• Get involved.
    We want your input!
    Apply for Membership and join the conversations about everything related to broadcasting.

    After we receive your registration, a moderator will review it. After your registration is approved, you will be permitted to post.
    If you use a disposable or false email address, your registration will be rejected.

    After your membership is approved, please take a minute to tell us a little bit about yourself.
    https://www.radiodiscussions.com/forums/introduce-yourself.1088/

    Thanks in advance and have fun!
    RadioDiscussions Administrators

Multipath Simulation for SSB FM Stereo

I realize this thread has become a long one. For those of you with interest, thanks for sticking around.

My desire in this debate has been to determine if there is validity of the stereo separation measurement method Brian describes in his paper on SSB. This is where he plots the composite response of a tuner at the output of the IF section. Then, he plots stereo separation based upon the mathematical matrix in the decoder. I found this approach interesting, and have wondered if this method would reproduce the same results as you would find from the complete, end-to-end transmission system method.

The stereo generator inside Omnia.11 was designed in MatLab, and it offers an easy way to generate DSB or SSB mpx signals. Following describes a test employed to recreate Brian's IF response method, and to see if stereo separation results support his findings.

I used the composite response he provided in his paper, and used that to model a linear low pass filter in MatLab. This filter model assumes the same coefficient values as he described. I used a linear filter, as this would replicate proper recovered coefficient levels, as he recovered at the IF output. This removes any phase anomalies from the exercise.

The model for the above described stereo generator was used to generate the mpx signal. The mpx decoder was also created in MatLab in order to recover the left/right channels, and measure stereo separation. The low pass filter, modeling the output of the IF section Brian describes, is inserted between the stereo generator, and decoder, along with a bypass switch, which is used to compare the effects of the filter (or any other simulated IF effect). This will simulate what effect the slope of the IF section will have on stereo separation. Additionally, the stereo decoder was fitted with a "coefficient correction" function, the same as Brian describes in his tuner, such that separation for a narrow IF can provide maximum separation at 1kHz.

The stereo separation results from this test do not parallel what Brian describes in his paper. Even when varying the value of the "coefficient correction" function, the separation results never approach his findings. I found this to be true for DSB and SSB modes. Actually, the SSB mode offered a bit better separation at high frequencies, compared to DSB. I surmise this is due to the LSB spectra being closer to "flat" as compared to the broader spectra required for DSB mode at high frequencies.

These results closely parallel what I found recently when doing the same tests using a stereo generator, exciter, and the same Sony receiver Brian refers to. I've read his paper numerous times, and he has privately described his method to me over the last year and a half. I understand his method, and what he is trying to illustrate.

Using the MatLab model was a way to recreate the end-to-end scenario, and it took into consideration the IF slope effect he describes. This MatLab test provided a way to generate the same altered mpx spectra as presented to a stereo decoder. The recovered left/right separation values were compared to what he gathered using his method. The MatLab results vary significantly to his posted results, and this leads me to believe there is something askew in his method.

-Frank Foti
 
Great explanation Frank, and good way to end this discussion on simulated tests. I would fully support a new discussion in the future on "actual" SSB testing, using real FM transmission systems. The availability of stereo generators with SSB will hopefully increase as the price point of DSP continues to fall. In turn, we should begin to see greater feedback from "real" users.
 
Frank, I don't understand. This entire thread you've complained about models and simulations and how they are not real world, yet now you're comparing one simulation to another to try and discredit Brian. What gives?
 
Not to speak for Frank, but reading this thread it is this:

Frank modeled the entire system, Brian modeled one portion of it.

It's like modeling the wing of an airplane but neglecting to model how it behaves when attached to the entire plane.
 
Yes, but a model is still a model, so you can't discredit using models by using a model yourself. I have read Brian's work over the course of a few years on tuner design and I can honestly say that there are maybe a handful of people who know what he knows. Frank is a processing guy, so just as I would not question his ability to create a distortion cancelled clipper, I wouldn't question Brian's models on how receivers behave.
 
Turnpike Tuner said:
Frank, I don't understand. This entire thread you've complained about models and simulations and how they are not real world, yet now you're comparing one simulation to another to try and discredit Brian. What gives?
Click to expand...

If you've read the entire thread, you will notice I did an end-to-end real world test, and the results differed significantly from Brian's. Instead of just hanging my hat on those results, I wanted to see if there was another way to determine if his method holds merit. Replicating the end-to-end system in MatLab provides such the opportunity, along with an additional means to look at any other possible anomalies within the system. The results of the model and real world system are quite close to one another.

The simulated model was done to confirm and further determine if Brian's method was valid. Through this entire discussion, I've asked to see a comparison of results between his method, and the complete end-to-end system. The MatLab simulation enables the ability to put his test data into the model and see if the results equate to his results. Bottom line is the end-to-end test results in MatLab, using his gathered data, do not equate to his results.

