AM Demodulation Quality Flex-6000 VS Others

Steve N4LQ

Or perhaps we can find room for a 6AL5 and use the extra diode for hardware AGC.

N7AIG

I have been thinking about the question posed earlier, "in a DSP, how is the carrier dealt with?". I concur that the typical implementation of AM detection is normally by way of envelope detection, formed as the RSS of the I and Q amplitudes. In that kind of situation, the Carrier normally appears as a DC offset, and is simply discarded. 

Another viewpoint is that you are forming the square root of the conjugate squared signal, which in the Fourier domain is a convolution among all the sideband spectra on both positive and negative frequency sides, and the Carrier, which is a delta function at DC. 

Either way, the Carrier sort of falls out as unimportant and even extraneous. I have to think back real hard to remember why it was ever considered important...

73 de Dave, N7AIG

N7AIG

sorry, my bad...  The conjugate squared signal is an autocorrelation in the Fourier domain (not a convolution), and that carrier is used to sweep across the sidebands to produce a nice strong audio component. So the carrier really does take part in the calculation. You'd want the carrier to be much stronger, (e.g., 4x?) than the sidebands in order for its contribution to the autocorrelation to stand out. If you don't have a carrier, then you have to inject one yourself.

N7CXI

4X is the standard recipe, yes.

Where it gets tricky is that one or the other of the sidebands will often be selectively faded or phase distorted, and the carrier may or may not have adequate SNR against the sidebands to correlate with anything. This goes back to the issue at the top of the thread. If you choose to address that, then various time-domain tricks are normally employed.

N7CXI

You also can't assume the frequency of the carrier is in the middle of the filter. The synthesized carrier has to correlate to the signal carrier when the SNR is good enough to do so precisely, but not update when it isn't. With the SmartSDR "SAM" demod you can be off to the side a bit on an AM signal using the "AM" demod, then click "SAM" and listen to it slide over to the signal carrier frequency and lock. If the signal carrier fades away, the SAM carrier just stays put until the SNR comes back up.

Steve N4LQ

Speaking of SAM. I notice hearing the signal slide over to the carrier frequency at times but not visibly. The scope's carrier stays put. I can drag the passband around but I don't hear the carrier return. I don't think PowerSDR worked quiet this way.

N7CXI

I wouldn't expect to be able to see the SAM carrier on the panadapter, but I don't remember how PowerSDR worked.  Out of sight, out of ... what was that? ;-)

Steve N4LQ

I'm willing.  
Email me steven4lq@gmail.com

n4lq

N7CXI

Burt, If you haven't resolved it by the weekend, I'll be home by then and happy to work with you. email is barberaudio@gmail.com . Jim N7CXI

Norm - W7CK

I have never gotten good audio reports from my 6700.  Its amazing that an average microphone can sound good on so many different rigs, but plug it into a Flex and you have to muck around forever just to get it to sound acceptable.  I think maybe the software still needs some fine tuning in this area.  I have to have different profiles for AM, FM and SSB - doesn't make too much sense.  We'll see what the future holds....

DH2ID

I get very good audio reports for my 6500 with a YAESU MD-100 microphone.

Ken - NM9P

Burt,
I am pretty much occupied all week, but Saturday Afternoon I might have some radio time and perhaps we could do some tests if you still need it.

Ken - NM9P

Steve N4LQ

Burt and I worked on this for about an hour yesterday. We about wore out the virtual sliders.  It seems like there is some degree of distortion no matter what we tried. Possibly he has some RFI on the mic cable but we didn't get into that. 

Ken - NM9P

Burt, I don't remember from your earlier posts...
Have you had the factory check/correction for the mike connector grounding issue done?  Perhaps, if you have the issue, yours only is showing up on AM?  Just a thought.

Steve N4LQ

Tell us more about that Ken. News to me.
Also we are running some issues together. Burt and I tested on SSB only and this thread is about AM receive so we've deviated off track a bit. 
Good advice though. 

