AM Demodulation Quality Flex-6000 VS Others

Steve N4LQ

While using my 6500 to check into the local 75 meter AM net I became frustrated trying to copy the other stations. I experienced this with the 6300 also.
 
AM amateur signals show up on the scope with a huge carrier and tiny modulation and the audio seems the same. Selective fading is so pronounced as to make speech difficult to make out, weak and distorted. No amount of AGC-T, bandwidth or other changes seems to help. No, NR wasn't on!
Using the SAM mode helps some however even it lacks good quality. 

So during the NET I fired up a few other receiver to compare. First my 75A4 then the RCA AR-88, IC-706, R390A and finally the TS-2000. All provided much better demodulation and lacked the severe distortion from selective fading. Of course I was using the same antenna. FYI The TS-2000 did better than the others. The other stations's carrier acted to quieten the receiver noise, as it should, and speech was clear and understandable.  The Flex doesn't seem to respond to the carrier and quieten the band noise as the others do. It's as if the carrier isn't actually there. 

Interestingly, AM stations on the BC band sound pretty good, great in fact at least until you open the bandwidth over 6khz then you begin to get a lot of hiss. I can't really claim that this is any better than my 70 year tube receivers however. 

So my question is....What have others experienced in comparison to other receivers? Is there hope for improvement with SSDR's demodulation ability? 

CW and SSB are just awesome and far better than the others, even my K3 (Which BTW also stunk on AM). Sorry to say, AM just doesn't cut it.
73

W7NGA

I'll disagree. I used to design receivers. I have a Collins 75S3C, Collins 75S3, multiple Collins 75A4s, Drake R4Cs, Drake R4, Collins R390A, National HRO50T1, NC183D, INC98, SX24, Icoms, Yaesus ... nothing compares to my Flex 6300 for AM demodulation. And then we also have SAM ...

dan  W7NGA
San Juan Island, Wa.

Steve N4LQ

I'm glad to hear that so what am I doing wrong? 
Seems like AM was even better when I had the Flex-1500. 
Any suggestions? I may resort to make a video for demonstration. 

Ken - NM9P

Where are you setting your AGC-T?  Even on AM you need to set it on a clear frequency to the "knee" or "Sweet Spot."  Many Flex 6000 receive problems are caused by having the AGC-T set too far to the right so that the quieting effect is not optimized.

Also, I usually use FAST AGC settings so that a quickly fading carrier won't leave the signal in the quiet zone for too long.

However, I have sometimes wondered if on AM the AGC could be set so it doesn't kick in for a few more dB, in effect being linked more to the modulation than the carrier?  That might remedy some of the "low AM audio detection" issues.

Steve N4LQ

Ken
While I agree with your suggestions, believe me, this has all been tried. Also I would suggest that one should not have to seek a "sweet spot" just to receive such a simple mode but of course, as I said, I have tried all manner of AGC-T settings and fast vs slow agc for the past year but still the old timer receivers provide cleaner AM on ham signals. 
I mention "ham" signals because on the AM broadcast band you can't beat the 6000 but for some reason, hams with their weaker signals and lower modulation percentages seem to be weaker and distorted. Especially bad is the phase distortion from fading. Sure you get some of that on regular diode detectors but it's certainly terrible on this version of SDR. The TS-2000 sounds just fine, not hi-fi but quiet readable. 

N7CXI

The hiss you're talking about is probably due to the 6000's flatter response. NONE of the older receivers had anything even approximating a flat response.

Have you tried setting an RX EQ curve that rolls off 3-6dB per octave above 3kHz? That might be a start on getting you there.

HTH,
Jim N7CXI

k0eoo

I'll throw my log on the fire here Steve....  I have many high quality AM receivers in the shack and none of them equal the consistent AM audio quality I get from my 6500 has as well as the SDR-1000 or Flex-5000 I owned....  So, you might have something amiss with one of your settings....  

Use Fast decay with AGC-T at just above the noise level (30-40 on 75m) and start with flat EQ and the slice audio gain at about 60-70 and the global audio gain set to suit....  YMMV....