Providing two end-to-end test environments, both of which produced the same results, further indicates that something is askew in his method.

-Frank Foti
 
fm-engineer said:
Great explanation Frank, and good way to end this discussion on simulated tests. I would fully support a new discussion in the future on "actual" SSB testing, using real FM transmission systems. The availability of stereo generators with SSB will hopefully increase as the price point of DSP continues to fall. In turn, we should begin to see greater feedback from "real" users.
Click to expand...

For those who may not be aware, this topic will be discussed in great detail at the NRSC meeting in Las Vegas. Also, we will be making available a stand-alone stereo generator that offers DSB and SSB modes. This will be priced within reason.

-Frank Foti
 
Turnpike Tuner said:
Yes, but a model is still a model, so you can't discredit using models by using a model yourself. I have read Brian's work over the course of a few years on tuner design and I can honestly say that there are maybe a handful of people who know what he knows. Frank is a processing guy, so just as I would not question his ability to create a distortion cancelled clipper, I wouldn't question Brian's models on how receivers behave.
Click to expand...

The model was done after doing the real world end-to-end test. The actual test results do not agree with Brian's results. The results of the modeled end-to-end system paralleled those of the actual system.

The MatLab model was created to enable test results that a real world system does not offer. This enables the ability to look at test results over a wide range, although the goal, here, was to see how the system responded using Brian's gathered data.

I'm not discrediting the use of a model. Most of our own designs began as models, but then we put them to task in the real world. I have all along questioned the modeled test results, when they were never compared to actual system results. In the end, I went the added mile and did both, actual and modeled testing. In both cases the results do not support his method.

I can easily see that Brian's an extremely knowledgeable person with regards to receivers. I do respect his views and knowledge. I hope he intends to assist with any ongoing SSB testing, as we'll see more activity on this topic this year. Although, on this separation topic, there is a disconnect somewhere, and that is what this discussion was about.

-Frank Foti
 
Multipath itself is not caused by enhanced L-R, but it's the dramatic blend of the receiver to mono which is the annoyance. I think what you're seeing is SSB causing many of these receivers to reduce stereo separation because something is wrong with the transmitted signal and, in turn, the blend when there is legitimate signal issues is not as dramatic.

I own the Sony CDX-GT700HD car stereo. It has one of the best FM sections I've ever heard in an auto environment. It, like the Sony tuner, is also subject to the abnormalities of SSB transmission because they have the same chipset.

Multipath is not a stereo issue, it is an FM issue and I think that's lost here. Just my $.02
 
Turnpike Tuner said:
Multipath itself is not caused by enhanced L-R, but it's the dramatic blend of the receiver to mono which is the annoyance. I think what you're seeing is SSB causing many of these receivers to reduce stereo separation because something is wrong with the transmitted signal and, in turn, the blend when there is legitimate signal issues is not as dramatic.

I own the Sony CDX-GT700HD car stereo. It has one of the best FM sections I've ever heard in an auto environment. It, like the Sony tuner, is also subject to the abnormalities of SSB transmission because they have the same chipset.

Multipath is not a stereo issue, it is an FM issue and I think that's lost here. Just my $.02
Click to expand...

Multipath is indeed an FM issue. But, as clearly stated in prior posts and articles, it is exaggerated by FM-Stereo based upon the added spectra of the DSB system (23kHz - 53kHz). With stereo content, the spectra of the L-R (23kHz - 53kHz) adds to the RF channel bandwidth, which adds to the sensitivity of multipath. When SSB is transmit (23kHz - 38kHz), the channel bandwidth is reduced, and the sensitivity to multipath, when stereo content is transmit, is also reduced. This has now been confirmed by a few dozen FM stations, here in the USA and abroad.

BTW: We also have the same Sony CDX-GT700HD car stereo receiver. It does possess some strange effects in both DSB and SSB operation.

-Frank Foti
 
Going to be putting SSB on one of my FM clients using an Omnia.11 late next week. The tower site is low, only 200' HAAT so it is prone to multipath. Will let everyone know the results after I turn it on.

My point about the models wasn't that models are not useful. But a partial model, not an end to end model isn't very useful. It's like looking at one crater and determining the structure of the entire moon from it. Gotta look at the whole picture first and then check each detail, not the other way around.
 
Here was Frank Foti's presentation at NAB about SSB FM Stereo:
http://www.youtube.com/watch?v=eSDg31HmbK0

As demonstrated towards the end, the Sony HD radio does not totally drop to mono, but there is a drastic reduction of stereo separation. Nonetheless, Frank does provide many examples to drive home the point that reducing the transmitted bandwidth does reduce multipath and improve the S/N ratio at the receiver -- but in that case, wouldn't turning off the HD, RDS, and SCA reduce it even further?
 
Status
This thread has been closed due to inactivity. You can create a new thread to discuss this topic.


Back
Top Bottom