Bob-N4HY

Steve:    Sorry to have taken so long to respond.  I am assigned this issue by Flex as I am the architect and original author, back on PSDR, of the original demodulation approaches we have taken to AM, both standard HPF (sqrt(I^2+Q^2)) which is remove DC to low cut off, as well as SAM (which is phase lock to the carrier, but otherwise demodulate as if double side band).  SAM is mostly immune to the overmodulation which occurs when selective fading wipes our the carrier or reduces its power well below that of the sideband power.

I have listened to your fantastic A/B of SSDR and Kenwood TS-2000.  I hear, with my own ears, two things.  One the AGC-T is set much better in the TS-2000 and finally, there is a major difference in audio response between the SSDR and TS-2000. 

I am going to suggest a way to settle this and give us actual measurements.

I would request that you download and install SpectraVue if you do not have it already and attempt to give us as close in time as possible A/B comparisons of the AUDIO OUTPUT of the TS-2000 and the SSDR running through SpectraVue.  If you could do the youtube process again it would be most helpful.   I apologize in advanced for all the extra work and thank you in advance for bringing this issue to us and understaking the measurements.   Some are going to LOVE the nearly flat group delay, amplitude response and other things that are the staple of the Flex system.  Others, such as you,  will prefer that which they are accustomed to.  For Flex, "It's only software" but any such modifications have to folded into the build, test, roll out, deal with unintended consequences of the changes process Flex must go to with its finite software resources.
 
73s, respectfully,
Bob N4HY

Steve N4LQ

Hello Bob and thanks for the interest. 

A spectrum analysis would be helpful I suppose. I tried to get SpectraVUE working but got too frustrated with all the hoops.  You probably have another receiver there you can do the A/B testing with and get better results. 

I will say that there is nothing particularly special about this TS-2000. It happened to be handy to make the video. I get the same results when using my WWII AR-88.

Perhaps I could hook up a O'scope and demo this better however take a careful listen to this video I made. BTW: I am using an Android phone for the recording so all you can do is listen to each receiver and form an opinion. The scope can come later.

https://www.youtube.com/watch?v=rab-LwAL34w

 Steve

k0eoo

Hello Steve,   (I notice Bob is looking at this issue but I just wanted to pass on my observations on your video while we are waiting his findings....)

I finally took a look and listen to your YouTube video, nice job by the way....  First thing I noticed is you had the bandwidth on the flex set too narrow.  You were listening to Robert, W0VMC and is signal was 10kHz wide (+/-5kHz) and the flex window showed about 4kHz (+/-2kHz), much too narrow!  On the TS-2000 you had 4kHz set which is 8kHz bandwidth in AM mode.  So we are comparing 4kHz BW on the flex with 8kHz BW on the TS-2000.....

I usually run my AM BW at 8 or 10kHz depending on the width I see on the panadapter, and the EQ as shown in the clip:



In addition I noticed you had the Rx EQ set to peak in the mid-band and be flat in the lows and highs; so you were lacking in amplitude in the presence band which helps in clarity.....  But, because you had the BW set too narrow to hear Roberts mids and highs, EQ becomes a mute point....

Dennis, k0eoo

Steve N4LQ

Nice try Dennis but believe me none of these has an effect on phase distortion due to selective fading. 

The TS-2000 was set for a "HI-CUT" of 4000 hz. The total bandwidth is actually 8 khz or 4 khz each side of zero-center frequency. I can open it up to a max of 5 khz HI-CUT and with the Lo-Cut set for zero we have 10 Khz max BW.

Flex's bandwidth indication showing 10 khz is really 10 khz total or 5 khz on each side of center. 

Adjusting the EQ is lots of fun but it cannot compensate for distortion. If you notice, I made several adjustments while filming and I tried to maximize the vocal range between 300 and 3500 hz. 

Ken - NM9P

Steve,
I just watched your video.  I agree that you have more phase distortion on your Flex than on the TS-2000.  You have SOME on the Kenwood, but it is not as often and not as deep as on the Flex.  Hmmm.

I would like to hear more, perhaps with the RX EQ flat.
Also, could you test with AGC OFF but adjusted so that it isn't overloaded?
It may reveal whether the distortion is in the actual demod routine or the AGC routine.

