question for Flex 6K users doing JT65 or JT9

  • 1
  • Question
  • Updated 3 years ago
  • Answered
When I start SDR, I notice there is no DAX IQ channel selected:
Is that normal?
Photo of Bill-W9OL

Bill-W9OL

  • 726 Posts
  • 56 Reply Likes
  • embarrisingly stupid

Posted 3 years ago

  • 1
Photo of Ken - NM9P

Ken - NM9P, Elmer

  • 3875 Posts
  • 1180 Reply Likes
If you haven't selected one before on that band, which would be saved by the slice persistence, or have a profile with a DAX channel selected, then it probably defaults to "none."
Photo of Bill-W9OL

Bill-W9OL

  • 726 Posts
  • 56 Reply Likes
I don't really understand what the DAX IQ does in relation to JT65
As far as I can tell, WSJT-X woks whether I enable it or not.
Photo of WH6HI - Pat

WH6HI - Pat

  • 263 Posts
  • 35 Reply Likes
WSJT-X does not use I/Q signals as it represents a portion of undecoded band width and JT65/9 only uses the audio stream.  No need to activate DAX I/Q as you have discovered, is not used. 
Photo of Chuck

Chuck

  • 55 Posts
  • 0 Reply Likes
Bill,
I operate JT65 and JT9 using a Flex 6500, software is WSJT-X. You said your using DAX "I/Q", shouldn't you just be using one of your straight DAX audio channels as WSJT-X does not use I/Q input.

With my DAX channel selected and upon changing bands the DAX channel persistence does not hold. Known defect and is work in progress from what I understand from FRS.
73

Chuck
(Edited)
Photo of Bill-W9OL

Bill-W9OL

  • 726 Posts
  • 56 Reply Likes
No I didn't say I was using DAX I/Q.
I clearly said that WSJT-x works whether I enabled it or not.
and I also said
I don't really understand what the DAX IQ does in relation to JT65

I now have an answer. It doesn't do anything which is why I didn't detect any difference. So I will continue not to use it.
I was just curious as to whether it had some part.
Photo of Chuck

Chuck

  • 55 Posts
  • 0 Reply Likes
My apologies for not correctly reading your post.
Photo of George Molnar, KF2T

George Molnar, KF2T, Elmer

  • 1529 Posts
  • 541 Reply Likes
DAX I/Q streams are not supported by WSJT/X or JT-65HF (and variants), so having those streams engaged will not affect operation. Perhaps someday. Right now, your DAX audio streams (DAX1-8) are the best way to go.
(Edited)
Photo of Bill-W9OL