By the way, I'm a retired design engineer for Audio Research Corp and enjoy the superior audio quality the Flex radios provide, especially on AM and SSB...  Good luck, Dennis, k0eoo

k0eoo

There should be no "search for the SWEAT spot" necessary for AM....  To go from SSB to AM all I do is change the AGC decay from FAST to SLOW or MED....  Something must be wrong with something else in your setup if you have tried the setup Ken or I put forth...  Wish I could visit your shack to hear what you're hearing....

Steve N4LQ

AM from BC stations and hams using 100% modulation and decent signals sound awesome on the 6000 but when someone is using a rice box or low level screen modulation it's really hard to decipher. The older receiver run rings around it. 
I'm trying to find time to set this up so I can record a video. I may also run some test using the signal generator and various levels of modulation to see if I can pin this down. 

Steve N4LQ

Here is a video I made to attempt to show the difference between the 6000 and other receivers on AM. 
Note: Same speaker is being used via separate inputs on the Bose. Same antenna.
No amount of juggling the AGC-T or EQ helps. 
https://www.youtube.com/watch?v=2HBg6uc8CNY

N7AIG

Sounds to me like the Kenwood has a high-frequency emphasis that you may not be aware of. Even your own voice as narrative lacks the crispness and clarity of the Kenwood in the film clip.

Most natural sounds have a roll off of around -6 to -12 dB / octave above 1 kHz. Voices have a dead zone between the first and second formants around 1 kHz.

So in my own systems here, running through the EQ of a MOTU 828mk3, I use a low frequency cutoff of 100-200 Hz, a high shelf filter of about 6 dB boost above 1 kHz, and a Q3 notch filter centered at 1 kHz. (instead of boosting the highs, I actually cut the lows with a low shelf filter by -6 dB at 1 kHz - that helps avoid overdriving the output)

73 de Dave, N7AIG

Steve N4LQ

The Kenwood has 3 preset EQ positions in the menu, Bass boost, F. Pass and High boost. F. Pass is set to emphasize the vocal range which is where I had it set. High boost is really crisp. You can also set custom levels via the computer.
All this is beside the point when it comes to selective fading distortion however.

k3Tim

Hi Steve,

A few questions if I may.  In the video the Flex has the filter bandwidth (passband) reduced compared to the signal showing up on the Panafall.  How many kc. wide was signal and the filter setting?  The original query states the AM on Flex sounds bad, somewhat implying SAM sounds fine,  Is this inference correct - ie. SAM sounded good on the Amateurs AM signal?  I assume you've tried the broadcast AM in MW and HF and results for AM versus SAM were?  
For the other receivers do you know the exact method of AM demodulation?

With the update to SAM, setting the RX EQ, and running the audio thru PC to perform equalisatiion, I was simple blown away by Japan Radio broadcast booming in from the Pacific in the early morning hours.  Their signal is 12Kc wide, and the SAM de-mode makes it sound like studio quality.

_..--
 TiM  

Steve N4LQ

Tim: Thanks

Filter width: Varied throughout the video from maybe 4 khz up to 8 khz. This has no effect on "selective fading distortion" but certainly on the fidelity. The TS-2000 uses a separate hi-cut and low-cut passband adjustment to set the bandwidth. I had low cut at 200 khz and varied the hi-cut from 4 khz up to 5khz which is max.

I only compared AM to AM since the TS-2000 does not have SAM however I can tell you that even in SAM mode I still experience bad distortion with the smallest amount of fading. I had first referred to listening to AM guys here within 100 miles on our 75 meter AM net during the late afternoon. SAM helps some but not totally and is somewhat bothersome to use in a net where everyone is using boatanchors and as much as 5 khz off frequency. 

Again I will say that the Flex/ssdr sound very nice on the AM BC band for whatever reason I can't explain. 

I haven't listened to Japan lately so I can't comment. 

Finally....as I mentioned previously, I can use any of my old receivers and they sound better than this however the lowly TS-2000 does sound better on AM than about anything else I've got. 

