Pages: [1] 2 
|
 |
|
|
Author
|
Topic: poor sideband suppression (Read 4693 times)
|
|
S53DZ
schon länger dabei
 Offline
Posts: 58
Ich liebe dieses Forum!
  
|
 |
poor sideband suppression
« on: 04. September 2018, 21:57:08 »
|
|
Hi,
I have made some more measurements, this time of an unwanted side-band.
The results are in the attachment.
This is probably off-topic for this thread. But as it is related to the last FW I am posting it here.
EDIT3:
- it is OVI UI used
- it is sideband and not 12 kHz issue
SRI for that
For the overview I have measured the ratio in USB and LSB mode separately.
The last column shows the respective noise floor for the chosen BW.
Walking along the filter window the suppression of the unwanted side-band remains the same. Only out of filter BW the suppression improves (which is logical).
EDIT:
Yes, RX IQ auto = ON
This means that at about 10 MHz the IQ imbalance results are the worst.
EDIT2:
Correction: The f (MHz) column should show a constant 10.100 MHz!
73 Bojan
|
|
|
|
|
DD4WH
positron
alter Hase
Offline
Posts: 462
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #1 on: 05. September 2018, 19:10:08 »
|
|
Hi Bojan,
thanks a lot for your measurements!
One question I have:
* was your tone at 1kHz?
* if yes, your results only state the opposite sideband suppression in 1 kHz distance from the carrier
* it would be interesting to have the suppression measured in different distances to the carrier, I probably wasnt clear enough in explaining that
BTW: I have a new version with (presumably) much better sideband suppression and (hopefully) the same satisfying performance in suppressing the 12kHz aliases.
Could you test that new version? (will probably be issued tomorrow!)
Thanks in advance!
73 Frank
|
|
 Logged
|
-----------------------------------------
Teensy Convolution SDR
https://github.com/DD4WH/Teensy-ConvolutionSDR
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #2 on: 05. September 2018, 19:58:21 »
|
|
Hi Frank,
my measurements of an unwanted side band were done like I wrote in the previous post. Perhaps I simplified too much?
For each chosen BW I did the following:
- I have measured the wanted side band signal level first,
- I re-tuned to the unwanted band
- I have measured this level about at the middle of filter curve,
- I have walked (tuned) across the filter band pass to see the possible difference in response for different offset from the carrier.
(I repeated it also by just switching USB/LSB between w/unw.)
Actually USB and LSB in my table is just to show that the behavior is very symetrical to the carrier.
As there was almost no difference I stated just the first (center) value.
EDIT:
My "tone" was at 750 Hz, when trying with 500 Hz and for test tried also with a 300 Hz BW
Yes, I will do this again for the next FW version to see the difference.
EDIT2:
In preparation for next FW I have done measurements with higher input level. It seems that I was influenced by a higher degree of dBm error at very low signal levels. The results now are more balanced.
These are my new reference figures to compare with.
FW 2.9.51
f (MHz) BW USB unw. com.
Hz dBm dBm dBc
10,100 0k5 -44 -97 -53
10,100 1k8 -43 -95 -52
10,100 2k3 -43 -94 -51
10,100 3k8 -43 -94 -51
|
|
73 Bojan
|
| « Last Edit: 05. September 2018, 22:00:44 by S53DZ » |
Logged
|
|
|
|
DD4WH
positron
alter Hase
Offline
Posts: 462
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #3 on: 06. September 2018, 07:17:35 »
|
|
The new firmware version 2.9.52 should combine both:
* reliable suppression of aliases at 12kHz spacing
* provide enough opposite sideband suppression for every HAM radio circumstance
Please test!
73 Frank
New audio path:
- lowpass filter I & Q with 43 tap (mcHF) or 83 tap (OVI40) lowpass filter @48ksps sample rate
- downsample-by-4
- heavy Lowpass Hilbert transform Filter +45 / -45 degrees with 199 taps @12ksps sample rate
- etc. . . .
|
|
Logged
|
-----------------------------------------
Teensy Convolution SDR
https://github.com/DD4WH/Teensy-ConvolutionSDR
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #4 on: 06. September 2018, 10:39:00 »
|
|
Hi Frank,
I will try to do it with FW 2.9.52 later today.
While doing side-band measurements with a high input level, such as -44 dBm, I have noticed the behavior similar to the repeating auto codec reset. Is it possible to get to this situation when a strong enough CW signal is applied to the input and tuning in the vicinity of it?
73 Bojan
|
|
Logged
|
|
|
|
DD4WH
positron
alter Hase
Offline
Posts: 462
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #5 on: 06. September 2018, 10:47:56 »
|
|
| Is it possible to get to this situation when a strong enough CW signal is applied to the input and tuning in the vicinity of it? |
|
No, I dont think so.
