Welcome to the new FlexRadio Community! Please review the new Community Rules and other important new Community information on the Message Board.
If you are having a problem, please refer to the product documentation or check the Help Center for known solutions.
Need technical support from FlexRadio? It's as simple as Creating a HelpDesk ticket.

Other side band not properly suppressed

2»

Comments

  • Tim - W4TME
    Tim - W4TME Administrator, FlexRadio Employee admin
    edited November 2018
    No, we have not acknowledged this as a problem.  But I have entered an issue in our bug tracker (#4544) to investigate the reported behavior.
  • K2CB Eric Dobrowansky
    K2CB Eric Dobrowansky Member ✭✭
    edited March 2017
    Ok, thank you Tim.
  • Sergey R5AU
    Sergey R5AU Member ✭✭
    edited March 2017
    Sounds good Tim, thank you !
  • Ken - NM9P
    Ken - NM9P Member ✭✭✭
    edited June 2020
    My own testing results:

    SET UP:

    Testing was done using my 6500 in FDX mode.  Two pans, one slice in each pan.
    Transmitting in Slice A  into the Transverter port on 75 Meters.  Receiving on Slice B with Antenna

    Since this is a test of the SSB Generation and filtering, I didn't feel the need to test with any actual RF power output.

    Using Audacity, I created four different test .wav files for this test.
    (Audacity signal level of .35, because any higher wold overdrive my transmit audio)

    1) Using the standard IMD testing tones of 700 & 1900 Hz.  
    2) Using Sergey's tones of 100 & 1000 Hz.
    3) 30 seconds of "Pink Noise."
    4) 30 Seconds of "White Noise."

    I was skeptical of the 100/1000 tones because they are harmonically related and might generate in-band distortion that might not give a good picture of what is really going on.  But I tried them anyway.

    The tones were sent via the voice keyer feature in ACLog 5.8, which allows sending tones via the DAX TX Audio channel.  I tested using 4 or 5 of my standard Mic Profiles, including my ESSB 6K, Rag Chew 3K, DX 2.7K, and Contesting 2.3K.  I also turned PROC On & OFF, and rolled the low cut filter setting up and down from 0 to 350 Hz on all tests.

    I set the DAX TX Audio level so that the tones would drive the rig one notch below the "red" audio level on the TX Panel.

    I realized that the one test I didn't do was with the TX EQ completely FLAT.  (How did I forget THAT one?)  But I DID try all manner of LO CUT settings.

    Test 1:
    Using tone set #1.  (700 & 1900 Hz)
    I detected NO undue IMD or opposite sidebands and could not produce any.

    Test 2:
    Using tone set #2 (100 & 1000 Hz)

    If I took the TX filter Lo Cut all the way to 0 (Which I never use) I was able to get a pattern similar to what Sergey posted, although my 100 Hz opposite sideband was down 25 dB from the main sideband.

    If I rolled the Lo Cut up in frequency, by the time I got to 250 Hz, there was no significant opposite sideband tone.

    THIS SURPRISED ME a bit:  If it turned PROC OFF, then the rejection was a little worse.  With Lo Cut of 0, the rejection was perhaps 20 or 15 dB down.  (My experience with most rigs is that PROC actually makes things WORSE, not better.  It must be the PROC algorithm.)  

    But even with PROC OFF, if I turned LO Cut up past 200-250 Hz, the opposite sideband disappeared.

    At no time were there any opposite sideband products beyond about 250 Hz, 


    Test 3 - Pink Noise.

    With pink noise running, and the Lo Cut at 0 Hz, there was a little bit of opposite sideband mostly 30 dB down, declining rapidly as the audio frequency increased, and NONE of it extending beyond about 200-250 Hz.

    As I moved the Lo Cut up, the opposite sideband quickly attenuated.  By the time I got to 250 Hz, there was none at all.

    Test 4 - White noise.

    Test results basically the same as with Pink noise.


    SUMMARY AND CONJECTURE:

    I do not think this is an IMD problem.  

    So I don't think it would be remedied with Adaptive Pre-Distortion routines.  (To prevent that argument spilling over into this thread.  APD will have its own benefits when it arrives, but this is not that.) 

    I think it is a filtering skirt issue.

    It seems to be real, but I wouldn't classify it "serious" under most normal situations.

    With my "normal" usage, with my PR-22 mic and my particular mic settings (My usual Rag Chew Mic Profile setting has a low cut of 65 Hz, and even my ESSB Mic Profile only goes down to 50 Hz) I do not see any significant opposite sideband problems.

    BUT....

    If the audio has a VERY strong bass presence, the Lo-Cut of the filter is not filtering it all out.

    I suspect that when the TX filters were redesigned for lower TX latency and better turn around time, that the skirts were loosed up enough that they are letting a little through, especially if the user is using a "hyper-bass" profile.  