Steve N4LQ

N7CXI

Steve,
Any chance you could cut the Flex RX EQ at 63 and 125 to the minimum, and see if that has any effect on the distortion you're hearing?

Thanks,
Jim N7CXI

Steve N4LQ

Jim
No....Cutting the LF audio doesn't improve the distortion however......

I was able to make a big improvement by simply eliminating one sideband while in AM mode. IOW...Set the bandwidth to 6 khz, center the AM signal in the passband then grab the edge and pull one of the sidebands in toward the carrier thus eliminating it. Most of the selective fading stops and audio fidelity is unchanged. I assume the sidebands are out of phase due to fading, which is typical, however the net effect is more drastic in the 6000 than in most receivers. 

Doing this enabled me to copy a weaker AM ham on 3590 khz just 5 khz away from a huge AM guy on 3885 khz. The TS-2000 was totally swamped by the stronger station, K4KYV who runs a virtual AM broadcast station!

Well this is a workaround and more needs to be done.

Steve N4LQ

N7CXI

I don't know how AM is demodulated in SmartSDR - I'll guess just the magnitude of the I&Q stream if it's like other common DSP demodulators. It will be interesting to see how AM is demodulated in the TS2000. If it's done in hardware, I should be able to find it in the schematic. Thanks for the detailed testing - you've made an interesting set of observations.

N7CXI

I had a look at the TS2000 service manual. The AM demod is done in the DSP, so there is no opportunity to learn how it works.

Since the 6000 series and SmartSDR have the benefit of much lower overall distortion, (including no filter phase distortion) there's no reason our 6XXX radios couldn't have the best demodulators available. The problem may be identifying what's the best and implementing it. It's an interesting problem to have, since it could be fixed as easily after-the-sale as before.

As an active AM experimenter and founder of two DSP engineering companies, I suppose I should look into it. Overcoming entropy (laziness) is usually the problem around here, at least... ;-)

73,
Jim N7CXI

k3Tim

Laziness is a good motivator...

I checked the TS2000 manual also  

{   Google:  "TS-2000X_Service_manual.pdf"    }

and agree with you on AM detector / DSP.  

On page 15 it says:

In AM mode, a 455kHz signal passes through the AGC amplifier (Q51) and amplifier (Q48 and Q45) and is detected by D58. The detection signal retrieved for the AGC is rectified, passes through the DC amplifier (Q39) for AGC control and goes to the Q37 gate terminal (G2).

but believe this is referring to the AGC voltage under AM mode.

From "Digital Signal Processing in Communication Systems", by Marvin Frerking (aka. W0EKC) one can demodulate the AM signal by:
   
     demod-AM = sqrt(I^2 + Q^2)

----
The text is recommended for anyone wanting to know the digital processing details.  Be advised a "little" math is involved...

_..--
k3Tim 


PS:  listened to a two hams running AM in AM mode and never heard distortion.  On SW bands don't hear it either except when there is a deep carrier fade, as one would expect.

N7CXI

"   demod-AM = sqrt(I^2 + Q^2)"

That's the IQ stream magnitude I was referring to. It's the most basic form of AM demodulator. Easy to code.

Burt, SmartSDR 1.4.3 changed the game for the TX EQ. I have mine reworked to a degree, but still have lots of room for improvement. It's at least possible that the image I attached will show up... Your mic and voice will no doubt be different, but I've had compliments with this curve and the PR781.

Jim N7CXI

James Whiteway

Bert, I think using the built in profile for the FHM mic may not always work as well as setting things up. I have a Heil PR781 and have used the included profiles with it, and "most" of the time, everyone says it sounds like me. But, I have also found that with some careful tweaking of the settings, I can get a clearer signal that I have been told, sounds more natural. I think the profiles that are included in SSDR, should be viewed as a starting point and not a carved in stone absolute. Sometimes, depending on the person's voice, the default profiles will work fine. Other times, maybe not.
james
WD5GWY

Ken - NM9P

It is only a matter of getting the adjustments right. The 6000 is capable of excellent audio even on AM. But you cannot use the processor at anything more than NOR. And it might be best not to use it at all. And getting the EQ right takes work. The default profile may not be right if it wasn't changed with the 1.4 update. The EQ characteristics were changed with 1.4.

k3Tim

Having a local receiver helps one in setting the mic/equ up (statement of the obvious).  Even a Softrock SDR should be good enough.  