But you are right, we should take care that the input level of the test signal does not overload the ADC. So a maximum input level of about -53dBm (S9+20dB) is probably a nice idea to be on the safe side for the tests.
73 Frank
|
|
Logged
|
-----------------------------------------
Teensy Convolution SDR
https://github.com/DD4WH/Teensy-ConvolutionSDR
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #6 on: 06. September 2018, 16:00:03 »
|
|
Hi,
These are my figures for FW 2.9.52.
FW 2.9.52
f (MHz) BW USB unw. com.
Hz dBm dBm dBc
10,100 0k5 -53 -106 -53
10,100 1k8 -53 -104 -51
10,100 2k3 -53 -104 -51
10,100 3k8 -53 -101 -48
|
|
I can spot no change in the unw. side-band suppression.
All the settings remained unchanged for a direct comparison.
Only the input level I changed to -53 dBm. But the same result is with -44 dBm.
73 Bojan
|
|
Logged
|
|
|
|
|
|
DB4PLE
positron
Urgestein
Offline
Posts: 1278
|
|
Re:poor sideband suppression
« Reply #8 on: 06. September 2018, 18:09:05 »
|
|
Hi,
the data used in the scope and waterfall display is also used to calculate the dbm/S-meter values. This data extracted before the main signalprocessing (since this uses decimated data and is of much lower bandwidth than the scope).
All changes in the past releases did not affect the processing of the data which is used by the scope/waterfall and hence the dbm values should not change. Since the problematic signals for instance the 12khz are basically byproducts of the signal processing which happens after the scope/waterfall data processing, these never showed up in the dbm & S-meter. Francois video showed that beautifully. We heard the 12khz signal but the dbm Value remained basically unchanged.
With the twin peaks syndrom and IQ imbalance, this is a different story as this is already in the original input data, and this has an effect also visible on the meters.
This is at least what I learned from trying to figure out what this signal processing does while working on the code.
73
Danilo
|
| « Last Edit: 06. September 2018, 18:09:50 by DB4PLE » |
Logged
|
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #9 on: 06. September 2018, 19:16:58 »
|
|
Thank you all for this explanation or better putting some light to it.
But now I am quite unhappy although -53 dBc is already quite good figure.
If I can hear some strange low level distorted audio (Andreas, you are right that it is somehow suppressed) in place where in the spectrum I can still see the signal and on S meter I can correspondingly measure it, then I am concerned about the way this suppression is made.
73 Bojan
|
|
Logged
|
|
|
|
DF8OE
Administrator
Offline
Posts: 6289
Stellvertr. OVV I40, Jugend / Nachwuchsreferent
|
|
Re:poor sideband suppression
« Reply #10 on: 07. September 2018, 03:58:23 »
|
|
Test procedure:
RX 10.100MHz USB, 2.9KHz LPF, injected test signal 10.1002MHz, Signal level -33dBm. I do hear very strong signal. If I switch to LSB I do hear very weak signal, assuming S1...S2, nearly eaten by the noise floor.
If I tune test signal to 10.1003MHz, signal is eaten by the noise floor.
If I reduce input level to -43dBm, 200Hz signal is eaten by the noise floor, too.
So I can state mcHF (this is the RX I have tested with) is very, very good. And OVI40, which does have a steeper filter, has the best suppression of wrong sideband compared to all commercial TRX I own. So I am very, very happy with the existent suppression.
vy 73
Andreas
|
|
Logged
|
Wenn der Wind des Wandels weht, nageln die einen Fenster und Türen zu und verbarrikadieren sich. Die anderen gehen nach draußen und bauen Windmühlen...
qrz.com-Seite von DF8OE
-----------------------------------------------------
>>>> Die Inhalte meiner Beiträge dürfen ohne meine explizite Erlaubnis in jedwedem Medium weiterverbreitet werden! <<<<
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #11 on: 07. September 2018, 05:24:56 »
|
|
Hi Andreas,
Thank you very much for your comment, rather than your measurements.
I can see that you are satisfied by this kind of audio suppression.
But, please, try to understand me when I say that stating "assuming S1..S2" is not very persuasive to me.
Can you confirm that the levels measured with S-meter in dBm in your case, on your mcHF RF board, are in the range of figures that I have measured? Otherwise you are telling me that I have to dig-in for a HW issue. Hi. Considering that I think the unw. side-band suppression is a HW / SW matter.
EDIT:
As long as your CW signal in the spectrum is above the noise floor, you can measure it's level. However, when nearing the floor the accuracy is falling by going to wider filter BW. And also the floor is rising. Here, with strong level signal at the RX input, the unw. CW signal is well above floor.
