3CCE]

Hello Bojan,

I forgot to tell you abt the lower input level.
Yes I did, level was abt S7 at the time and the mirror rejection was abt the same 40dB's but instable caused by the noise. So I decided to increase the level considerably.

Jos,

[s53dz]

Jos,

Thanks a lot for your comments.
Yes, I too noticed the very sharp behavior of I/Q settings, so the influence of temperature drift is high here.
I have rechecked this Issue thread for whole and I can repeat what was already said: Nothing can replace the initial I/Q error coming from the HW! Although I am happy to see this important improvement in the mcHF life using a simple linear interpolation. Respect.

73 Bojan S53DZ

[PA3CCE]

Bojan

I just did some tests at 3.6MHz :

I put the 26dB attenuator after the frequency generator and measured -40dBm carrier on the mcHF.
Then I adjusted the mirror at minimum, that level was -78 dBm on the mcHF, so -38 dBc.
The gain setting was -7 and the phase setting -12.

Now I exchanged the 26dB attenuator for one of 40 dB.
The mcHF showed -54dBm, so 14 dB's less that is excellent.
Then I re-adjusted the mirror rejection.
Now the gain nsetting came out at -12 and the phase setting stayed at -12
Conclusion : gain setting is level dependent or compression @ -40dBm .
However, level of mirror was -91 dBm --> -37dBc (let's say it's the same as before).

Will continue tomorrow.

73
Jos

[PA3CCE]

Should be compression @ -26 dB,. sorry

[PA3CCE]

Bojan,

The way I measured the mirror rejection by reading the displayed values on the mcHF screen is NOT correct, sorry.
I found that without signal on the antenna plug my noise level indicates -128 dBm on the mcHF screen.
When I connect the signal generator via the 40dB pad on the antenna connector I read -54 dBm on the frequency , but whern I tune off frequency the noise level now shows -93 dBm on the screen.
I had set AGC off in the menu.
So that's the value I always will measure when I turned down the mirror completely (abt -40 dBc).
Anyhow, the reality will be better. Right ?
How did you measure the dBc,, on the audio?
if so, it should be a selective measurement on 1 kHz to be correct..
Maybe better on 12kHz on a spectrum analyser, i will try that.
Regards
Jos

[PA3CCE]

Bojan,

On 12kHz I only see the carrier on the spectrum analyser obviously.
When I measure the audio out on my spectrum analyser using mcHF's smallest bandwith on SSB (1.3-2.3 kHz) the difference between max signal @ 1kHz and the noise is just a bit more than 40 dB.
So I never will be able to measure more than that abt. -40 dBc, though it will be more.
How did you manage, is your noise level better than mine ?

best regards
Jos

[s53dz]

Jos,

With a 50 Ohm dummy load (40dB+10dB attenuator) at the RF input and using a CW filter I have mcHF displaying an S1 unit. And shows around -130dBm. But at this measurement I have used the SSB filter instead since it is more convenient and also it is not important to go really into that deep. Then I have some S2 units. The dip itself is really too sharp and it quickly becomes irrelevant when you go off the freq. It would be informative to see the IQ unbalance when inside the band.

I forgot to mention that the TCXO temp at the time was nearly constant 45 deg C.
It is also helpful if the signal generator has PN at least -50dBc at 24kHz offset.

Measuring TX mirror is quite straight forward. Of course it would help to have a narrow BW filter to notch the carrier out but it is not feasible. I am happy with that.

Added(s53DZ):
My RX IQ measurements could be biased since I have trusted the S-scale of mcHF and then calculated dBm out of that. I have measured the input signal level with external power meter and it was -49,5dBm. mcHF showing -54dBm.
Added 2(s53DZ):
An example: S9+20dB = -53dBm, S2.7 = -110.8dBm, difference = 57.8dB
TX IQ measurements are ok.

73 Bojan S53DZ

[DG9BFC]

hello jos this is how i would (try to) measure that test transmitter set to rx qrg ... crank up till you see an s2 or so now tune test transmitter to mirror crank up till you again see s2 (or what you had with first setting) (added level is mirror surpression) retune test tx to rx qrg ... (now you will see s9+ something) ... no ... do not use that readout!!!! (cause you never know how accurate s metre is!!!) add attenuator till you again see s2 added damping is your mirror surpression dg9bfc sigi Am 12.12.2016 um 14:32 schrieb PA3CCE:

…

Bojan, On 12kHz I only see the carrier on the spectrum analyser obviously. When I measure the audio out on my spectrum analyser using mcHF's smallest bandwith on SSB (1.3-2.3 kHz) the difference between max signal @ 1kHz and the noise is just a bit more than 40 dB. So I never will be able to measure more than that abt. -40 dBc, though it will be more. How did you manage, is your noise level better than mine ? best regards Jos — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#233 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AQk8wKyqtiqdpb7XbR9yjj_1HZc912mdks5rHVrogaJpZM4IA_K5>.

[PA3CCE]

Bojan,

Thanks for the reply.
I now understand you first measure the input level with an external power meter and then calculate the mirror level from the S reading on the mcHF.

What I did is:

1. Reading the dBm value on the mcHF at the signal frequency and subtracting the dBm reading at the mirror frequency.

2. I observed the audio spectrum on a spectrum analyser and measured the peak value of the 1 kHz signal to be 40dB over the noise .
   Since the mirror was turned down into that noise I only could say "mirror suppression is > 40 dB".