    The worst case scenario was with the 100/1000 two-tones, with PROC OFF, and Lo-Cut at 0 Hz.
    By increasing drive to the absolute limit, at the edge of the "red" I could generate an opposite sideband signal of the 100 Hz tone that was 15-25 dB down.  But It quickly improved with more normal settings of the Lo-Cut filter.  At 50-100 it was still there, but reducing to about 25 dB down.
    Between 100-200, it moved to about 25-35 dB down.  at Lo-cut of 250 Hz, it was gone completely.

    Note.....If one is using what I have called a "Hyper-Bass" profile strong enough to cause significant blast-through to the opposite sideband, I wonder how intelligible the signal would be, especially at lower signal levels (see my you tube tutorial on my Audio Profiles), but that is a different story.

    It might be that on SSB/Digi we may want to have a latency or filter skirt option for transmit similar to the one on receive that lets us select between two modes:

    1) "Standard" mode (the way it is now) with lower latency & quicker faster turn around time, and 
    2) "Brick Wall" mode (The way it used to be)  with higher latency, but sharper filters for those who want them.

    I will need to test these tones again with a completely FLAT audio profile to see if there are any changes from these initial observations.
     
    Hope this helps...

    Ken - NM9P
  • Sergey R5AU
    Sergey R5AU Member ✭✭
    edited March 2017
    Ken thank you for perfect notes and you participation.
    You are absolutely right - no issue at all with IMD - I didn't ' see worse then -35db and next improvements only on step with clean signal implementation.
    BTW I'll test RX filters also and result should be very close to TX filters that I sense on strong carriers however, suppression level of unwanted sideband a part of phase error in IQ algorithm and should not so depends on the low cut filter skirt.
    So, on the fact we are recognized point of addressable improvements here.
  • Tim - W4TME
    Tim - W4TME Administrator, FlexRadio Employee admin
    edited March 2017
    I'd like to provide some very preliminary feedback as we discussed this issue in great detail this morning during our weekly defect review meeting.  We have a fairly good idea what might be happening and decided on a testing methodology to validate our suspicions.  We believe that a change to the firmware to reduce the latency of the transmit chain on phone is responsible the behavior described above by Sergey and Ken.  If this assumption is validated, then we have a course of action to mitigate the issue.  At this time I do not have any ETAs on when testing will start as we have to manage the current development workload and integrate this unplanned activity into the development schedule, but it does have a high priority.  Thanks to everyone for the information provided.
  • DH1RK
    DH1RK Member ✭✭
    edited March 2017
    Hello Tim,
    I did an additional test, too.
    1. Pink Noise test, result good.
    2. Stepped Sine from 20Hz ... 500Hz, not okay from 20...220Hz

    In test 2. all harmonics (2nd, 3rd order cut @ 220 Hz).

    See the attached pictures and link to my video.
    Hope this helps to identify the root cause.
    Thank you in advance.

    Link to all:
    https://app.box.com/s/nf7s3da5g8by543recxgi3b8nddqsfjz

    PinkNoise:
    image

    Stepped Sine:
    @20Hz
    image

    @100Hz
    image

    @220Hz
    image

    Measurement video see attached link above.

    Hope this helps.

    73 de DH1RK, Ruediger
  • Tim - W4TME
    Tim - W4TME Administrator, FlexRadio Employee admin
    edited March 2017
    Thanks, Ruediger.  I'll add this info to the open issue.
  • ctate243
    ctate243 Member ✭✭
    edited March 2017
    Lets hope that the fix in question does not reverse the latency reduction that has been instrumental in improved performance in rapid QSO situations.
  • Eric-KE5DTO
    Eric-KE5DTO Administrator, FlexRadio Employee admin
    edited March 2017
    We are likely to handle this by increasing the filtering depending on the low cut filter.  The closer to 0 you are, the more filtering is necessary (and thus, more latency).  Where you do you usually have your Transmit Filter Low Cut set for contesting?
  • ctate243
    ctate243 Member ✭✭
    edited March 2017
    Sri eric I just sent a note out on Alpha we can chat there..
  • K2CB Eric Dobrowansky
    K2CB Eric Dobrowansky Member ✭✭
    edited March 2017
    Please don't forget, a lot of us are NOT contesters, and are into ESSB to some degree. We do utilize the low end of the rx and tx audio range to some degree. 
    A lot of us bought Flex radios for this very reason (unlimited TX bandwidth).

    I hope Flex has not forgotten which portion of the ham community got them to where the are today; it certainly was NOT the contesters!   IMO, if would be a foolish move to focus entirely on the contest community going forward, and ignore the original user base. 