Bill-W9OL

  • 726 Posts
  • 56 Reply Likes
Thanks George, That answered my question.
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Actually
The 1.5-devel versions do support DAX I/Q. I'm trying that out right now. Seems to work FB,
after taking the audio level on the DAX-IQ stream. I've the IQ stream set to 192k, the level slider in WSJT-X set to the 6th index mark (+0dB), preamp in SmartSDR set to 0dB, on the waterfall in WSJT-X i have Zero, and Gain levels set at 0, the water fall is clear of any noise between the actual signals, Im seeing Average signal levels on WSJT-X level meter in the mid 40-50 dB range, AGC set to off, and I'm trying to discover where to set AGC-T at. AGC-T 0 - 100 don't seem to affect the waterfall much. I'm looking at AGC-T 100 right now and still only see signals and no noise in the waterfall. but lowering it to AGC-T to 0 appears the same. I'll try to see if i can determine if AGC-T has any effect on the decoding ability but AGC-T 0 is decoding everything JT65 and JT9 thats visible. I'm wondering if AGC-T is having any effect at all, it's almost like WSJT-X with DAX-IQ is the perfect, it just works automatically, audio setup. Very interesting,  WSJT-X 1.5.0-devel r-5007 is what I'm using, There is a Candidate (beta) release, Version 1.5-rc1 download availiable for download which is probably a newer version than the alpha version I'm using. I think think the wider bandwidth of the IQ audio, accounts for the increase in the WSJT-X level meter. BTW i've got the RX filter set to 5k in SmartSDR. The dynamic range in the waterfall palatte seems huge.  I just set WSJT-X to RX on DAX-IQ RX 1 and to transmit on DAX TX 1, switched DAX IQ 1 with the DAX menu on the left hand menu. and have DAX TX 1 enabled for TX on the DAX control panel. 
(Edited)
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Pat
I haven't been seeing any reversals like that. However after experiencing some ghost images of strong Signals from outside the passband showing up in the  WSJT-X waterfall I've Switched back to not using DAX-IQ today. If Ghost are getting aliased into the passband it's not worth doing it that way. All i saw of them were above 4k in WSJT-X. 
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
George 
As far as I know it's not documented anywhere, I might have tried it earlier if it had been. The DAX-IQ items had been visible in the Soundcard chooser, possibly all along. But they never worked before. I only tried it again because it was being discussed and got curious enough to try using them again. A good experiment, but maybe no useful result. I'll keep messing with it. 1.5.0-rc1 seems better than than the alpha version I'd been working with. I'll write something up and put it online once i get thru looking for new ways to tweak it. There's got to be some optimal way to juggle the WSJT-X input level, DAX-RX level, AGC-T level, and preamp settings. Some combination should result in more decodes and more DX. Repeatable results are what i'm trying to discover the possibility of not having to factor in AGC-T levels was what got me excited about it. SmartSDR and DAX can deliver whatever level WSJT-X expects. so I've simplified the setup so far by leaving the WSJT-X slider set for 0dB. I think I've stumbled onto a few new tricks , Explaining those tricks is another matter, but I'll share what I find when i know it works.  
Photo of George Molnar, KF2T

George Molnar, KF2T, Elmer

  • 1529 Posts
  • 541 Reply Likes
Thanks for the "experimenters notes " Jay. I've been using 5.4 kHz bandwidth and AGC SLOW here, with the slider about midway. Seems to work very well. For JT modes, at least. RTTY and PSK seem better with fast or no AGC and much narrower bandwidth.

Having fun experimenting!
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
George
Thanks for the feedback, I have been running WSJT-X with AGC turned off and with the threshold set @50. and DAX-RX on 20. the WSJT-X slider set centered for 0dB gain. that seems to work fine. But can I do better? Im trying  to keep the noise floor out of the decoders while trying to boost the weak signals as much as possible. Yesterday I discovered a new usage for the WSJT-X waterfall, Of course the waterfall has no effect on what gets decoded, but it does let me see what effect the preamp, and AGC-T level have on the "Character" of the audio being sent to WSJT-X. The AGC-T threshold with AGC on, controls the amount/percentage of the audio that is the noise floor. ie.  If the level is set too high the audio has more noise floor in it. That's nothing new, what's is new to "me" at least is using the WSJT-X waterfall to see the effect the AGC-T threshold is having, During the "quiet intervals" the WSJT-X waterfall should show little or no signs of the noise floor. I can see this best with the "Negative" palatte selected and the zero and gain waterfall controls adjusted to increase the sensitivity of the water fall. At any given DAX-RX level as you increment the AGC-T level the quiet intervals will be pure white bands, increment the AGC-T level another unit and specs of the noise floor start to appear. decrementing back to the previous unit and it clears back up. At that point, I think I've located the noise floor in the audio, Next, with the goal of maximising the weakest signals, i begin incrementing the DAX-RX level, which increases the loudness of the both the noise, and signals hitting the decoder. as I increment the DAX-RX level at some point the noise specks begin to appear in the quiet intervals again but the signals will all be stronger, decrementing the AGC-T threshold might remove more of the noise thats getting thru so at that point its basically turn one control up a unit and the other one down a unit, Alot like adjusting a manual antenna tuner. in other words I'm trying, again, to tune out any of the noise present while getting the signals as loud as possible. I haven't found any combination the 2 controls that might be overdriving the decoders yet. Like manual antenna tuners there are multiple combinations that acheive the clean/loud result. Once I figure out how to adjust to get that result quickly and repeatably, I then want to find out just how much noise the decoders can handle before the noise begins to affect the decoding. there has to be a "magic" "sweetspot". "Turn it up as as loud as you can get it" is not the answer I'm looking for. The answer is not missing any of the -25 and lower signals hitting my antenna.
Yesterdays, (all day) experiment was doing this with AGC set for fast, Today i'm going to do it all over again with AGC set to slow. and see if I get any better results. I happy with both the AGC off, and AGC fast results but I still haven't finished the experiment. Many times with AGC enabled a single loud JT9 will desense the rest of the passband, so I'm trying to factor that problem in too. 