Ken - NM9P

UPDATE:  I also noticed that the distortion is more prevalent in the lower registers on the Kenwood, but is more in the higher frequencies on the Flex....

Steve N4LQ

Ken
I just made another video demonstrating the clipping in AM mode. Really I should use an o'scope but I think you will be able to hear it. 
Remember the old receivers with diode type "noise limiters"? Well this is about what it sounds like. Perhaps this will offer more insight.
https://www.youtube.com/watch?v=rab-LwAL34w

k0eoo

Steve, Where in the video did you show the Flex BW set to 10kHz?  The Flex BW did not cover Roberts whole signal.  I could not see the scale on the Flex screen and from the panadapter you can plainly see the highlighted blue BW area did not cover even half of Roberts transmitted signal, so as far as I saw and heard the comparison was not apples to apples.....  

Steve, I don't want to point fingers, I'm just a customer as you are and everyone's perceptions are valid even if we don't all agree....  Flex is looking at the issue so if something is amiss they'll find it I'm sure....

Regards, Dennis, k0eoo

Steve N4LQ

Dennis:
With the Flex bandwidth set for 5.6 Khz the carrier is centered within the passband and there are 2.8 khz bandwidth on each side. Only one sideband is necessary for demodulation of AM. The Kenwood TS-2000's bandwidth is indicated as lo-cut and hi-cut. lo-cut was set to 0 and hi-cut to 4000. That gives a total bandwidth of 4 khz or 2 khz for one sideband. The maximum I can set this for is 5 khz or 2.5 khz for one sideband. That's one reason the Kenwood sound a bit bassy and the reason I reduced the Flex to 5.6 khz with is the most narrow pre-set bandwidth on Ssdr in AM mode. 
That is pretty close to apples vs apples. So in reality, the Flex was set slightly WIDER then the Kenwood. As for Robert's signal width. I can't do much about that. 3 khz one side of zero is plenty = 6 khz total. 

Steve N4LQ

I have a postulation:
The 2 sidebands become out of phase with each other and produce distortion. 
Kenwood demodulates only one sideband. Flex demodulates both. 
By eliminating one sideband we reduce the opportunity for distortion. 
Please view these 2 videos.

Typical AM BC signal with one sideband removed
https://www.youtube.com/watch?v=3G0v9Uftdx8

Ham on 75 meters with one sideband removed
https://www.youtube.com/watch?v=8fbFeVKnd8c

Bill -VA3WTB

I wish I knew what is going , I read were many people think Am mode is really good, but not for you

Steve N4LQ

Trust then verify.

N7AIG

Not quite sure what you are calling "phase distortion"... but I did the math. The old formula of taking the square root of the I^2 + Q^2 signals *only* works properly for "analytic signals" -- effectively a single sideband signal. This is the classic envelope detector.

When you feed a double-sided spectrum into a classical envelope detector, you invariably generate demodulation products at the expected DC and audio frequencies, and *also* at double the audio frequencies. That has to sound incorrect. That does not happen if you feed only one sideband through.

So your approach of feeding only one sideband through the AM detector is, in fact, the correct thing to do.

73 de Dave, N7AIG

N7AIG

... or else, you should send the I & Q signals through a Hilbert transform and combine with the original signal, to generate an analytic signal ahead of the envelope demodulator. It's easier to just feed one sideband through. It also avoid potential phase cancellation. (phase cancellation due to differential propagation, but not phase distortion).

Ken - NM9P

Perhaps the routine could be modified to decode only one sideband, but have a button that allows the user easily to select which one, depending upon interference?

N7AIG

Furthermore, the classical envelope detector really only works properly when only a single tone is present in the modulation. Anything more generates sum frequency components at reduced levels, equal to their amplitude product.

The spectrum of an envelope detection is the autocorrelation of the signal spectrum, and so you can expect a low level fuzziness to any real signal demodulated in this manner.

The very best you can do is to use SSB demodulation with a synthetic carrier, but then you have to get the carrier frequency just right to avoid the Donald Duck sounds. But this generates no spurious signals in the floor of the demodulated sound.