VY 73 Bojan
|
| « Last Edit: 07. September 2018, 05:40:15 by S53DZ » |
Logged
|
|
|
|
DD4WH
positron
alter Hase
Offline
Posts: 462
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #12 on: 07. September 2018, 07:45:29 »
|
|
Hi Bojan,
yes, I agree that measurements are important and objective, but our ears are the best audio measurement instruments available on the market with more than 120dB dynamic range, the mcHF/OVI40 is definitely worse in that respect ;-).
I still have problems in interpreting your measurement results, there seems to be a misunderstanding:
* for opposite sideband suppression, it makes no difference which audio filter you choose. The opposite sideband suppression is determined by two things: 1.) IQ SIDEBAND SUPPRESSION: a correct IQ phase & amplitude balance and (at least equally important) 2.) HILBERT SIDEBAND SUPPRESSION: the performance of the Hilbert filters that do the 90 degree phase shift between I & Q in order to cancel out the opposite sideband (because I & Q are already 90 degrees apart and if you shift them another 90 degrees, the opposite sideband is cancelled out).
* the audio filter (the one that the user can choose, eg. 500Hz BW, 1k4 BPF etc.) comes AFTER the Hilbert filter and is a simple IIR software filter. This IIR has nothing to do with sideband suppression, because it does not work on I & Q, but on single channel audio.
* Second thing: I made a mistake in recommending using the dBm-display for "measuring" opposite sideband suppression. Danilo has already pointed us to this ;-).
The audio path goes like this:
* I & Q signals enter the digital domain via the ADC
* I & Q signals are phase & amplitude corrected in software
* I & Q signals go into the FFT for the dBm display & the spectrum/waterfall
* I & Q signals go into decimation & Hilbert filtering for the audio signal
* etc.
So, the dBm display & the spectrum display shows the IQ sideband suppression that is achieved by the software correction of I & Q. It does NOT show the HILBERT SIDEBAND SUPPRESSION, because the Hilbert filters come after the dBm display.
The HILBERT SIDEBAND SUPPRESSION would have to be measured on the audio signal itself. That is why Andreas trusts his ears ;-). One solution for a more objective measurement would be to measure the audio output with a scope or an AC voltage measurement instrument directly at the headphone output. [settings: tone at 500Hz above the carrier frequency, LSB, CW filter with 300Hz bandwidth and 500Hz centre frequency]
(that is also the reason why you cannot measure a difference between versions 2.9.50/51/52 with the dBm display, because the IQ sideband suppression is exactly the same in these versions and only the Hilbert sideband suppression differs hugely in these versions: so you hear it, but you dont see it ;-))
* I interpret from your Email that you are not sure whether there could be a hardware issue regarding sideband suppression? I cannot see anything in that direction: there is -of course- hardware related IQ phase and amplitude imbalance in the I & Q signals. But as soon as these imbalanced I & Q signals enter the ADC, this imbalance is corrected in software with an accuracy that leads to a maximum of 65 - 70dB suppression, which seems to be the maximum achievable with single precision floating point calculations. You simply cannot achieve 65dB in hardware! Have a look at EMRFD (Experimental Methods in RF design by Hayward, Campbell & Larkin): their best hardware phasing approaches can achieve 50dB for SSB audio signals (and involves very carefully selected components and shielding and is restricted to one fixed filter bandwidth).
So, please excuse my wrong recommendation of using the dBm display to measure Hilbert filter sideband suppression! That was my fault!
Comparing the opposite sideband suppression of different commercial transceivers with the mcHF/OVI40 is a good idea. Use the same filter, the same distance of carrier signal and trust your ears (or your audio level measurement instrument) :-).
73 Frank
|
| « Last Edit: 07. September 2018, 07:53:34 by DD4WH » |
Logged
|
-----------------------------------------
Teensy Convolution SDR
https://github.com/DD4WH/Teensy-ConvolutionSDR
|
|
|
|
|
S53DZ
schon länger dabei
Offline
Posts: 58
Ich liebe dieses Forum!
|
|
Re:poor sideband suppression
« Reply #14 on: 07. September 2018, 19:02:00 »
|
|
Hi all,
Thank you for this explicit and fine explanation. It is more clear now to me how and where it is processed. It is my fault that I did not read it first. So it is different computing on the audio signal for this suppression. It is clear now that my measurements could not show any change.
Unfortunately I have no noise-like BW limited modulation signal here to check the behavior of Hilbert at suppressing such shape of unwanted signal. I am curious what Andreas's ears would hear in that case, hi.
Anyway, this suppression is good achievement! But I still have trouble of having a displayed signal, which is then not heard. 
Andreas, I see the point in your quote, my observation was about mixing the word dBm (S-meter figures), which are an RF signal category with a level of an AF signal.
73 Bojan
|
|
Logged
|
|
|
|
Pages: [1] 2 
|
|
|
|
|
|
|
10M1sideband.pdf