Sorry for the inarticulate explanation, I'm still learning the details, learning is fun, and might even be useful. You never know until you try your hardest. There just might be a "magic" sweetspot. I realise this isn't a new approach. Everyone does this, this is just me doing it. I'm just trying to verify it's the right approach, and be able to explain why, it is.

I'm also sorry if this post is adding to the noise floor in the Community.

Jay - NO5J
(Edited)
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Correction: in my first post to this thread I stated, "The 1.5-devel versions do support DAX I/Q." . Which is completely incorrect. They can connect to a DAX-IQ stream to get audio, but it's really not a useful way of getting that audio. My apologies, if I'm wasting anyones time.
Photo of Bill-W9OL

Bill-W9OL

  • 726 Posts
  • 56 Reply Likes
I'm using the 1.5.0 rc1
When I tried to use the IQ, it really bolluxed up my settings.
It changed DIGU to USB and changed the bandwidth down.

I'll have to experiment a bit....As I only use voice when I actually have to for a ATNO, I'm not to up on that side of the Flex as it pertains to IQ.

Leaning more than one new thing a day is pretty much beyond my abilities.
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Well experiment then, If "it" changes something, change it back. Never let the "It's" win!. I Just started out tonight running WSJT-X with DAX-IQ, and with v.1.5.0-rc1, so I'm learning it too, for the first time. But it's still WSJT-X, just Improved. I usually approve of improvements, and try to stay current. But I lapsed for a couple of weeks with WSJT-X. I'm going to catch back up. If I know the answers. I'll share. I'll bet it's not documented, as usual, so I skip the reading and just do. So ask I might have figured the answers out. 
Photo of Larry da Ponte

Larry da Ponte

  • 159 Posts
  • 15 Reply Likes
What is the potential benefit of using DAX IQ over the slice DAX audio with WSJT?
Photo of Jay / NO5J

Jay / NO5J

  • 1402 Posts
  • 209 Reply Likes
Alex 
Are you increasing the passband on the upper end or the lower? the JT9 portion isn't usually that crowded and starting at XX075.9 might gain a few mistuned JT65 signals. Locally it seems the next mode up the band is popular from XX080.0 on up. so widening the upper range is pointless most of the time.   
Photo of Jay / NO5J

Jay / NO5J

  • 1402 Posts
  • 209 Reply Likes
Alex 
Your right, there are some ghost signals appearing above 4k sometimes. when the IQ sample rate is set to 192k, I'll try an IQ sample rate of 48k for a while, to see if it has any effect on them. 48k samples might mean Windows wont have try converting the sample rate. I'm not sure it does with the IQ streams though.
(Edited)
Photo of Jay / NO5J

Jay / NO5J

  • 1402 Posts
  • 209 Reply Likes
Alex 
The ghost images starting about 4k in the waterfall seem to be the best indications that my trick should be considered a failed trick. They're in the WSJTX waterfall but unseen on the SmartSDR panafall, they dont seem to be bother decodes below 4k but I have no way of knowing. Maybe I should inquire with Flexradio if there might be some optional way to tap a DAX RX audio ahead of the AGC and have that level set by the preamp. That might have the benefit of the simpler audio setup without the ghosts.  BTW when i select either R or L instead of mono in WSJTX the ghost seem to stop. they might return though I'll keep trying.
Photo of Alex - DH2ID