    Personally, I run the tx filter setting at 0-4000Hz, then control the actual passband via external audio gear. I can then rapidly switch my bandwidth via operating profiles in the rack gear or virtual sw audio chain.
  • ctate243
    ctate243 Member ✭✭
    edited March 2017
    Eric D.  Please dont turn this into a bashing of a particular group.  Its probably not the healthiest approach to suggest things are done because who got flex where.  There are multiple communities withing the community and the team does their best to balance those needs.  The team will come up with a solution but hopefully one that will work for everyone.
  • Ken - NM9P
    Ken - NM9P Member ✭✭✭
    edited March 2017
    I think the idea of variable skirts (and consequentially higher latency), based upon proximity to the Zero-beat line is potentially a good solution.  As one who does almost EVERYTHING in the hobby with my 6500, I find that I seldom engage in "Rapid QSO" situations with my ESSB or Rag Chew profiles that have Lo-Cut set below 100 Hz.  So a little more latency wouldn't bother me at all.  A few more milliseconds is nothing in a long Rag-Chew.  My Main Rag-Chew profile is 65-3000 Hz

    My various "High Performance/High Intelligibility" or weak signal Mic Profiles have the Lo-Cut anywhere from 115-300 Hz, depending upon the conditions.  My Main 2.3 KHz Contest profile is 300-2600 Hz.  My Main DX_2.7 Profile is 165-2900 Hz. 

    I would not want to sacrifice turn around time in a contest or DXing situation using those profiles.  But chances are, with the testing I have done, they aren't a problem even with the current filtering.  

    The "sticky wicket" here might be Digital.  I have already been setting my Lo-Cut to 25 or 50 Hz for WSJT-X and other modes that use a wide bandwidth.  On 160 Meters, I even roll the low end of my receiver filter up to about 150-175 to block the WSPR signals when I am using JT-65/9.  Finding a compromise between digital turn around time and proper filter sharpness is going to be a challenge.
  • ctate243
    ctate243 Member ✭✭
    edited March 2017
    Yep seems like there are some options out there that will fit the bill for everyone.. the right solution will work for all.  Its great these guys identified this and made the data point available and from what I can see it looks like an equitable solution is on the table to re-mediate it.  Not sure about the digital stuff tho..
    
  • K2CB Eric Dobrowansky
    K2CB Eric Dobrowansky Member ✭✭
    edited March 2017
    Chris, I am not bashing anyone.

    I'm not the only one who has valid concerns. To many, it appears Flex's main focus shifted from the development of SSDR to the Maestro, and from the Maestro on to the Power Genius XL. And let's be honest, the last few SSDR releases have been mostly long overdue fixes and tweaks, and no new earth shattering features, sans the USB cables. And there are still items that have not been addressed yet, like something as simple as the screen formatting issue where the five right most boxes don't display properly, and the right edge not going full screen. Yes, simple things, but things that should have been fixed long ago. So no, I'm not bashing, I'm just being honest.

    The real question is, how did this issue slip thru the Alpha and Beta process? It's not the first time things like this have happened. It seems like two steps forward, one step back, on numerous occasions. Maybe Fkex should re-evaluate their testing teams and look to add some of the more technically astute users, such as the ones here who are discovering these sort of issues after the fact. When new features are added, have select testers who continue to test and ensure previous portions still function properly, in addition to those testing the new functions.
  • DH1RK
    DH1RK Member ✭✭
    edited March 2017
    Eric - KE5DTO
    Can you give more details about the filter (type, order, data type, etc.).
    Did you investigate different calculations for the filter coefficients (denominator base, calculation iteration with multiplication / division)?


  • Tim - W4TME
    Tim - W4TME Administrator, FlexRadio Employee admin
    edited March 2017
    Eric,

    We appreciate your concerns.  And I can assure you that the solutions we provide to issues are balanced and accommodate all users and operating styles.
  • Eric-KE5DTO
    Eric-KE5DTO Administrator, FlexRadio Employee admin
    edited March 2017
    They are FIR filters of varying size (e.g. 256-2048 samples).  Lots of effort has gone into optimizing the performance of these filters both computationally and to balance the sharpness (skirts) with latency/delay.
  • Tim - W4TME
    Tim - W4TME Administrator, FlexRadio Employee admin
    edited March 2017
    Since the context of this post was a problem report by Sergey (and may I say it included excellent information to define the issue), verified by others, subsequently acknowledged by FlexRadio and a resolution is forthcoming in the next release of SmartSDR, I am inclined to close this post.  An appreciative thank you to everyone who provided valuable and constructive feedback that allowed us to identify the root cause of this issue quickly.

Leave a Comment

Rich Text Editor. To edit a paragraph's style, hit tab to get to the paragraph menu. From there you will be able to pick one style. Nothing defaults to paragraph. An inline formatting menu will show up when you select text. Hit tab to get into that menu. Some elements, such as rich link embeds, images, loading indicators, and error messages may get inserted into the editor. You may navigate to these using the arrow keys inside of the editor and delete them with the delete or backspace key.