So both methods gave the same result > 40dB, but can't say how much due to the noise.

[PA3CCE]

Hi Sigi,

Thanks for your suggestion.
However, I am satusfied with the result > 40dB, knowing at all frequencies it is ( probably mostly far) better than that.
Anyway a big improvement compared to the old situation.

Thanks agn es
73 de
Jos

[dalerobins]

Hi all,

I'm after some advice regarding the IQ Phase/Balance.
I recently built my mcHF and am using the amazing firmware by DF8OE (Latest version)

One thing that is bothering me is on 10m to completely remove the "tone" my IQ Phase is around +124, this seems quite high to me as I don't have much adjustment left.

Frequency correction was on -12Khz to do the adjustments.

My board is a V0.5 from Chris, no modifications at all.

Is this normal or is something wrong, I am worried maybe I have not done something correctly, will having my phase so high cause any issues? - If something is wrong what do I need to do to correct it.

Many thanks, and keep up the great work!

de 2W0ODS IO81

[df8oe]

I think your 74xxx74 is at the border of its working capabilities and has very much phase differ at higher frequencies.

We have investigated that usage of 7474 manufactured by "potatoe chips" is working much proper. It is capable to work up to 1GHz so 120MHz (30MHz * 4) is not a problem at all.

Unfortunately Chris has redesigned 0.5 unluckily:
Vdd of 7474 in 0.5 is now 5V - but potatoe chip can only be used at 3.3V. So you have to modify Vdd for divider if you swap to potatoe chip.

Easier way is to try another manufacturer of 7474 to get a faster one...

73 de
Andreas

[dalerobins]

Andreas,

Thank you so much for a quick reply, was not expecting to hear anything today.
Merry Christmas to you.

I see, so it is not anything I have done when winding toroid or transformer. That is a relief.

I will look at the Potatoe Chips (That is a funny name), or a faster 7474.

Will everything work okay on my mcHF as it is, or is it causing a problem having the phase so high on 10m, should I not worry about it?

Love your firmware, it is working very well here. Noticed a couple of couple of minor bugs (display corruption when changing S-Meter mode and you do not have the dBm/Hz display enabled, enabling it removes the corruption. Also waterfall you cannot see the > after magnify)

[df8oe]

H Dale,

Merry Christmas, too! I am spending some mnutes on GitHub every day - Christmas time included :)

There will be not a big problem. Your adjustings are within possible range. But I guess linear calculation for all other bands will not be as exact as they can be if the differences between 80m and 10m in settings are high. Phase error may be non-linear increasing at higher frequencies so it will be better if you reduce hardware phase errors as good as you can do it...

Take a look at the datasheet from semipotato: http://www.potatosemi.com/datasheet/PO74G74A.pdf

73 de
Andreas

[dalerobins]

Thank you again Andreas,

One thing I forgot to mention and I hope I am not causing any confusion here, the IQ Phase I adjusted on 10m is for TX, that value is at 124, I have not done the Rx ones yet but will do those today.

Also does anyone else have higher SWR readings on higher frequencies with a known good dummy load, I it registers 2-3 on the SWR meter. I hope this is not causing the radio to put out a lower output power.

[df8oe]

Hi Dale,

output power is not reduced at higher SWRs.

There is a design fault for tandem couplers. Please take a look at our German Project Group page for modification RF-04-H-023. Sorry it's available in German actually :/ The toroid windings (many windings) which lead to PA output must be removed and connected to BNC side (it's connected to the wrong "side"). If you correct this readings will be more accurate.

RX and TX IQ settings are not cross-dependent - so no care to first do only one setiing.

73 de
Andreas

[dalerobins]

Is this still an issue with v0.5 board, I was ignoring any "04" mods as I thought they were not applicable. I will go back and look at all the modifications on your Wiki again. Just wanted to check with the IQ TX too see if it is still 7474 for Tx and if +124 is still okay for Tx. I can use google translate, if not my close friend who lives in Heidelberg can help me. I love Germany it is one of my favorite countries, when I worked for HP I spent a lot of time there, mainly in Boeblingen. My friend moved from the UK to Germany and would not come back here.

…

On 26 December 2016 at 14:19, DF8OE ***@***.***> wrote: Hi Dale, output power is not reduced at higher SWRs. There is a design fault for tandem couplers. Please take a look at our German Project Group page for modification RF-04-H-023. Sorry it's available in German actually :/ The toroid windings (many windings) which lead to PA output must be removed and connected to BNC side (it's connected to the wrong "side"). If you correct this readings will be more accurate. RX and TX IQ settings are not cross-dependent - so no care to first do only one setiing. 73 de Andreas — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#233 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AXoNHaiukZ0IvAl4MnCZeRgJdXVRX6prks5rL8z3gaJpZM4IA_K5> .

[df8oe]

Hi Dale,

124 is "the edge". But if you cann suppress mirror deeply it will work. For better results in calculated settings try faster flip flop 7474...

We do not have completed mod comparison 0.4/0.5 because none of the fw developers own 0.5 boards. Because Chris removed live debugging headers on 0.5 there is no mitivation to change to this revision.... But I think many modifications are NOT implemented in 0.5 boards... Unluckily you have to test this by yourself :/

Good luck - 73
de Andreas