Alex - DH2ID, Elmer

  • 878 Posts
  • 157 Reply Likes
Jay, you're right of course, about the 192k sample rate.
I will look into the ghosting when I get back from Italy.
Tapping DAX RX audio ahead of the AGC would be a very good idea, but I'm not sure how that can be implemented.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
I just set the AGC-T control to "OFF" ... Does this not do what you want?
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 971 Reply Likes
The DAXIQ data is derived from a wideband receiver (Mbps) that is before any slice receivers -- no controls in slice receivers will affect DAXIQ.  No volume controls, AGC-T, DSP controls, operate on the DAXIQ data.  The DAXIQ data will be centered at the center of the panadapter.

DAXIQ data is, however, stuffed into a Windows virtual sound card for programs that need access to data through a sound card.  Because of this, Windows controls on those sound cards will affect the data.  So will the limited dynamic range of Windows sound cards.  For best performance, the digital mode software needs to be written to access the DAXIQ data directly from the radio before Windows gets it's grubby hands on it.  Note that every other radio's data that gets shoved into Windows sound cards suffers the same fate -- we just have a better way to access the data and get the complete dynamic range out of the data and avoid the possible resampling issues in Windows sound cards.

DAX data does pass through many of the slice controls -- tuning, noise mitigation, etc. all operate on DAX data.  We do bypass some controls that do not make sense for digital programs such as mic equalizer, processor, etc.
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Steve 
Is there any way to get a 0dB level out of a DAX RX channel and be sure it's a 0dB level?
AGC-T combined with the DAX RX slider leave me guessing. 
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
Well DAX is digital data.  There is no "0dB" level in digital data.  In the analog world, you typically specify a level than is some number of dB below your analog overload point -- a level that you prefer to receive.  In the digital world, you have so many bits to represent the data.  Let's say that it is 16 bits.  In the 16-bit world, each data sample can be a number between -32768 and 32767.  In the digital signal world, we would look at the number of dB represented here which would be 6dB per bit or 96dB.  The lowest number is referred to as "-96dBFS" or 96dB below full scale (FS). The number 32767 is called 0dBFS because it is full scale.  

The issue in the digital world is that in a composite bandwidth you will be receiving, there is not just one signal.  There may be multiple signals and there will be noise.  If your sine wave peaks at 16000 (-3.1dBFS) and another signal of slightly more strength appears and they periodically align (the peaks of their sine waves occur at the same time), the two numbers will add.  If the sum of these numbers is greater than 32,767, you have an overload.  Overload = distortion.  If the signals are weak, it's not just other signals you're worried about, it is also noise.  The noise will add just the same.  

So for a digital mode, you want to run at "some dB below 0dBFS that keeps you from going into overload."  As a radio manufacturer, it's hard to know this level.  Doing JT65, how would the radio know if your neighbor is going to show up with an S9+30dB signal at the same time you're listening to things at the noise floor?  We really can't.  But what if you turn everything way down where 0dBFS is ABOVE the overload of the radio itself?  This would make it where you could never overload from a strong signal because the radio overload would occur first.  Well this would work great if you had that dynamic range in your digital data system.  Unfortunately, Windows limits us to a small number of bits.  If we had 32-bit audio, this would all be a mute point -- the data would unconditionally never overload.  

So you need to look at the DAX control panel and keep the audio level well below the far right so that the data is never overloading.  This will give you the best SNR and keep you from overloading Windows sound cards.  (Alternately, convince your local digital mode programmer to accept native data not passed through Windows sound cards ;-) )
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Steve
Thanks, I knew it would be a complicated answer, My only goal was to feed WSJTX at what it thinks is a 0dB level and it Displays as a 30dB level (go figure) I've got 2 methods of doing that that seem to be repeatable, and are likely all I need. Thru trial and error and many late nights staring at 20m WSJTX with no audible or visible signals present so Im just hearing band noise. a setting of AGC off, AGC-T 50 preamp 0dB, and DAX RX set to 20. i get 30dB in WSJT-X as long as the slider in WSJT-X is set for 0dB gain. I'm usually close enough with the 50/20 setup. the other way I do it is again trial and error testing with my FLEX5000 I convinced my self that at 2am, empty 20m band noise from my dipole is +10dB stronger than my Dummy load. So to line up WSJT-X I switch to the Dummy load and set DAX-RX to the magic "20" and then adjust the AGC-T slider to get a 20dB (Usuallythats AGC-T 50) reading in WSJT-X which when switched back to the Antenna rises to 30dB on the meter. I'd be happy if SmartSDR could just ouput a 600Hz CW tone that would hit WSJT-X at a 30dB level.