I tried the AM transmit mode with input to BAL and it sounds fine in local receiver.  Same with FM / SSB.

k3Tim

KY6LA_Howard

The new Slice Record and Playback function works rather well... it now includes the ability to record Equalized and Processed Audio from your mike and from DAX.   At least with SSB (I have never tested AM) it effectively alleviates the need for a second receiver to set up your audio

N7CXI

No, it doesn't work on AM.

It's just OK, though. We AM experimenters are a rugged bunch - we don't need no
monitor nohow. ;-)

Jim N7CXI

Tim - W4TME

I have entered this problem report into our bug tracker for additional investigation. Thank you for the defect report.

Barry N1EU

I'm a little late to this thread but that video seems to show the TS-2000 set for a 4KHz audio passband and the Flex set for a 2KHz audio passband (after you narrowed it with the mouse early in the video).  That's the fundamental reason the Kenwood sounds so much better.

Barry N1EU

Steve N4LQ

Barry etc. I started with 6 khz bw on the Flex then 5.6 khz and tried various widths by dragging the bw bar. "Sound" or hi-fidelity isn't really the issue here. The issue is susceptibility to phase distortion due to selective fading. Actually, setting the Flex for a wider bandwidth exacerbates the problem while dragging one side of the passband toward zero tends to improve things slightly, having eliminated one sideband  

A question I have is...What role does the received carrier play in the demodulation process? In a typical diode detector the relationship between carrier strength and sideband strength is important. If the carrier fades below the sideband strength, overmodulation occurs and the resulting audio is distorted however what goes on inside a DSP chip is beyond my pay grade. 

 

N7CXI

We don't know what flavor or variation of AM detector is used in SmartSDR, *so what follows is speculation.* 

The "AM" detector *could be* a simple "AM = sqrt(I^2 + Q^2)" magnitude detector, as described elsewhere in this thread - an envelope detector in analog terms. In this case, the relationship to the carrier is only that found in the signal - subject to selective fading and all other forms of phase distortion.

Beyond the textbook magnitude detector we venture into the realm of pure speculation, since details vary widely between implementations. Most of the more sophisticated detectors use some variation of a synthesized carrier, phase-locked to the signal carrier. (SAM) The more sophisticated ones make some attempt at identifying/correcting selective fading by comparing the integrated magnitude of each sideband against both the signal carrier and the synthesized carrier, as well as each other. That computed relationship is used to modify gain envelopes that control the relative levels of the carrier, synthesized carrier and the sidebands before the actual demod. The detector can also usually be switched between modes to force LSB, USB or DSB or "auto" sideband detection modes. "Auto" mode allows the detector to select which sideband to use based on the above calculations, as well as user input for threshold parameters.

My reason for this extended speculation is to point out that AM detectors can be very simple or very complex. For usn's amateurs, the beauty of it as deployed in "distributed" SDR schemes like the Flex radios is that both firmware and software can be updated after the fact based on the development priorities of the vendor. In the case of SmartSDR, I believe it's even possible for a third party to sign up for a slice IQ stream and perform custom demodulation on it. Now whether that can be routed back to the hardware audio I don't know, but if not it seems to me that DAX or even a direct connection to the PC audio device could be used to work around it.

In other words... If we don't like the AM demod, some enterprising young tinkerer could build one we do like, even if Flex doesn't give it the priority we would prefer.

My .02,
Jim N7CXI

Steve N4LQ

Perhaps we should implement the 1N34 mod?

N7AIG

ROFL!! I love it...

73 de Dave, N7AIG