I'd happily call that a Zero.

Still calibrating by the seat of my pants!
Jay - NO5J  
(Edited)
Photo of Jay / NO5J

Jay / NO5J

  • 1406 Posts
  • 209 Reply Likes
Steve 
Don't get me wrong Complicated, is just fine.
I can usually google my way to partial understanding, and thank goodness for hyperlinks and wikipedia. 
Photo of Guy G4DWV/4X1LT

Guy G4DWV/4X1LT

  • 1687 Posts
  • 387 Reply Likes
@ Steve N5AC
"So you need to look at the DAX control panel and keep the audio level well below the far right so that the data is never overloading.  This will give you the best SNR and keep you from overloading Windows sound cards. "

PMFJI with a slightly OT point, but I read somewhere that for FLDigi, it is best to keep the DAX-RX display as far to the right as possible without it always staying slammed up against the 0 end-stop. Is this a Bad Thing® too?

TIA.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
If you are using a narrow bandwidth and can ensure a single signal in the passband, this is a correct approach.  You want it as far to the right as you can without overloading.
Photo of George Molnar, KF2T

George Molnar, KF2T, Elmer

  • 1529 Posts
  • 541 Reply Likes
Interesting experiment. One note on using DAX I/Q then; since the I/Q data is centered on the pan adapter, not in the slice, it is possible that transmission and reception could occur on different frequencies. If the slice is centered in the pan, it's cool. But if the slice is move even a little in relationship to the pan, it could get confusing.
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
DAXIQ is not really designed for QSOs ... it's really for demodulating large swaths of the band with tools such as CW Skimmer.  If you're trying to conduct a QSO with DAXIQ as the receiver, it's probably going to frustrate you...
Photo of Jay / NO5J

Jay / NO5J

  • 1402 Posts
  • 209 Reply Likes
Update:
From the guy who was trying out DAX-IQ RX with WSJT-X. I've already reverted back to using a regular DAX RX channel, The Ghost signals appearing in WSJT-X's waterfall, that Alex reported are all the reason needed to not use DAX-IQ with WSJT-X. It just adds "noise" into the work flow for WSJT-X to attempt to decode. which it can't, because the ghosts aren't JT signals, they are only interfering with the real JT signals. and likely will affect decoding in a negative fashion. Ultimately the only benefit I could find was the possibly coincidental  0dB level that DAX-IQ was providing. It made it simpler to acheive a working audio level setup between SmartSDR and WSJT-X. It was worth experimenting with, but since decoded QSO's are more important than easy audio setup. Your better off adjusting the DAX-RX level, AGC-T level, preamp,and WSJT-X input level to acheive correct SNR. As always, there are usually several ways to setup anything. One of those ways will be the right way, and the others probably aren't. To find out which is which, you have to try them all. 
(Edited)
Photo of Steve - N5AC

Steve - N5AC, VP Engineering

  • 1007 Posts
  • 969 Reply Likes
The other key problem is that WSJT expects that the tuning for the radio that it sends to the radio will set the 0Hz point (carrier) in the received data and transmitted data.  If ESJT sends a tune command through CAT, it only tunes the slice and not the panadapter.  So the location reported by WSJT will not be where the signal actually is.  Then when it goes to transmit and reply, it will be on the wrong frequency.

Assuming that WSJT doesn't know how to digest IQ data, the "noise" is probably images of good signals.